Information Systems constant changes and refactoring processes eventually result in design debts, one of them being related to the database management. Redundancy and complexity are often found in databases and eventually affect the overall system performance. In this study, the analysis conducted was based on graph analysis, which is an essential technique in several fields demanding data management. The investigation involved the study of relationships and connections in a real-world financial organization database, between SQL queries represented as nodes and edges in a graph structure. By analyzing the graph structure and properties, it was possible to identify important nodes, detect clusters of related data, and uncover hidden relationships and redundancy. The results indicate that 50% of the database queries had medium to high similarity in subgraphs, which allows the organization to gain valuable insights into their data, make informed decisions, and optimize their database performance.

The use of Educational Robotics as a support tool in the teaching-learning process for children with learning difficulties is the core of our investigation. In this work, we try not to distinguish between (normal) difficulties caused by lack of interest or another solvable problem in the classroom or those caused by well-defined disorders in medicine. In an initial systematic review, in which we did a general search of articles and dissertations in conferences and journals using a well defined set of keywords that covered the most well-known learning problems and that also met the keyword "educational robotics", we identified that there are very few works on this topic, around 14 contributions. Given the novelty of the topic, we carried out a more in-depth analysis of these and other works, including several issues, not only related to learning difficulties. Our main question to be answered, from then on, was to identify how much and how educational robotics can be an adjunct in the teaching-learning process, for these students with difficulties. With this in mind, we defined the quali-quanti research methodology to identify the remaining points, where we adopted questionnaires as a data collection instrument. We provide forms with questions for students, teachers, and monitors from selected schools, among those that have educational robotics activities (all public schools) in the city of Natal, RN, Brazil. Through an analysis of the data obtained from the forms, we observed an increase in the use of robotics in education, in general. However, we mainly identified that, although students with learning difficulties or disorders are found in educational institutions, they are not completely included in robotics activities. Nonetheless, after compiling and analyzing the data, we reached positive results, which verify the possibility of application to improve learning in these children. That is, the majority of students and a good number of teachers and monitors agree that robotics has managed to improve the ability of students with difficulties in learning, especially transversal content. Regarding "how", we noticed that some factors are important. The first factor is the ease of understanding the robotics itself, brought about by new software and hardware tools developed by this pretty new scientific community, which greatly facilitate its use as a support tool. Furthermore, we note that the interest of teachers is another factor in the "how" that can collaborate in trying to improve the issue of learning difficulties. We noticed that interest improves with the use of these new tools, demystifying robotics, which is no longer a big deal for lay teachers. Furthermore, the existence of complete support systems for its application, free of charge and freely available, involving content and tools on the Web, in addition to an effective continuing education policy for teachers, have contributed to the tool becoming increasingly more effective, mainly, as highlighted, in content transversal to robotics.

The spectral efficiency gains introduced by multiuser MIMO systems render them relevant schemes for current and upcoming generations of mobile communication networks. Due to the intrinsic complexity of the mathematical models of these systems under realistic conditions and the interdependence between the processing steps of the transmitters and receivers, machine learning is an option that allows the complete system to be designed by training a noisy autoencoder. This paper proposes a neural network architecture for end-to-end optimization of a multiuser MIMO system. The performance of the system, measured in terms of symbol error rate, was compared to an M-PSK baseline with zero-forcing equalization and least-squares channel estimation. Simulations were performed using a Rayleigh fading channel model and the realistic 3GPP TR 38.901 model. A bidirectional channel estimator, based on the interpolation of sparse pilots, is proposed, reducing the control signaling to less than 3% in exchange for a fixed 10 ms delay. The results show that signicant gains can be achieved by applying the proposed model, but those vary with respect to the estimation errors during the pilot transmission times.

Potholes in urban roads represent a significant issue, impacting both user safety and vehicle durability. This study introduces an innovative approach integrating machine learning on edge devices, with an emphasis on YOLOv8 and FOMO models in the context of TinyML. We utilize a specialized dataset, containing annotated images, to effectively train these models for accurate pothole detection. The focus of the research is on optimizing the performance of these models for devices with limited computational resources, aiming for real-time efficiency and reduced energy consumption. This work not only provides effectively trained models but also introduces an adaptable framework for pothole detection, ensuring practical and efficient implementation. Furthermore, we present a complete pipeline for pothole detection at the end, validating the models' accuracy and efficiency. This approach proposes a robust and energy-efficient solution for the automatic recognition of potholes, significantly contributing to improvements in urban infrastructure maintenance.

Amid increasing advances in the detection capability of modern radar systems, stealth
technology aims to reduce the radar cross section (RCS) of a platform, making it difficult
to detect. Among the principles on which stealth technology is based, active systems
have stood out due to the various advantages they present. Due to their spatial filtering
characteristics, frequency-selective surfaces (FSSs) have been widely explored as
active systems. Considering the strong out-of-passband reflections in various direction,
bandpass FSSs can significantly reduce monostatic RCS, however they can cause large
bistatic RCS. In this context, a new structure, called frequency-selective radome absorber
or frequency-selective rasorber (FSR), has attracted the attention of researchers in recent
years. This type of structure combines the design of a bandpass FSS with a resistive
absorber FSS, to obtain two absorption bands (A) located on both sides of a transmission
band (T), also known as absorption-transmission-absorption (A-T-A). Therefore, it
is capable of reducing out-of-passband reflections in other directions. In this work, two
dual-polarization A-T-A FSRs based on interdigital resonators (IR) are proposed, whose
lossy FSS designs feature elements with four incorporated resistors and different layouts;
split-ring, crossed dipole, and square frame. The two rasorbers presented provide good
operational bandwidth, and have passbands around 9.7 GHz and 4.78 GHz, whose minimum
insertion losses (IL), under normal incidence, are 1.0 dB and 0.77 dB, respectively.
equivalent circuit models (ECM) for the proposed FSRs are presented with the aim of
explaining the parallel resonance in each case. The rasorber prototypes are fabricated
with low-cost FR-4 substrate in both layers, and measured for demonstration. The results
are very coherent when compared with those obtained in simulations. Due to their
symmetries, the presented FSRs have good angular stability. Therefore, it appears that
the proposed rasorbers are capable of reducing the RCS of communication systems operating
in different bands, contributing to making them stealthy. Finally, performance
comparisons with other FSRs implemented with parallel resonance in previous works are
presented.

Systems involving multiple Unmanned Aerial Vehicles (Multi-UAVs) performing area scanning procedures have garnered attention in recent years. Developing these systems faces significant challenges, with the robustness of the data communication network being a crucial aspect for efficient collaboration among UAVs. This work is dedicated to specifying a data transport protocol for a communication network within a squadron of Unmanned Aerial Vehicles, focusing on scanning the rocket impact area over the sea in the SPACEVANT 2 project.

The research encompasses the key characteristics of communication networks for multi-UAV systems, considering application-specific aspects such as path and trajectory planning and sensor data collection. Strategies are proposed for the rocket impact area scanning, analyzing characteristics of networks best suited for these strategies, and defining a suitable protocol for the application. Utilizing Xbee Pro 900HP S3B modules in a network architecture integrated with a computer-controlled embedded hardware platform, a test plan is devised and implemented to assess the protocol's performance. Performance evaluations include measurements of bandwidth (throughput), latency, and indicators on communication links among network nodes, such as retransmission, packet loss, flow control, and congestion. This study demonstrates the robustness and reliability of the protocol in the proposed network architecture for multi-UAV systems.

We propose the creation of a set of independent variables, derived from the analysis of phenomena, government interventions and events that may influence the spread of the SARS-CoV-2 virus, whether positively or negatively in relation to the trend of the curve of cases and deaths. This approach aims to identify relevant characteristics and parameters applicable to these scenarios, using regression techniques. To achieve this, we carried out studies using machine learning methodologies to determine which variables should be selected and applied to the regression models. Our strategy included collecting and cleaning data, evaluating the generalization of models, and applying machine learning techniques, such as regression using Seasonal ARIMA, clustering, dimensionality reduction through principal component analysis and neural networks, in addition to creating informative data visualizations. As a result, we compiled and proposed the use of a set of independent variables to predict the number of deaths from COVID-19, aiming to increase accuracy and reduce the standard deviation in relation to real values, and avoid underspecification of the problem. The main contributions of this work include the investigation of the causes and possible relationships between phenomena, government measures and events in the spread of the virus in urban areas, such as cities and countries. The results of this study can serve as a complementary resource to assist governments and authorities in making decisions in the face of possible future pandemic scenarios, such as the one faced during the COVID-19 crisis, which, although still present, has a lower mortality rate.

This work proposes an integrated antenna with frequency selective surface (FSS) with textile substrate for applications in medical continuous monitoring systems with reduced specific absorption rate (SAR). The antenna and FSS were fabricated on cotton and polyester textile substrates, used in textile healthcare products, based on the technical standard ABNT NBR 13734:2016 - Textile healthcare products, for operation in the ISM 2.4 GHz frequency band. The system operates at a frequency of 2.4 GHz. Three textile materials with varying proportions of cotton and polyester were chosen to design the antennas and FSS. The electrical characterization of these materials was carried out. The antenna is designed for a filled ground plane and a rectangular patch fed by a microstrip line. The FSS was designed using square loops. The experimental operation of the frequency selective surfaces within the range of interest was observed through the transmission coefficient as a function of frequency. Phantoms were fabricated to emulate human tissues skin, fat, blood and muscle, and a thermal analysis was carried out to evaluate the impact of continuous exposure on signal transmission from the antenna integrated into the fabricated phantoms. The results demonstrate a reduction in SAR due to the effective rejection of the ISM band signal by the FSS of around 66.67% compared to the system with the active antenna without the presence of the FSS.

The Internet of Things (IoT) is a paradigm where computing and connectivity capabilities are embedded into objects, connecting them to the Internet. Acknowledged as a crucial and emerging technological area, IoT holds significant potential to enhance quality of life, optimize industrial processes, and offer more applications to everyday objects. With the increasing number of IoT-connected devices, there arises a need for infrastructure to manage the vast volume of generated data. In this context, Edge Computing stands out by processing data close to its source, leaving only heavier processing tasks for central servers. Edge processing enables the development of optimized machine learning algorithms, known as Tiny Machine Learning (TinyML). By employing lightweight and optimized algorithms, TinyML offers advantages such as reduced latency, improved energy efficiency, and increased autonomy for devices operating in remote or isolated applications. In the field of TinyML, implementing machine learning techniques on resource-constrained devices like microcontrollers poses significant challenges, including outlier detection and correction. This work addresses the outlier problem, crucial for both academic research and industrial applications in domains such as energy measurements, health data, industrial systems, and automotive applications. Three algorithms were developed: TEDA-RLS and TEDA-Forecasting for outlier processing, focusing on detection and correction, and TEDA-Ensemble for forecasting. The developed algorithms are based on the concept of Incremental Learning, as they have the capability to continuously learn from new data inputs. The proposed algorithms were compared with established techniques in the literature such as XGBoosting, Long Short-Term Memory (LSTM), Linear Regression, and k-Nearest Neighbors (KNN), yielding promising results with low error rates and minimal energy consumption. Finally, the outlier processing algorithms were successfully embedded in a microcontroller, confirming the feasibility of the TinyML approach.

The global SARS-CoV-2 pandemics compelled governments, institutions, and researchers to assess its impact and develop strategies based on general indicators to achieve the most accurate predictions possible, in order to help managers mitigating its effect. While known epidemiological models were naturally used, they often produced uncertain forecasts due to insufficient or missing data. In addition to data limitation, various machine-learning models such as random forests, support vector regression, LSTM, auto-encoders, and traditional time-series models like Prophet and ARIMA—were employed, yielding impressive yet somewhat limited results. Some of these methods struggle with precision when handling multi-variable inputs, which are crucial for problems like pandemics time series prediction that require both short- and long-term forecasting. In response to this challenge, we propose a novel approach for time-series prediction that utilizes a stacked auto-encoder structure. Our model uses $n$ internal autoencoders to process the input and generate different latent spaces for this respective input. Then these different latent spaces are concatenated and the expanded latent space is obtained. We conducted an experiment using previously published data series on COVID-19 cases, deaths, temperature, humidity, and the air quality index (AQI) in São Paulo City, Brazil. This experiment assessed the suitability of our model for short-, medium-, and long-term forecasting. Furthermore, we directly compared our proposed model with two existing works in the literature that have already undergone expert scrutiny. The first comparison places our model among those that use one network for feature extraction and another for predicting the pandemic trends. The second comparison highlights our model's effectiveness in multi-series forecasting of pandemic indicators. The results suggest that our proposed model possesses strong capabilities in both feature extraction and multi-series forecasting, offering improvements over the two comparison works. Finally, the model demonstrates promising forecasting accuracy and versatility across datasets of varying lengths, making it a standout option for time-series forecasting tasks.

This work aims to develop a non-destructive system for measuring humidity using electromagnetic waves in concrete blocks used in masonry. This system applies the Structural Health Monitoring (SHM) technique to the blocks analyzed. Changes were proposed in a microwave system for measuring relative humidity of hollow concrete blocks presented in the literature, aiming to improve the sensitivity of the system. The operating frequency range was changed from 2.5 GHz to 5.8 GHz. This reduced the dimensions of the antenna arrays used in the system. Five block samples were used in this work. The frequency selective surface (FSS) used as a sensor was modified from a double circular ring to a square loop, which presents a single-band frequency response. The results obtained in this work showed a high correlation between the frequency variation and the relative humidity level. First, second and third order functions were compared to see which one would best relate the resonance frequency to the relative humidity. The second order function presented the best results, considering the Mean Absolute Percentage Error (MAPE), the Root Mean Square Error (RMS) and the coefficient of determination. A sensitivity analysis of the sensor was performed showing that the new proposed system is more sensitive than others presented in the literature.

In this research, a new method for face representation is presented, which utilizes Bidirectional Generative Adversarial Networks (BiGANs), showing significant progress compared to conventional methods of video transmission using MPEG-2 compression techniques. In scenarios such as online meetings, our approach takes advantage of the inherent bidirectional capabilities of BiGANs in virtual environments to produce compact yet highly expressive facial representations. As a result, the amount of data required for transmission is reduced. The effectiveness of our approach in generating high-quality synthetic face images that closely resemble the original faces was demonstrated through our experiments, which were conducted on a dataset consisting of 813 frames of an individual's face. Furthermore, the method's capability to preserve higher values of the Structural Similarity Index (SSIM) and Peak Signal-to-Noise Ratio (PSNR) highlights its potential to generate synthetic facial images with minimal degradation in quality. This makes it an encouraging approach for real-time online communication, especially in situations with limited bandwidth.

High-performance computing, a dynamic field within computer science, provides the processing power necessary for algorithms across diverse domains. Large-scale supercomputers are indispensable for tackling complex problems; however, their size and complexity make them susceptible to failure. This underscores the criticality of employing fault tolerance techniques to mitigate the impact of interruptions or failures. These methods are instrumental in addressing hardware and software malfunctions and preemptive scenarios such as resource reclamation by cloud providers.

Given the imperative for fault tolerance, we present the Dependability Library for Iterative Applications. This library offers a versatile solution for bulk synchronous programs. The proposed library simplifies the integration of fault tolerance capabilities into the applications, offering high configurability options and allowing users to select which functionalities they want to utilize in their applications. The principle is to reduce efforts in implementing fault tolerance approaches, allowing project groups to focus on developing their specific problem.

The library offers a range of fault tolerance features, including checkpointing, replication, and heartbeat monitoring. Checkpointing saves the application's state at intervals, allowing it to resume from the last saved point after a failure. Replication ensures reliability by allowing a backup unit to take over in case of failure. The library has detected possible failures using the heartbeat monitoring method and potential resource reclamation. The proposed library is compatible with user-level failure mitigation, which allows programs to continue operating after crashes, minimizing downtime and ensuring continuous operation.

Our proposal was successfully applied to the geophysical problem of full-waveform inversion, a standard algorithm for oil and gas exploration geophysics processing. This application serves as a high-performance practical scenario for analysis. All features were rigorously validated, and the overhead in this problem was thoroughly analyzed using more realistic examples. In our experiments, the application did not lose all data processed until the failure moment, and it could continue execution even in the presence of node failure, with minimal overhead. This work also shows other case studies in the initial stage of applying the library and discusses some fault tolerance concepts and related works.

In Software Defined Radio (SDR) most analog functions such as downconversion and channel selection are moved into the digital domain. With these features, an SDR can be easily programmed and adapted to new standards. However, its implementation using a Nyquist rate analog-to-digital converter (ADC) has been challenging and very power-hungry, which makes it unsuitable for mobile applications. SDM (Sigma-Delta Modulator) using LC resonators are well suited for this purpose, as they can convert a narrowband around a radio frequency (RF) center frequency. In the literature, high-order LC-based SDM uses nodes between LC tanks for feedforward or feedback coefficients, to obtain the desired noise transfer function (NTF). These nodes are often an additional source of noise, non-linearity, and power consumption, affecting overall SDM performance. In this project, we aim to explore the functionality of N-Path filters to implement a programmable Sigma-Delta Bandpass Modulator that meets various communication standards. Therefore, it was necessary to carry out a bibliographic and technological research on the state of the art of N-path filters, continuous-time and discrete-time Sigma-Delta modulators. There was also the development of high-level models for different topology categories of Sigma-Delta Modulators implemented with LC and discrete-time filters to serve as comparison models with the topology design methodology using the N-path filter proposed in this work. A high-level model of the N-Path filter was developed, and the work is currently in the stage of integrating the N-path filter and SDM. The results obtained from the SDM implementation with LC filters show that the modulator has a good agreement with its DT equivalent. In initial experiments, the implementation of SDM with an RLC circuit equivalent to the N-path filter proved to be promising, presenting results similar to those obtained with the discrete-time topology.

The area of Neurophysiology of Social Behaviors in Rodents has the general objective of investigating the electrophysiological correlates of social behaviors in rodents. Examples of studies in the area include the investigation of neural processing disorders and synaptic plasticity in animal models of autism and changes in synaptic plasticity in animal models of epilepsy. This area uses sophisticated and expensive equipment, and some of the existing platforms are proprietary brands, obsolete, and using old technology. There is a prominent need to improve this equipment to aid research, seek cheaper alternatives, with open construction, and use more up-to-date components. With that in mind, in this work, we propose improvements in the software platform that receives data and controls the necessary operations in the experiments using the behavioral box. Specifically, we contribute on three fronts. At first, we developed software aimed at automating a behavioral box in a tactile discrimination task, which, in its new version, uses the Arduino microcontroller. Also, we developed the software part of the Supervisory System, in Python, which communicates with Arduino, for visualization and definition of the tasks to be carried out in the behavioral box. Finally, we developed basic computer vision routines for locating the rodent within the behavioral apparatus, indicating its position and orientation. These developments collaborate so that the box can be automated, reducing manual interaction with the experimenter, and thus contributing to improving the process of studying tactile discrimination in rodents.

This work aimed to develop a real-time test platform for systems associated with the tactile internet area. The proposal comprises a master device, a communication channel and a slave device. The master device is a tactile glove (wearable technology) that works as a tactile interface based on vibratory feedback, responsible for enabling interaction with virtual or remote elements. Additionally, the communication channel introduces a bidirectional connection of variable latency. In turn, the default slave device is a robotic Phantom Omni manipulator integrated with the Matlab/Simulink environment and the robotics toolbox. The platform can generate tactile sensations such as coarse roughness, fine roughness, smoothness, dripping and softness. Furthermore, the results of the proposal have been adequate and allow testing various algorithms and methods related to tactile internet.  Therefore, the developed system has the capacity to serve as a basis for a wide variety of use cases related to tactile internet, providing a diverse opportunity for exploration and application. In this sense, the proposed platform serves as a foundation for a use case focused on tactile exploration of product textures on an e-commerce website, enabling interaction with virtual clothing through online shopping capable of providing tactile stimuli. In the scenario of this use case, the slave device is a cursor that functions as an element controlled by the master device and a mouse, which allows interaction with different clothing textures, generating specific tactile sensations on the tactile glove. In order to enhance the platform performance, Virtual Network Functions (VNFs) were introduced. VNFs are virtualized processes implemented and executed in cloud services, capable of enhancing the scalability and agility of a network, while also enabling more efficient utilization of infrastructure resources. In Fifth Generation (5G) mobile networks, tactile internet connections are associated with ultra-reliable low-latency communications (URLLC) services. This category poses challenges due to its stringent requirements of low latency in a few milliseconds, low packet loss probability, and high network availability. Thus, the application of VNFs during tactile internet connections in 5G networks allows for adjustments of connection parameters to maintain good connectivity, ensuring the low-latency criterion and a high degree of realism in tactile interactions. Consequently, the use of VNFs in URLLC communications aims to fulfill the requirements of tactile internet systems while maintaining an acceptable level of Quality of Experience (QoE), which is a metric defined by the Telecommunication Standardization Sector as the degree of satisfaction or annoyance of users with an application or service.

Solar trackers, besides serving as support for photovoltaic modules, follow the move- ment of the sun throughout the day and try to keep the plane of the photovoltaic modules perpendicular to the sun’s rays, thereby increasing the performance of the photovoltaic module. With the aim of demonstrating the effectiveness of using the automatic orien- tation method of photovoltaic solar modules in following the seasonal movement of the sun as a way of improving energy production, this work analyzes two systems: a solar tracker and a fixed-axis system. The first part simulates two photovoltaic plants with solar tracking and fixed-axis systems using the software PVsyst, consisting of a total of twenty modules. The results show that the solar tracker improved the efficiency of electricity production, with a gain of 25.97% during the day and an annual gain of 25.6% more than a fixed system. The second part presents a proposal for an automatic solar tracking system (with a controller) and a fixed-axis system (without a controller) based on the equations of the sun’s movement angles. This was modeled using the software Matlab/Simulink using Fuzzy logic as the controller, with the aim of evaluating the use of solar radiation throughout the day and to build a real system with the same characteristics as the pro- posed simulation. The gain of this system was approximately 17% to 35% during some hours of the day for the system with a controller. Both software simulations showed that photovoltaic systems that track the movement of the sun during the day and year are more advantageous compared to fixed-axis systems because they take better advantage of solar radiation.

It is essential to monitor the levels of Non-Ionizing Radiation (NIR) that the general population may be exposed to and compare them with the limits defined in current regulations, bearing in mind the rapid rise of telecommunication services and the perspectives of an extremely connected. Indoor environments (\textit{indoor}), such as residences and shopping malls, are places that meet the interests of measurements, mainly due to the presence of different NIR sources and the location of these sources about end users. This work presents NIR measurements in a \textit{shopping center} and residences in Natal, RN, Brazil. At shopping, a set of six measurement points was proposed, following two criteria: places with a large flow of people and the presence of one or more Distributed Antenna System (DAS), co-located or not with Wi-Fi access points. The results are presented and discussed in terms of distance from the DAS (conditions: near and far) and flow density of people in the mall (scenarios: low and high number of people). The highest mean and highest peak electric fields measured were 1.96 and 3.26~V/m, corresponding respectively to 5\% and 8\% of the limits defined by the International Commission on Non-Ionizing Radiation Protection (ICNIRP) and Agency National Telecommunications Agency (ANATEL). In the homes, measurements were taken at 40 Wi-Fi access points where the highest exposure level was 4.66 V/m (7.64\% of the limit), verified for the network load on the network of 2.4 GHz (Situation 1), at a value of 4.66 V/m (7.64\% of the limit). The results are discussed regarding the measurement situation, brand and time of use of the equipment. Measurements were also carried out in 51 microwave ovens. Only one exceeded 50~W/m² (limit determined in current regulations), caused by rust that compromised the door's structure. Thus, two types of repairs were carried out: the use of epoxy putty to fill the opening in the door (Repair 1) and the replacement of the outer surface of the oven (Repair 2). Only Repair 1 solved the leakage problem with an average of 0.1841 W/m² and a peak of 0.4222 W/m². As for the other 50 microwave ovens, the relation between power density and factors such as worst point, brand, time of use and state of conservation of the device were analyzed.

This work aims to propose and validate an approach for the generation and visualization of association rules and sequence rules obtained from the content access history data of a Brazilian journal. The proposed approach is composed of four phases: exploratory data analysis (EDA), data preprocessing, generation of association and sequence rules, and visualization of results. The algorithmsAprioriand FP-Growth were used to generate the association rules. To generate sequence rules, the algorithm used was SPADE. Parallel coordinate graphs were used to visualize the association rules and graphs for visualization of sequence rules. An outstanding aspect of the proposed approach is the visualization of the rules obtained using graphic resources to enhance the analysis of the results in support of business decisions and contribute to mapping the users’ access profile. The proposal was validated by using data from user access to a digital news portal.

The amount of data in real-time, such as time series and streaming data, continues to grow. Analysing this data the moment it arrives can bring immense added value. However, it also requires much computational effort and new acceleration techniques. As a possible solution to this problem, we propose a hardware architecture for Typicality and Eccentricity Data Analytic (TEDA) algorithm implemented on Field Programmable Gate Arrays (FPGA) for data streaming. TEDA is based on a new approach to outlier detection in the data stream context. The suggested design has a full parallel input of N elements and a 3-stage pipelined architecture to reduce the critical path and thus optimise the throughput. To validate the proposals, results of the occupation, throughput and power efficiency of the proposed hardware are presented. Compared to other software platforms, the design reached a speed of up to 693x, with a throughput of up to 10.96 MSPs (Mega Sample Per second) with a dynamic power of 16mW. Besides, the bit-accurate simulation results are also presented for different application scenarios with multiple sensors ranging from applications in Industry 4.0 environments to the Internet of Medical Things (IoMT). This work is a pioneer in the hardware implementation of the TEDA technique in specialised hardware.

The use of Corrective Switching technique to eliminate overloads in electrical systems, addressed in several studies in the last decades, has proved to be a very effective load flow control technique. Some of these studies are based on electrical circuit theories or Relief Function methodologies. Network switching causes changes to the system configuration and consequently the short-circuit levels of the buses are changed. These changes can lead to a network insecure operation when the protection parameters are not modified, since the overcurrent protection functions are based on the values of the rated currents and the short-circuit currents, which are previously defined in the protection studies. Adaptive Protection techniques makes it possible to recalculate the network protection parameters in real time, for any system configuration, enabling network operation with optimized protection settings. The purpose of this workis to implement the Corrective Switching technique, using, as a restriction condition in the choice of switching variants, the adjustments of the network protection parameters. If the network operation is not safe for the configuration proposed by the relief variants, new network protection parameters are calculated, as well as coordination and selectivity, is adopted and the Adaptative Protection technique is implemented. The network is automatically reconfigured through the Self Healing technique. 

This work proposes the development of one-terminal traveling wave-based transmission line
protection methods for HVAC (High Voltage Alternating Current) and HVDC (High Voltage
Direct Current) systems. The effect of the sampling frequency on the protection is considered,
which until now has not yet been investigated for one-terminal methods. Inaccuracies on the
wave velocity estimations, which classically yields to problems on methods based on traveling
waves, are addressed and solutions to such problem are presented. The proposed methods have
been evaluated by means of computer simulations. The proposed earth fault distance protection
for HVAC transmission lines was evaluated using a real commercial relay with time domain
based protection functions. The results related to the proposed earth fault distance protection
for HVAC transmission lines show that the proposed function, when associated with other existing
protection functions, can achieve a quite remarkable enhancement for the dependability and
velocity of the transmission line protection. The results concerning the method based on traveling
wave reflections show that it is possible to protect most point-to-point transmission lines in
a fast way, without the need for communication, and independent of the estimation of the wave
speed. Finally, the results concerning distance protection for meshed HVDC systems demonstrate
the applicability of the proposed method for such systems. The method showed operating
time below 2 ms for a line of 500 km length. This operation time meets the likely requirements
that HVDC meshed systems will present. In addition, the method presented selectivity for a
4-terminal system interconnected by 5 transmission lines.

The scanning area procedure of the probable impact region is a mandatory step in the pre-launch 

phase of any rocket. The objective is restrict the presence of unauthorized vessels, avoiding 

possible accidents and, consequently, preserv human lives. Considering the high operational cost 

of this procedure, normally performed by a P-95 aircraft, the SpaceVANT II project proposed the 

execution of the area scan by a fleet of Unmanned Aerial Vehicles (UAVs) that, acting in a 

cooperative way, have the ability to execute the mission with lower cost and operational complexity. 

In this context, it is essential to create a communication network able to ensure the exchange of 

information between the aircraft that make up the system. One of the most used alternatives to 

provide this communication network is mobile ad hoc network (MANET) protocols adapted for 

flying ad hoc networks (FANET). However, managing the dynamic behavior of this type of network 

in order to minimize disconnections and enable the necessary conditions for the operation of 

applications on board aircraft is not simple and has found limitations in the solutions adopted. In 

this perspective, this thesis proposed a management based Software Defined Networking (SDN) 

called SD-FANET. Design on a distributed and hierarchical architecture, the SD-FANET control 

plane allows centralized management and assigns the controller the ability to program the 

switches, changing the data forwarding from a holistic view of the network. The SDN approach 

allows the use of different path selection algorithms, which adds more flexibility and improvement 

to the forwarding process. To validate the proposed solution, experiments were carried out in the 

emulated environment provided by Mininet-Wifi and a proof of concept based on a prototype using 

a communication protocol developed to enable transmission of images in ZigBee communication 

links embedded in DJI Phantom 3 Standard drones. The results showed that the throughput 

obtained by SD-FANET was 22.2% better than MANET protocols and 5.4% in relation to the local 

SDN controller. The average recovery time provides by resilience mechanisms was about 1.5 s.

In oil fields with many wells operating by Sucker-rod pumping, due to the lack of early diagnosis of operating conditions or sensor faults, several problems can go unnoticed. These problems can generate large losses, such as increased downtime, increased OPEX (Operational Expenditure), decreased efficiency and lost production. In practice, the identification and diagnosis of operating conditions are carried out from surface and downhole dynamometric cards, using pre-established standards, with human visual effort in the operation centers. This task requires a lot of time and work, in addition to requiring experience, as it can be influenced by subjective factors. However, in recent years, with the facilities inherent to Machine Learning (ML) algorithms, several researches on the subject have achieved good results in the diagnosis of operating conditions, showing that ML can be used for this purpose. However, it is still common to have doubts about the difficulty level of the dynamometer card classification task, the best algorithm, the best shape descriptor, the best metrics and what is the impact of the imbalanced datasets. In the search for answers to these questions, this work used real data from 38 wells operating by sucker-rod pumping in the region of Mossoró, RN, Brazil. More than 50,000 cards have been classified by specialists and distributed in eight modes of operation and two common sensor faults in this field. Sixty tests were performed and divided into seven groups. Three algorithms were tested (with and without hyperparameter tuning): Decision Tree, Random Forest and XgBoost, in addition to pipelines provided by Automated Machine Learning (AutoML). The descriptors used were: Fourier descriptors and Wavelet descriptors, in addition to the load values of the downhole dynamometric card. Balanced and imbalanced datasets were also tested. The results confirm the feasibility of applying ML for diagnosis of operating conditions and sensor faults in sucker-rod pumping systems, since 75% of the tests reached accuracy above 92% and a maximum accuracy was 99.84%.

The manufacturing processes advancement in silicon wafers has enabled integration between electronics and photonics. Data will no longer be processed with electrons for the technology sectors but with photons, providing higher processing and transmission rates. In this context, the health and well-being sectors, for example, will have new equipment, allowing remote and individual monitoring and diagnosis, enhancing the prevention and treatment of diseases. Thus, this research proposes a photonic device with graphene applied and a surface plasmon resonance structure for sensing applications. Initially, they are concepts related to photonic devices and their different applications. Photonic crystals, plasmonic devices, graphene, and their use in optical guides are also presented. Then, the methodology used in the research development is explained, dealing with modeling and simulation. The results are presented and analyzed as the structure’s spectral behavior for different glycerol concentrations, which reinforces the potential for point-of-care applications considering its compactness and simplicity of construction.

This work is an extension of previous studies and presents the research and development of a pulse width programmable gain integrating amplifier for chip systems. One of the main advantages of the proposed PGA architecture is that it uses few components and the number of gain values is independent of the circuit area. In addition, the architecture can be configured to operate with single-ended or differential signals.

Before starting the Integrated Circuit (IC) development, equations defining the uncertainties of the circuit were obtained and a discrete circuit was designed, simulated, and developed, which went through a series of experiments to validate the proposed architecture.

The implementation of the architecture in an IC aims to achieve higher operating frequencies, allowing its use in a wider range of applications. The integrated circuit was designed using the TSMC 0.18 µm CMOS technology, presenting an average power dissipation of 123.2 µW and occupying an area of approximately 0.065 mm2.

Transient simulations were performed with and without noise, DC, AC, PSS, PAC, Corner, and Monte Carlo, evaluating means, uncertainties, relative and absolute errors related to various parameters. This suggests that the circuit was carefully designed and tested, and the presented results provide a solid basis for the application of the proposed PGA in a wide range of analog conditioning applications.

In this work, a numerical method for the computation of observer-based output feedback controllers is proposed for fuzzy Takagi-Sugeno (T-S) systems subject to constraints, based on set-invariance theory. Positively Invariant (PI) polyhedral sets are used to ensure that state and control constraints are satisfied at all times. Sufficient conditions are established for a polyhedron defined in the augmented state space (state + estimation error) to be PI. From the invariance conditions, a bilinear optimization problem is formulated to simultaneously calculate the controller and observer gains and the positively invariant polyhedron that guarantee the satisfaction of the constraints. The two types of observers found in the literature of fuzzy T-S systems are considered: the first considers the membership functions dependent only on the system output, while the second refers to the general case, where these functions can be associated with any state variable. In the simplest case, although the membership functions depend only on the output, the estimated state feedback results, in general, in controllers with better performance and with larger sets of admissible states associated with them than the output static feedback control. For the general case, as membership functions depend on non-accessible states, an estimation mechanism is needed to calculate these variables. In both cases, this role is played by the fuzzy observer T-S. The problem of tracking a constant reference signal is also considered, for which the concept of robust positive invariance is used in conjunction with an Integral-Proportional (I-P) controller. Sufficient conditions are established for a polyhedron defined in the augmented state space (state + estimation error + tracking error integral) to be PI in the presence of a constant reference signal. Several numerical experiments illustrate the effectiveness of the proposed approach.

Funding agencies and research institutions often employ quantitative methods and scientometric techniques to evaluate scientific groups. These evaluations typically rely on a single type of metric, whether it is based on counts (such as the h-index) or derived from Complex Network Analysis. However, the use of multiple measurement approaches and the proper exploration of the temporal dimension of academic production is still a recurrent issue. Particularly, an underexplored approach involves combining these indicators with Machine Learning and Graph Embedding techniques, which could enhance the evaluation process of research groups. In this context, this study proposes a Graph Data Science-oriented methodology to analyze scientific teams over time. Through a case study, the results suggest the feasibility and suitability of the proposed method for quantitatively assessing research groups. The presented approach has the potential to provide strategic and proactive insights for scientific teams, contributing to a better understanding of their dynamics and limitations.

Amyotrophic Lateral Sclerosis (ALS) is a neurodegenerative disease that irreversibly impairs an individual's motor system and functional abilities, even causing the progressive loss of communication skills and autonomy. Technological resources based on Augmentative and Alternative Communication, Computer Vision, and Machine Learning are essential for developing digital health solutions to enable the communicative process and autonomy that, consequently, promote improvements in the quality of life and survival of people with ALS. Focused on a Human-Computer Interaction (HCI) approach based on images of the eyes from a simple camera not mounted on the body, this work presents an assistive technology resource for Augmentative and Alternative Communication for people with ALS. The approach proposed in this work consists of an algorithmic model capable of recognizing the state of the eye (open or closed) in real-time and interoperating with Autonomus, a digital health solution designed by the Laboratory of Technological Innovation in Health at the Federal University of Rio Grande do Norte (LAIS/UFRN) for the communication of people with ALS. The model consists of four methodological processes: (i) image acquisition; (ii) Face detection; (iii) eye detection; and (iv) classifying the state of the eye, which is the foremost step for Human-Computer Interaction. An algorithmic study with a control group was conducted to evaluate the model's overall performance and the Convolutional Neural Network (CNN) classification ability. The results related to the proposed model for classifying the state of the eye in real-time are promising and reach significant values of accuracy and f1-score above 92%. The results also point to the viability of developing low-cost assistive technology resources that guarantee universal access, health promotion, well-being, and reduced inequalities, which go beyond improvements in the communicative process of people with ALS. Therefore, the object of study of this work is also to enable and promotes the exercise of rights, citizenship, fundamental freedoms, and health care for people with ALS.

Osteoporosis is a silent and still underdiagnosed condition, with a mortality rate higher than several types of cancer, especially when patients suffer fractures. The gold standard equipment for the diagnosis of densitometry, Dual-energy X-ray absorptiometry (DXA, or DEXA), which requires an invasive and costly procedure, is scarce in countries considered middle or low-income, thus hindering timely access to diagnosis. In this context, a portable device, known in the literature as Osseus, was developed for the screening of patients who need the densitometry exam, i.e., to qualify the referrals of exams to the DEXA equipment. The thesis aimed to validate the Osseus device using machine learning techniques. For this, the planning and data collection of 505 patients who underwent the DEXA exam and the test on the Osseus device were carried out, of which 21.8% were healthy and 78.2% were diseased (they had low bone mineral density or osteoporosis). Therefore, to implement the studies and develop the research, the database was separated into 80% for training and validation (5-fold cross-validation) and 20% for testing. The performance obtained in the test base with the best model (Random Forest) corresponded to sensitivity=0.853, specificity=0.871, and F1(harmonic average of precision and recall rate)=0.859. The results showed that the most relevant variables to indicate the individual health status were age, body mass index (BMI), and the attenuation measured in the Osseus. When compared to the results of DEXA scans, the model has proven to be effective and consistent in screening individuals with osteoporosis and facilitating early diagnosis of the disease, which consequently entails improved productivity and reduced costs for surgery, treatment, and hospitalization. This way, by qualifying the referral of patients from primary care to the specialized network, the Osseus can improve the osteoporosis care network and consolidate itself as a resource of easy access in primary care, also impacting the reduction of waiting lines in the specialized network of the BrazilianNational Health System (SUS).

Since December 2019, the COVID-19 pandemic caused by the SARS-CoV-2 virus has intensely affected the world. In the case of a novel virus identification, the early elucidation of taxonomic classification and origin of the virus genomic sequence is essential for strategic planning, containment, and treatments. Deep learning techniques have been successfully used in many viral classification problems associated with viral infection diagnosis, metagenomics, phylogenetics, and analysis. Considering that motivation, this work proposed an efficient viral genome classifier for SARS-CoV-2 using the deep neural network based on the stacked sparse autoencoder (SSAE). For the best performance of the model, we explored the utilization of image representations of the complete genome sequences as the SSAE input to provide a classification of the SARS-CoV-2. For that, two datasets were explored: based on k-mers image representation and based on CGR image representation. The dataset based on k-mers image representation was applied in the experiments of different levels of taxonomic classification of the SARS-CoV-2 virus, and the dataset based on CGR images was applied to the experiments of classification of SARS-CoV-2 variants of concern (VOC). For the experiments of taxonomy classification, the SSAE technique provided great performance results, achieving classification accuracy between 92% and 100% for the validation set and between 98.9% and 100% when the SARS-CoV-2 samples were applied for the test set. These results indicate that our model can be adapted to classify other emerging viruses. For the experiments of SARS-CoV-2 variants classification using CGR images, the SSAE technique provided even better results, achieving classification accuracy of 99.9% for the validation set and 99.8% for the test set. Finally, the results indicated the applicability of this deep learning technique in genome classification problems.

Dual connectivity (DC) technology (or, more generally, multiple connectivity (MC)) is an important feature in the initial journey of New Radio (NR), the access network of 3GPP to 5G, being the basis for the first ways of implementing this generation of communication systems. With MC, users can be simultaneously connected to legacy technologies (4G-LTE and Wi-Fi) and the new 5G-NR technology. However, such technology poses additional challenges for the network, especially with regard to energy management and the numerous connectivity configuration options. This work aims to conceive strategies based on machine learning capable of efficiently exploring the performance of MC towards 5G eMBB (enhanced Mobile Broadband) and 5G URLLC (Ultra Reliable Low Latency Communications) in heterogeneous networks. The idea is to dynamically configure the best set of MC parameters to provide increased data throughput (eMBB target) and a more robust connection (URLLC target), being relevant in scenarios with the presence of small cells using mmWaves, for example, both considered essential to meet 5G specifications.

In this work, waveguides based on directing light through media with sequentially refractive indices in the direction of propagation are proposed. The core guiding model of these structures is based on quasi-periodic sequencing of the refractive index that forms the signal propagation region. In this context, well-known optical fiber models are modified to adapt to the segmentation structures of optical signal guiding regions.

These modifications enable the development of new optical device models, such as lasers, filters, and optical sensors. One of the most traditional segmentation models applied in optical waveguides is the Bragg Grating, also known as Fiber Bragg Grating (FBG), where the sequencing in the propagation region follows a periodic pattern. More recently, some quasi-periodic sequence models have been applied in photonic crystals, primarily considering the cross-sectional area of the waveguide. In this context, a new segmentation strategy is proposed, based on quasi-periodic sequencing of the refractive index along the direction of propagation.

Structures of core fractionation in waveguides will be investigated by applying quasi-periodic models, such as Fibonacci sequences, Thue Morse, Period Doubling, and Octonacci, with the aim of comparing their performance with waveguides that use Bragg Grating as a segmentation strategy. For this study, a mathematical formulation based on the finite element method, in conjunction with the optical beam propagation method, will be used as a simulation tool. Parameters of transmission and reflection at the interfaces between the different media composing the applied sequences will be analyzed.

Due to their great efficiency, security and flexibility, mobile robots are being increasingly used in industry. However, their positioning control is a great challenge due to the non-linear nature of this plant and the difficulty of estimating certain parameters, for example, the friction effects. In this work, non-linear controllers are applied to the trajectory control of an omnidirectional robot under the effect of unmodeled dynamics. The control approaches used in this work were both non-linear control strategies, Feedback Linearization (FBL) and Sliding Modes (SMC) both incorporated with an intelligent compensator utilizing Recurrent Neural Networks in order to assist the control by estimating uncertainties. The chosen architecture of the neural network was based in the need to compensate more complex dynamics and at the same time the restriction of computational complexity so that it could be embedded in the hardware of a mobile robot. The stability properties were proven by the principle of assintotic stability proposed by Lyapunov and the performance of the strategies were verifed through both simulations and experiments using Robotino, an omnidirectional mobile robot produced by Festo Didatics and a performance gain was observed when compared with the neural network without the recurrence.

Preterm birth (PTB) is a phenomenon that brings risks and challenges for the survival of the newborn child. Despite many advances in research, not all the causes of PTB are yet clear. It is currently understood that PTB risk is multi-factorial and may also be associated with socioeconomic factors. In order to analyse this possible relationship, this work seeks to stratify PTB risk in Brazil using only socioeconomic data, extracting and analysing those clusters that present relevant PTB divergence, all of which will be found by automatic clustering processes using a series of unsupervised machine learning methods. Through the use of datasets made publicly available by the Federal Government of Brazil, a new dataset was generated with municipality-level socioeconomic data and a PTB occurrence rate. This dataset was processed using two separate clustering methods, both built by assembling unsupervised learning techniques, such as $k$-means, principal component analysis (PCA), density-based spatial clustering of applications with noise (DBSCAN), self-organising maps (SOM) and hierarchical clustering. The methods discovered clusters of municipalities with both high levels and low levels of PTB occurrence. The clusters with high PTB were comprised mostly of municipalities with lower levels of education, worse quality of public services -- such as basic sanitation and garbage collection -- and a less white population. The regional distribution of the clusters was also observed, with clusters of high PTB located mostly in the North and Northeast regions of Brazil. The results indicate a positive influence of the quality of life and the offer of public services on the reduction of PTB risk.

With the growth of technologies for the automation of electrical systems, it is important to create ways to maintain the supply of electrical energy even in abnormal situations, with fast and low-cost resolutions, keeping load restriction as a last resort. The Corrective Switching technique is a tool capable of controlling the power flow of meshed networks by changing the network topology. The main advantage of this control methodology is that it does not involve additional costs, since its implementation depends on maneuvers by elements already present in the network. In previous works, several techniques and methodologies were developed to reduce the computational time needed to implement the technique, envisioning its application in the operation of systems in real-time. The present work aims to develop a computer program capable of identifying overloads in branches of the system and reconfiguring the network in real-time through the Corrective Switching technique. The simulations of electrical networks in real-time will be carried out through RTS. The computer program with the techniques and methodologies used to choose the switching variant will be developed using Scilab software, which will exchange information with the simulation in real-time in RTS. To improve the performance of the technique in the simulations, linearization techniques were also used to gain computational time. Finally, analysis of the results obtained from simulating the corrective switching technique in real time is presented.

In December of 2019, the first case of COVID-19 was found in Wuhan, China, and in April of 2021, there were already 136 million confirmed cases. Due to the virus fast propagation, the scientific community has been making efforts to develop viral classifications techniques for the SARS-CoV-2. In this work, was developed, using a set of techniques from Genomic Signal Processing, a new proposal of genomic data representation of six viruses from the Coronaviridae family, which the SARS-CoV-2 belongs to. Then, the accomplished mapping was applied in a deep learning architecture for the samples' viral classification, obtaining accuracy of 94% e 91% for the sequences resized for the sizes of 64 and 128, respectively, also obtaining sensibility of 100% for the vectors with size 64. Lastly, given the mutation rate of the RNA virus, new variants emerged, and with them the possibility of an increase in cases. It was then, using the developed technique, carried out an analysis of the evolution of four variants of concern in three viral classification procedures, the results obtained aided comprehending the phylogenetic relationships between the variants.

Este trabalho ou projeto de detecção de qualidade relevante, detectou e detectou ativamente um dispositivo de previsão de previsão e comparação com índices de dados de medição de previsão/final de qualidade com um microcontrole capaz de enviar como informações apresentadas para um banco de todos os dados. Por ser um projeto que necessita de grande escala para que seja possível capturar índices de qualidade do maior número de pontos possível para obter dados com uma granularidade muito alta, o desenvolvimento está sendo pensado semper visualizando o custo-benefício dos componentes para que replicar, e também o desenvolvimento faz parte de um projeto maior possível, que deve ser disponibilizado à comunidade uma plataforma completa, capaz de fornecer dados de qualidade do ar em tempo real.

Distributed generation systems, which use renewable energy sources, are increasingly present
due to the need to meet energy demand and environmental restrictions. Generally, such
generation systems are connected to the electrical grid by means of power converters and,
their integration into it, can produce instability in the electrical system. With the advancement
in power electronics and the wide range of applications involving single-phase voltage
converters, it is possible, with the appropriate control strategy, to dodge instability problems
and also to offer improvements in power quality. For this, several control techniques have
been studied for this purpose, however, the classical techniques fail for not having robustness
to parametric uncertainties and are dependent on the system model. Thus, the control by
sliding modes has gained space in the control of converters because it is a robust technique to
parametric uncertainties and is independent of the system model, but the presence of
chattering in the controlled variable in steady state is its main disadvantage. This master
thesis presents a sliding mode control strategy applied to single-phase grid-forming converters.
The proposed strategy employs linear systems analytical tools to determine the of closed-loop
poles' suitable location when constrained to the splip surface. Furthermore, it is using a
simplified structure for the internal current control. Simulation results obtained from a single-
phase VSI connected to the grid operating under different operational conditions demonstrate
the effectiveness of the proposed solution.

The benefits of frequency-selective control of electromagnetic wave flow make frequency selective surface (FSS) a continuous subject of study in applied electromagnetism areas. With a wide range of applications in various engineering fields, FSS can be used, for example, as an antenna efficiency improvement tool, used for selective electromagnetic shielding, and even for fault detection in concrete structures such as bridge pillars. One particular class of FSS is the Complementary Frequency Selective Surface (CFSS) which has very interesting features together such as high angular stability, multiple transmission and/or reflection bands, and unit cell miniaturization. The analysis and design of this type of structure are commonly performed using commercial software that implements the so-called full-wave methods (FWM) which, despite having reliable responses, requires a license to use and demands high computational effort, impacting a longer time to obtain results. The Equivalent Circuit Model (ECM) provides approximate equations that, by calculating impedances related to unit cell geometry, model the behavior of frequency-selective surfaces in the frequency domain faster due to simplified formulas, which can be implemented in free software in many different programming languages, showing good agreement when compared to full-wave methods. This research aims to optimize the ECM when applied to complementary frequency selective surfaces formed by rings, uniting the interesting characteristics of both the cascade structure and the geometry. The study is paralleled with output techniques from transmission line theory and genetic algorithms for greater effectiveness of the method.

Nonlinear dynamical systems play a crucial role in control systems because, in prac- tice, most of the plants are nonlinear. However, achieving control for nonlinear systems is not simple though many methods have been developed. There are still some problems to be solved, as robust control balance in humanoid robots and the modeling inaccuracies of the autonomous underwater vehicle, which has a small-pitch-angle. Usually, a Lyapunov function is used to perform a control and stability analysis of a nonlinear system. The procedure for obtaining a Lyapunov function is not a simple task. There have been many efforts and numerical methods in the literature on how to estimate Lyapunov functions for several kinds of systems. An artificial neural network is a useful tool for generat- ing functions. Motivated by this, we investigated the capability of a neural network to compute Lyapunov functions and provide a deep neural network to compute a control Lyapunov function without any linear approximation for nonlinear systems. Moreover, we examined the equilibrium point stability and obtained an estimation of its region of attraction contained in the set. Numerical examples and experimental simulations using some nonlinear systems, such as the inverted pendulum and the rotary inverted pendulum, are performed and compared with some conventional control techniques.

The advancements in power electronics have supported the widespread penetration of wind energy conversion systems (WECS) in electric grids. In this sense, the system based on the Double Fed Induction Generator (DFIG) has a large share of installed wind farms power generation technology, in this configuration the stator is directly connected to the grid and the rotor is interfaced by means of two back-to-back power converters, which have a crucial functionality in the machine control by means of voltages applied to the rotor terminals in order to ensure that the stator voltages and currents have the same frequency as the electrical grid, besides controlling the active and reactive power and are directly related to the harmonic distortion, thermal and mechanical stress levels in the WECS. Currently, several power converters topologies have been employed in order to improve the performance and increase the WECS processed power capability to support the network demand, in this sense, the multi-level converter topologies stand out and can be appropriately used to fulfill these requirements. On the other hand, one of the main features of the DFIG-based WECS is its high sensitivity to electrical disturbances, especially voltage sags, since the stator is directly connected to the grid, this feature highlights a fundamental challenge which is to provide fault support and in particular voltage sags to DFIG. Considering the relevance of this problem, this thesis presents a proposal focused on the improvement of DFIG supportability to a voltage sag in the grid by means of a recent multilevel converter topology, the Modular Multilevel Converter (MMC). The proposal is based on the use of the specific characteristics of the MMC through the presence of internal impedances, in order to provide significant damping of the DFIG transient components of the rotor and stator fault currents during a voltage sag in the grid, effectively contributing to maintaining the DFIG connection, while keeping the controllability and avoiding the protection activation, without the need of structural modifications or additional control loops. Moreover, the proposed structure using the MMC directly contributes to the improvement of the system performance by reducing the harmonic distortion rate and electromagnetic torque oscillation, reducing the heating and internal losses in DFIG windings and increasing the useful life of the gearbox, in addition to providing greater robustness, efficiency and reliability to the set of back-to-back power converters. Results obtained by means of simulations indicate the MMC effective damping capability of DFIG currents and the increase in supportability during a grid fault.

The Internet of Things (IoT) is a growing network of objects — sensors, devices, systems, and more — that capture, transfer data, and communicate with each other through communication protocols. In addition, such objects have the ability to produce potentially unlimited sequences of data, called data streams. In this sense, it is clear that the IoT is creating new opportunities for various sectors, as can be seen in an unconventional way in the automotive sector. It is known that as a consequence of the advancement of architecture, vehicles are becoming increasingly equipped with various sensors and computational power. And, from available interfaces, it becomes possible to capture and extract, in an automated way, information through sensors and communication protocols present in vehicles and enabling a scenario known as the Internet of Intelligent Vehicles (IoIV). One of the benefits of IoIV is the creation of diagnostic applications, such as characterizing the behavior of drivers. This type of diagnosis is an essential requirement since the way you drive can impact different contexts, such as traffic safety, fuel consumption, emissions, and maintenance, among others. Furthermore, solutions generally available in the literature for analyzing drivers’ behavior focus on supervised offline learning models, fed with the entire dataset for training and testing. On the other hand, such solutions do not handle data streams suitable for online learning, that is, without knowledge of subsequent data. In face of this reality, the objective of this work is to identify patterns in the behavior of drivers, from a methodology oriented to data streams and unsupervised online algorithms. The methodology is adaptable and flexible, and considers the historical-temporal relationship between the samples, adapting in an autonomous and evolutionary way, without the need for a supervised training phase. In order to validate the proposed methodology, a case study was carried out in a real scenario with different conditions, which allowed the identification of daily driving operations. The results indicated the feasibility of the proposal regarding the identification of event detection and indicators of driver behavior. Therefore, the methodology can contribute to several applications, such as industry 4.0 — customized maintenance, fault detection — smart cities and urban mobility — improvement of pavements, increase in the number of speed reducers and crosswalks, decrease
in the maximum speed of roads, among others.

Syphilis is a chronic infectious disease that remains a serious public health problem in much of the world. In Brazil, the rapid response to syphilis project, No Syphilis!, was created in 2018. This project has several strategies to combat syphilis, including the creation of a group of field researchers who worked in priority municipalities and produced thousands of text reports and added them to the platform. The objective of the work is to use the textual productions of the platform of field researchers of the No Syphilis project, LUES, for Text Mining to understand the impact of syphilis in the territory. The texts will be analyzed using some Data Mining (DM) algorithms. In addition, the answers from questionnaires from the "No Syphilis!" project will be used. with the purpose of finding relationships between the productions carried out and the questionnaires in the contribution of the reduction of syphilis. Associations of the data extracted from the questionnaires and reports with indicators of syphilis and its impact of the epidemic in the territory are tested. With this, it is possible to understand the most effective processes for reducing syphilis in the territories from the support strategy and the LUES platform.

In the oil and gas industry, a device known as the Pipeline Inspection Gauge (PIG) runs through oil and gas pipelines performing various maintenance operations. The efficiency of these operations can be increased by employing a closed loop speed control system. To get the speed, it is usually resorted to the use of odometers. Although such a method is relatively simple, it can cause certain measurement problems resulting from slippage between the odometer wheel and the duct. In order to contribute to the solution of these problems, the objective of this work is to develop a soft sensor (virtual sensor) to measure the velocity of PIGs from the pressure difference to which the device is submitted in the duct. A soft sensor is basically made up of two elements: a mathematical model of the system and sensors that measure the physical variables required by the model. To obtain the model, it is intended to use artificial neural networks. This model will be shipped in a Raspberry Pi to be installed in the PIG, which will also be responsible for obtaining the sensor data. The SIP testing pipeline from the Petroleum Assessment and Measurement Laboratory (LAMP / UFRN) will be used to evaluate the results. The proposed system is expected to be able to complement the use of odometers, thereby increasing the reliability of speed measurement.

The risk involving the use of Unmanned Aerial Vehicles (UAV) has grown directly with its popularization. It is a global concern of regulatory agencies to reduce these risks. In Brazil the National Civil Aviation Agency (ANAC) waives most requirements for UAVs up to 0.250 kg. In this perspective, the present study aims to propose the design and development of a micro UAV flying wing model with remote or autonomous piloting capability with low risk of collision with people and facilities (Micro-Brosh). With a maximum wingspan of 0.30 m and a final weight of 0.160 Kg, the Micro-Brosh UAV's main feature is its easy and safe operation, being able to be launched manually and with the realization of landing on grass cover, without causing damage to the aircraft. We introduce the proposed architectural design, with aerodynamic and performance analyses, demonstrating parameters such as lift, thrust, drag, stall speed, and flight time. In our tests, the prototype proved to be superior to almost all other UAVs in the literature studied, except for one that has similar size and flight time characteristics, but with greater weight and size than the Micro-Brosh. In all, 6 prototypes were built and analyzed in more detail, resulting in the final model that was validated through flight tests and bench tests. In addition to the safety advantages, another important aspect is that the costs associated with the construction and maintenance of the Micro-Brosh is below US$ 500, a value considered low in relation to the traditional UAVs surveyed.

O risco envolvendo o uso de Veículos Aéreos não Tripulados (UAV) tem crescido diretamente com a sua popularização. É uma preocupação mundial das agências reguladoras diminuir esses riscos. No Brasil a Agência Nacional de Aviação Civil (ANAC) dispensa a maioria das exigências para UAVs com até 0,250 Kg. Nesta perspectiva, o presente estudo tem como objetivo propor o projeto e desenvolvimento de um micro UAV modelo asa voadora com capacidade de pilotagem remota ou autônoma de baixo risco de colisão às pessoas e instalações (Micro-Brosh). Com uma envergadura máxima de 0,30 m e peso final de 0,160 Kg, o UAV Micro-Brosh tem como principal característica a operação fácil e segura, podendo ser lançado manualmente e com a realização do pouso em capo gramado, sem causar danos à aeronave. Introduzimos o projeto arquitetônico proposto, com análises aerodinâmica e de desempenho, demonstrando parâmetros como sustentação, empuxo, arrasto, velocidade de estol e tempo de voo. Em nossos testes, o protótipo demonstrou ser superior a quase todos os outros UAV da literatura estudados, menos um que possui características de tamanho e tempo de voo similars, porém com peso e tamanho maiores que o Micro-Brosh. Ao todo, foram construídos e analisados mais detalhadamente 6 protótipos, resultando no modelo final que foi validado através de testes de voo e testes de bancada. Além das vantagens de segurança, outro aspecto importante é que os custos associados com a construção e manutenção do Micro-Brosh fica abaixo de \$500, um valor considerando baixo em relação aos UAVs tradicionais pesquisados.

Currently designs developed in wireless network systems are on signal confinement. We observe the need to increase security and restriction of services, in which electromagnetic signals must remain only in a specific room, without external transmittance. Consequently, several researches are studying the use of Frequency Selective Surfaces (FSS) as a way to reduction electromagnetic interference (EMI) and to maintain the systems' compliance. With characteristic and distinct performance of the conventional passive absorbers of Salisbury and Jaumann, that present good absorption in the band of stop, but, low transmission in the external bands. The absorbers with FSS propose to act with transparency to the electromagnetic waves of the incident plane in the passing band, while absorbing the waves in the wanted frequency bands. Therefore, the objective of this work is to design a microwave absorber structure integrated with frequency selective surfaces with multiband performance. Studies described in the literature are analyzed to obtain such desired absorption characteristics, later structures are projected with variation of their respective parameters. Demonstrating mastery in the use of the commercial software used in the simulations, and understanding the effects of the parametric variations of the devices examined.

In this work, simplified geometries for microstrip antennas were evaluated and optimized, to select one appropriate geometry, aiming for its application of partial discharge (PD) detection, as a UHF sensor. To improve the performance of the proposed antenna, cascaded frequency selective surfaces (FSS), using double square loops as a unit cell element, were designed and evaluated. Antenna reflection coefficient measurements without and with the integrated FSS were performed and compared with simulated results. The results show that the insertion of the FSSs did not impact the operational bandwidth of the antenna, which comprises 100$\%$ of the characteristic frequency range, of the PD activity (300 – 1500 MHz). The insertion of the FSSs resulted in an average gain of 4 dBi, concerning the isolated antenna, increasing the sensitivity for PD detection, within the proposed frequency range. We evaluated the PD detection of the structure through measurements in three devices, a bar of a hydro generator, an oil tank with electrodes in the flat-tip configuration, and a potential transformer (PT). The proposed UHF sensor for PD detection could detect PD activity in all three scenarios and for levels lower than 10 mV. 

This work proposes a methodology for the detection and identification of failures in dynamic systems, through an online and evolutionary approach. The proposal is divided into three stages, in which data pre-processing and post-processing are carried out to increase the robustness of the methodology in the presence of outliers and noise, in the pre-processing the selection of characteristics, normalization is carried out of data, filtering and adding regressors, in post-processing time filtering is performed. In the processing stage, an adaptive and unsupervised approach is applied, through the Auto-Cloud algorithm, which performs grouping and classification of data streams. To validate this proposal, different evaluation metrics were used, such as Adjusted Rand Index (ARI), homogeneity, completeness, precision, f1_score, recall, and satisfactory results were obtained. Finally, the conclusion of this work is presented, in addition to proposals for future work.

Artificial intelligence, machine learning, and deep learning have proven to be powerful techniques for solving natural language processing problems, computer vision, and others. However, its computational and statistical performance depends on other factors of the algorithms used for training, the computing platform used, and even the numerical precision used to represent the data. One of the main machine learning algorithms is the Support Vector Machine (SVM), most commonly trained through minimal sequential optimization, contributing to low computational performance. As an alternative, algorithms based on Stochastic Gradient Descent (SGD) have better scalability and can be used as good options for training machine learning algorithms. However, even with algorithms based on the stochastic gradient, the training and inference times can become long depending on the computing platform employed. Thus, accelerators based on Field Programmable Gate Arrays (FPGAs) can be used to improve performance in terms of processing speed. This work proposes a fully parallel FPGA implementation of an SVM with SGD-based training. Results associated with hardware occupation, processing speed (or throughput), and accuracy for training and inference in various quantization formats are presented.

The problem of highlight removal from image data refers to an open problem in computer vision concerning the estimation of specular reflection components and the removal thereof. In recent applications, highlight removal methods have been employed for the reproduction of specular highlights on high dynamic range (HDR) displays; to increase glossiness of images in specular reflection control technologies; to improve image quality in display systems such as TVs; and to enhance the dynamic range of low dynamic range (LDR) images. However, the underlying processing required by state-of-the-art methods is computationally expensive and does not meet real-time operational requirements in image processing pipelines found in consumer electronics applications. In addition, these applications may require that methods work with a single frame in imaging or video streams. Consequently, this work has the objective of proposing a novel method for the real-time removal of specular highlights from a single image. The essence of the proposed method consists in matching the histogram of the luminance component of a pseudo-specular-free representation using as reference the luminance component of the input image. The operations performed by the proposed method have, at most, linear time complexity. In experimental evaluations, the proposed method is capable of matching or improving upon state-of-the-art results on the task of diffuse reflection component estimation from a single image, while being 5X faster than the method with the best computational time and 1500X faster than the method with the best results. The proposed method has high industrial applicability, and targeted use cases can take advantage of contributions of this work by incorporating the proposed method as a building block in image processing pipelines.

In this work, we propose to study and verify the potential of using Educational Robotics in the classroom to improve cognitive functions and at the same time act in the prevention of the occurrence of mental disorders, a problem that has hindered the socio-cultural development of many young people today. It is worth mentioning that there are studies verifying the effectiveness of Educational Robotics with children and adolescents in order to assist in the diagnosis and treatment of the autistic spectrum and in visual impairment, but we do not find in the literature a study similar to the one proposed in this work. Furthermore, it is already demonstrated in the literature that Educational Robotics improves the child's cognitive functions and also their and teachers' self-esteem. During the work, from the analysis of collected data, we were able to verify strong indications that children and adolescents who take educational robotics classes are less susceptible to major problems of depression, stress, among other types of mental disorders. To reach this inference, we developed and applied questionnaires for teachers and students, from which variables were determined to be analyzed, such as psychomotor, cognitive and affective. The questionnaires were applied in the practical phase of the Brazilian Robotics Olympiad that took place at the Federal Institute of Rio Grande do Norte, Natal, Brazil. These data were submitted to an analysis and later verification through qualitative and quantitative methodologies on the educational robotics tool used.

Industry 4.0 set a paradigm shift in industrial process monitoring and control. It installed several sensors in different parts of the plant, connecting these industrial processes with the Internet of Things and Cloud Computing. Although, the data generation growth demanded engineers build proper environments to process and store the absurd amount of data. This growth caused an increasing energy consumption, computational complexity and environmental degradation. Therefore to address these demands, this dissertation proposes efficient approaches to perform Fault Detection and Identification in industrial processes. The first approach consists of using Symbolic Aggregate Approximation (SAX) to compress process variables to reduce the load on data warehouses. Then, we train a Long Short-Term Memory (LSTM) neural network with those compressed inputs to perform fault detection. Finally, the second approach addresses efficient edge computing systems, performing LSTM neural network compression with pruning technique. The compression reduces the memory usage and number of operations of these networks, saving energy and accelerating inference speed in edge computation. To assess the performance of both approaches, we use the Tennessee Eastman Process (TEP) as the benchmark with classification metrics of accuracy, precision, recall and F1-Score. We are also going to analyze the compression efficiency of both approaches, studying their viability and parameter reduction in LSTM networks.

The squirrel cage induction generator is a robust and low cost alternative to variable speed wind energy conversion systems. For this type of generator, direct torque control, direct stator-field-oriented control, direct rotor-field-oriented control, stator-flux slip control and rotor-flux slip control can be used for maximum power point tracking. Stator-flux slip control and direct torque control have simple schemes with fast dynamic responses, since they do not have internal current controllers. However, stator-flux slip control presents high inrush-current and a poor dynamic response and direct torque control presents high torque ripples and poor performance at low speeds. On the other hand, rotor-flux slip control, direct stator-field-oriented control and direct rotor-field-oriented control have fast dynamic response, low inrush current and low error at maximum power point tracking, but have current controllers and offer a larger complexity in the control design. The objective of this work is to define the most suitable control strategy by evaluating the following performance indexes simultaneously: maximum power point tracking, total harmonic distortion, inrush current and dynamic response. Simulations are performed in order to analyze the performance of the control strategies. Tests were carried out for the five control strategies and the direct rotor-field-oriented control presented the best trade off considering maximum power point tracking, total harmonic distortion and dynamic performance, despite it does not have the best inrush current.

With the advancement and mass adoption of solutions in the fields of Internet of Things (IoT) and connected cities, the number of devices and sensors connected to the network tends to grow exponentially. In this scenario, the transmission and storage of the growing volume of data bring new challenges. When devices transmit potentially irrelevant or redundant data, there is increased energy and processing waste, as well as unnecessary use of the communication channel. Thus, local data compression solutions on the IoT devices themselves become increasingly attractive, enabling the elimination of samples that would have little or no contribution to the application, in order to significantly reduce the volume of data needed to represent the information. However, such devices present on the market today have serious storage and processing power limitations. In order to circumvent these limitations, the TinyML field emerges, which seeks ways to implement machine learning models in low-power devices. Given this context, one of the sectors that can benefit most from these new technologies is the automobile industry, as currently all cars produced must be instrumented with a series of sensors. In this way, by connecting an intelligent device to the vehicle, it is possible to process the data locally and transmit it to a remote server later. In this context, the present work proposes the development of a new online, unsupervised, and automatically adaptable data compression algorithm for IoT applications. The proposed approach is called Tiny Anomaly Compressor (TAC) and is based on data eccentricity and does not require pre-established mathematical models or any assumptions about data distribution. To test the effectiveness of the solution and validate it, two tests were carried out with different objectives. First, a comparative analysis on two real-world datasets was developed with two other algorithms from the literature, the Swing Door Trending (SDT) and the Discrete Cosine Transform (DCT). Finally, the proposal was embedded in an IoT device based on an Arduino and connected to a car to verify the impact of the algorithm on the processing time of the system's primary operations. Preliminary results show that it is possible to achieve significant compression rates without significant impacts on the generated error and system processing.

This work proposes a technique for counting people in an already populated environment. Initially, a survey is made of the technologies and solutions designed for this purpose. As a proof of concept, a counting solution is analyzed for a small number of people, applying machine learning to the descriptive statistics of an RF signal. Finally, the classification results are presented for a more realistic scenario, with up to 350 people in the environment, using a software-defined radio measurement system for data collection. The results show significant precision in counting the number of people by classification in groups of individuals.

Magnetic levitation teory attracts a lot of interest from the academic community currently, as it has a great potential in present promising technologies to the market, mainly in the passenger and cargo transportation sector and in the industrial sector, providing a great positive impact for society and contributing to the your well-being. Under this premise this work proposes to analyze the application of type-2 fuzzy control as a better alternative for control design for systems of this class, analyzing the positive characteristics of the application of this control technique in the face of type-1 fuzzy. The controllers were designed and tuned manually, in order to observe if the extra degree of freedom provided by the type-2 fuzzy technique presents a superior performance. The experimental results show that the type-2 fuzzy controller provides a superior performance to its type-1 counterpart in different aspects. With this methodology, the best characteristics of the different classes of control systems can be combined to achieve a more efficient and intelligent control scheme.

The space environment is a harsh environment for digital circuits. The effects of space radiation on the circuit might result in malfunction and, as a consequence, jeopardize the security and execution of space missions, manned or not. Therefore, there was a need to develop digital circuits which can withstand radioactive particle’s strikes and, nowadays, there are several processor technologies that can indeed work safely under such conditions. However, these technologies generally involve proprietary transistor fabrication steps or proprietary cell libraries which inhibit usage and development. There is, nonetheless, another set of architectural methods which operate on a higher abstraction level and are arguably as reliable as the mentioned techniques for a set of space missions. This work uses these techniques to develop the CEVERO, a general-purpose processor base on the existing PULP Platform, which uses architectural modifications such as redundancy and other methods to detect and correct faults, for deploying on aerospace missions.

With the increase in interest in renewable energies in the world, the importance of research in wind energy, methods, and associated technologies has grown. The increase in wind turbines connected to the electrical systems also brought with it new challenges related to the stability of the system and protection of its components. Differential protection is used in many electrical system components and for DFIG-based wind turbines this protection has been proposed recently. Therefore, this work proposes criteria in the decision-making of the differential protection based on the Park transform, which in addition to the restriction current also uses the voltage level and the direction of the currents at the stator terminals. In a simulation with DFIG operating at 1.51 MW, tests were carried out for internal and external faults to the protection zone, of the single-phase and threephase type with different levels of voltage drops. Inter-turn faults on the machine’s rotor and stator have also been tested with varying levels of defects. Thus, with the proper def inition of the operating logic, it was possible to show the effectiveness of the protection, with detection for all single-phase and three-phase faults within the protection zone, detection of faults in windings in the stator with 10% defect or more and detection of faults in the rotor windings with a defect of 20% or more.

Multi-objective optimization problems depict real situations and therefore, this class of problems is extremely important. However, even though it has been studied for decades, this class of problems continues to provide challenging situations, especially by the continuing lack of effective techniques. Among all the difficulties that we can find in the optimization of multiple objectives simultaneously, whether conflicting or not, one of the main difficulties found by the algorithms and existing approaches is the need for a priori knowledge of the problem, causing a predefined importance for each of the objectives. When dealing with this class of problems through reinforcement learning, two approaches are predominant: single policy (single-policy) and multiple policies (multi-policy). Algorithms and techniques that use the first approach suffer from the need for prior knowledge of the problem, an inherent characteristic of multi-objective problems. The second approach has other difficulties, such as: limiting the set of solutions and high computational cost. Given this presented context, the work proposes two hybrid algorithms, called Q-Managed with reset and Q-Managed without reset. Both are a hybridization of the Q-learning algorithm and the ε−constraint approach, respectively belonging to reinforcement learning and multi-objective optimization. In summary, the proposed algorithms work as follows: Q- Learning is used for environment exploration, while the ε−constraint approach is used for the environment dynamic delimitation, allowing to keep intact the essence of how the algorithm Q-Learning works. This delimitation has the following purpose: to impose the learning agent can learn other solutions by blocking actions that lead to solutions already learned and without improving them, that is, solutions to which the learning agent has already converged. This blocking actions feature is performed by the figure of a manager, where it is responsible for observing everything that occurs in the environment. Regarding the difference between the proposed algorithms, basically it is the choice of whether or not to take advantage of the knowledge already acquired of the environment after a solution is considered to be learned, that is, the learning agent has converged to a particular solution. As a way of testing the effectiveness of Q-Managed two versions, traditional benchmarks were used, which were also adopted in other works, thus allowing a fairer comparison. Thus, two comparative approaches were adopted, the first of which was through the implementation of third-party algorithms for direct comparison, while the second was done through a common metric to everyone who used the same benchmarks. In all possible tests, the algorithms proposed here proved to be effective, always finding the entire Pareto Front.

In recent years, advances in wireless communication systems and the growth in the number of applications have made it attractive to use compact multiband and broadband antennas for both commercial and military communications. The most common examples are found in the variety of portable wireless communication devices, such as: smartphones, handsets, palmtops and laptops, among others. The development of different wireless technologies, such as GSM (Global System for Mobile Communications), RFID (Radio Frequency Identification), Bluetooth, Wi-Fi and WiMAX, has served as a motivation to boost searches in search of lighter devices , compact and low cost, as is the case of planar microstrip antennas. The development of antennas with these characteristics is still a challenge to current research. This work aims at the study, design, manufacture and measurement of new planar antenna geometries, with openings in the ground plane and the use of FSS (Frequency Selective Surface), aiming to reduce its size, volume and weight as well as gain of antennas, to meet the demands of the services offered by the Mobile Systems 5G, for example. The insertion of air layers between the antenna and the EBG structure was also used aiming at a greater degree of freedom to adjust the resonant frequencies. Computational simulations with the objective of predicting the values of the parameters that characterize the proposed structures were performed through the CST and HFSS 13.0 softwares. Prototypes were manufactured and measured. A good agreement was observed between the simulated and measured results.

Several analytical models created since Amdahl's pioneering work have explored aspects such as variation in the size of the problem, memory size, communication overhead, and synchronization overhead, but delays in accessing data are considered constant. However, such delays can vary, for example, according to the number of cores used, the relationship between processor and memory frequencies, and the problem's size. Given different problem sizes and the large number of possible configurations of operational frequency and number of cores than current architectures can offer, speedup models suitable for describing such variations among these configurations are quite desirable for offline or online scheduling decisions. A new analytical performance model that considers variations in the average data access delay to describe the limiting effect of the memory wall in parallel applications in homogeneous shared memory architectures is presented in this thesis. The experimental results indicate that the proposed modeling captures the application's behavior well. Besides, we show that considering parameters that reflect the applications' intrinsic characteristics, the proposal presented in this work has significant advantages over statistical models such as those based on machine learning. Our experiments also show that conventional machine learning modeling may need about an order of magnitude more measurements than the proposed model to achieve the same level of precision achieved by the proposed model.

Syphilis in Brazil grew by over 5000% between 2012 and 2017, this drew the attention of the Ministry of Health, which in 2017 declared a syphilis epidemic Brazil. The hardest side of syphilis is in vertical transmission, that is, when a pregnant woman transmits Treponema Pallidum to her baby. The Federal Court of Accounts (TCU - Brazil), through a serious and in-depth audit regarding the increase in syphilis cases in Brazil, made several notes and recommendations, among them the need to develop tools and technologies that make it possible to monitor more efficiently the evolution of syphilis countrywide. However, when it comes to monitoring it is important to look beyond epidemiological issues, diagnosis, care, treatment and cure. These factors are important, for example, to eliminate vertical transmission of syphilis and, consequently, congenital syphilis, the latter being one of the greatest challenges for Brazil. In this context, it is necessary, therefore, that the actions to face syphilis in Brazil are coordinated between Health Surveillance and Health Care. These two areas, despite being conceived by the Brazilian Health System (SUS) as essentially articulated with each other, operate in a large part of the country in a dichotomous way, including in the Ministry of Health itself. This sometimes occurs due to the lack of integration between the teams that they work in different areas of health, or because the technologies made available by DATASUS are far from understanding the need to mediate a logic of integration between these strategic areas of SUS. The present doctoral thesis is, therefore, situated in this dimension, it brings in its discussion the development of a technological architecture that in fact is a digital health solution that mediates the integration between Health Surveillance and Health Care through technology. both, it incorporates in its flow the management of cases of Syphilis in the territory, that is, in the municipality, places where health is closest to the population. At the same time, it produces epidemiological and care indicators on syphilis. The construction and validation of this architecture was carried out in the city of Natal-RN and the entire discussion surrounding the development and application is presented in this doctoral thesis.

The expressive growth of modern data sets, combined with the difficulty of obtaining information about labels, has made semi-supervised learning one of the problems of practical importance in modern data analysis. In most cases, obtaining a set of data with enough examples to induce a classifier, can be costly, as it is necessary to carry out data labeling by a specialist. Unlabeled data is easier to obtain but more difficult to analyze and interpret. In of semi-supervised learning problem, there is a database formed by a small part labeled and a larger part not labeled, being possible two aspects: semi-supervised classification and semi-supervised clustering. With this, this work aims to apply models that use Deep Learning techniques in semi-supervised learning, where a neural network is trained, in this case, an autoencoder using unlabeled data. Then, an additional layer is embedded in the encoder. This new layer has its weights initialized by the K-means ++ algorithm and adjusted through the backpropagation algorithm using information theory learning. The labeled data is assigned to the clusters generated by the encoder, influencing the unlabeled, cluster by cluster, thus labeling the non-labeled data that was previously clustered. With the experiments carried out, it was noted that the satisfactory performanceof the proposed model when compared with other semi-supervised algorithms, both the classics such as self-training and co-training, as well as other more recent works, showing the proposed model feasibility for the learning semi-supervised problem.

The incessant development of implanted or promoted biometric sensors has led to the growth of Wireless Body Area Network (WBAN) technology. A WBAN can be used for many applications, such as monitoring of physiological signals and industrial communication applications. In this context, the textile antenna is an important element in wireless communication in smart fabrics, being the subject of recurrent research in industrial, military and medical applications. These antennas are flexible and, therefore, become convenient in applications where antenna rigidity is an obstacle This work presents the development of planar antennas with a textile metamaterial substrate in the 5G frequency range for WBAN applications, showing all process from making the substrates to the manufacture and testing of the antennas. Three types of textile substrate with different structure and composition were selected, and in each of them a conductive thread was inserted through different techniques (manual insertion, knitting and automatic sewing). The characterization of the materials was performed using simulations in the HFSS and the coaxial probe method, to obtain the electrical parameters. The antennas were designed and simulated to operate at 3.5 GHz, the best results being manufactured and tested in a Vector Network Analyzer (VNA). The results obtained in the simulations and tests were analyzed and the presence of a metamaterial characteristic was found in the loop and straight stitch geometries. Antennas with a textile metamaterial substrate had a lower resonance frequency than those with a dielectric textile substrate, indicating the possibility of reducing the dimensions of the device.

In the multi-core era, the size of the software operation space, i.e. hardware configurations that provide different software performance and energy consumption, is significantly larger. It becomes even more complex to choose a configuration that optimizes heterogeneous processors' performance and energy consumption. Heterogeneous multicore architectures offer flexibility in different core types and voltage and frequency pairings, defining a vast design space to explore. Furthermore, energy efficiency solutions are crucial on smaller devices as they can lead to longer battery life and a better user experience, including more complex applications. This thesis proposes a methodology to find performance-energy trade-offs for single parallel applications with dynamically balanced workloads running on HMP systems with a single instruction-set architecture (ISA). Our method devises novel analytical models for performance and power consumption whose parameters can be fitted using only a few strategically sampled offline measurements. These models are then used to estimate an application's performance and energy consumption for the whole configuration space. In turn, these offline predictions define the choice of estimated Pareto-optimal configurations of the model, which are used to inform the configuration that the application should execute. The methodology was validated on an ODROID-XU3 board for eight programs from the PARSEC Benchmark, Phoronix Test Suite and Rodinia applications. Energy savings of up to 59.77%, 61.38% and 17.7% were observed compared to the performance, ondemand and powersave Linux governors, respectively, with higher or similar performance. This method aims to provide an optimal start point for a runtime energy manager to make better decisions according to the given application's performance and energy consumption requirements and running system. Therefore, this thesis also proposes a strategy using the Pareto-optimal configuration selected by our models as an appropriate start point for a runtime support framework called Nornir.  This framework performs a  local search dynamically for a more desirable configuration of cores and frequency adapting to workload fluctuations and external interference.  Also, we extend our power model to predict the whole device's consumption, i.e. the sum of all internal components' consumption. This hybrid approach was employed on an ODROID-XU3 board on two multi-thread applications. Nornir starting with Pareto configuration can achieve up to 50% of energy savings compared to random starting configurations. We also observed that the performance and interactive Linux governors consumed up to 1.62X more energy than Nornir using Pareto.

This thesis proposes a versatile methodology for measuring, modeling, and predicting errors as the RMSE in the depth and the RPE in the camera positioning using data captured from an RGB-D camera mounted on the top of a low-cost mobile robot platform. The proposed method is divided in three stages, with the very first one consisting on the creation of ground truth data for both 3d points (mapping) and camera poses (localization). The next stage is the acquisition of a data set for RMSE and RPE errors calculation using the mobile platform with the RGB-D camera. Finally, the third step is to model and predict the errors in the measurements of depth and positioning of the camera as a function of distance, speed, and vibration. For this modeling and prediction stage, a simple approach based on MLP neural networks is used. This results in two networks, the N rmse Z for the depth error prediction and the N RPE for the prediction of camera positioning error. Experiments show that the N rmse Z and N RPE have an accuracy of ± 1% and ± 2.5%, respectively. The proposed methodology can be straight used in techniques that require an estimation of the error dynamics, as for example probabilistic robotics for localization and mapping, with RGB-D cameras mounted on UAVs, UGVs, and also USVs (including sailboats). Tasks that use RGB-D sensors, such as environmental monitoring, maintenance of engineering works, and public security, could rely on this approach to obtain the error information associated to the camera measurements (depth and positioning).

The wave equation is pervasive in mathematical physics and engineering, and we need to solve it repeatedly to simulate wave propagations in software. The spectral element method, one of several approaches for the numerical solution of the wave equation, discretizes the underlying domain in a mesh made of elements and nodes, and traverses every element and every node at each time step as it marches the target equation through time. We propose a memory reordering algorithm, meant to be used with the spectral element method, that rearranges mesh-related data to reduce the number of cache misses and boost locality of data reference, thereby improving the execution speed of the mesh traversal process. We devise a spectral element method formulation for 2D waves over unstructured meshes made of triangles, and we pair it to our memory reordering algorithm to construct an acoustic wave propagation simulator. Our experiments show that the reordering technique based on Hilbert space-filling curves performs well in meshes of different granularities, and also when the variation in element sizes is either small or large. In addition, we compare the proposed approach with three other memory reordering schemes, and find that our algorithm runs between 9% and 25% faster than the alternatives we tested. We recommend this memory reordering algorithm to any application that requires successive traversals across domains.

Alarm and event logs make up a voluminous and dormant historical repository of tabular-like data, commonly undervalued or overlooked in manufacturing. Although they are a potentially rich source of relevant information about the monitored plant or process, these records are taken for analysis only as a last resort, mainly due to the difficulties imposed by the large volume and low expressiveness of those databases. Such oversight is no longer acceptable in the contemporary data-oriented scenario, already ubiquitous in several productive sectors and gaining prominence in traditional manufacturing, especially due to the advent of the Industry 4.0 paradigm. Therefore, it is proposed to transpose these bases to a more expressive and flexible representation domain, allowing a more proactive exploration of the episodes reported in the records and, consequently, entailing more agile incident, anomaly, compliance, and performance analysis tasks. For such, from the recognition of an ontology, entities, attributes, and associations virtually immersed in the operational context described in the records are mapped into a Knowledge Graph (KG). The approach uses Exploratory Data Analysis, Natural Language Processing, Network Analysis, Multivariate Analysis, and Composite Indicators techniques to derive a myriad of aspects, properties, and relations from data, which are incorporated as hierarchical, temporal, and similarity relationships (edges) between identified entities (nodes). The visualization of the KG is dynamic and interactive, with different visualization modes and levels of detail. Evaluation scenarios are designed to demonstrate the effectiveness of the approach.

The needs of new wireless communication systems demand that radiant devices are increasingly stable and reliable. In recent years, several proposed solutions are observed in the literature to improve the performance of planar antennas applied in modern wireless communication systems. This work presents the use of ceriferous substrates as a biodegradable solution, of low cost, electrically stable and with many application possibilities to improve the operation of planar antennas in wireless communication systems. Although the use of organic substrates is not new, a variety of manufacturers have lines of boards for planar circuits based on organic compounds. This work proposes in an innovative way the use of ceriferous substrates for the development of miniaturized microstrip antennas. Planar antennas were developed for applications in IEEE 802.11 ax protocol services. Throughout this work, the process of characterizing the proposed material, making the substrates, designing and manufacturing the proposed antennas, numerical analysis and experimental characterization will be presented. Four types of ceriferous substrates based on carnauba wax were selected, and the main characteristics were determined by analyzing the material's electrical properties, X-ray diffraction and thermogravimetry. With results from the characterization of the substrates, the radiating devices were designed and numerically analyzed by the Finite Element method, using the Ansys HFSS® software. The antennas were experimentally characterized in the frequency range from 1GHz to 8.5 GHz, where it was possible to verify that the material used as substrate presented stability and good properties for applications in planar circuits, as in the case of microstrip antennas. There is also the possibility of miniaturization of the radiant elements using carnauba wax as a substrate, when compared to the same structure designed on a standard commercial substrate such as FR4, a miniaturization factor of 44.5% was achieved in the volume of the miniaturized antenna on wax substrate.  

Photovoltaic solar energy has proven to be a viable alternative that contributes to sustainable development and ensures energy supply around the world. However, the exponential growth of installed capacity in recent years has highlighted the need to ensure the safe operation and reliability of photovoltaic systems. In this context, faults in such systems are a crucial issue, since they can significantly impact the generated power, reduce the useful life, and cause potential risks in operation. Thus, this research applied artificial intelligence techniques to detect and diagnose faults in photovoltaic modules. The faults identified by the proposed methods are short-circuit modules, string disconnection and partial shading. In addition, multilayer perceptron neural network algorithms, probabilistic neural networks, and a neuro-fuzzy method were developed, combining a neural network with fuzzy logic. All trained algorithms were used from simulated and tested experimental data from three different photovoltaic systems. Moreover, training situations in which the dataset is contaminated by random noise were also considered. The results indicated maximum accuracy of 99.1% for the lack of short-circuited modules, 100% for string disconnection and 82.2% for the lack of partial shading. Furthermore, the analyzes allowed to reaffirm the robustness of the multi-layer perceptron network for fault detection in photovoltaic systems, even with the presence of noise in the training data.

In this work, angle synchronization method and fault location method were proposed, based on the fundamental frequency phasor approach, independent of the electrical parameters of the transmission line and of iterative processes. In the synchronization algorithm proposed the line capacitive effect is taken into account and only the analysis of steady-state voltage and current phasors is required, eliminating the need for settings based on line electrical parameters. In addition, it identifies angles in any quadrant and is insensitive to operation conditions of the system. Real operating conditions of the Brazilian system are considered to assess the algorithm performance under different scenarios, including light, moderate and heavy line loading cases, and various power factors and line lengths.  The obtained results show that the synchronization proposed solution is reliable, straightforward, accurate, being useful for practical two-terminal phasor-based applications when traditional common time reference sources are not available. The tool developed for fault location requires pre and post-fault current and voltage phasors, work for all types of faults (symmetrical and asymmetrical), and for transposed and untransposed transmission lines. The evaluation of the fault location algorithms is performed through fault simulations by the ATPDraw software. The obtained results prove the efficiency of the method in finding fault location with good precision.

The tasks organization among computational nodes affects the performance of the program. In computationally expensive applications such as geophysical problems, the impact is more significant. The imbalance caused by inefficient task scheduling can generate an application non-scalable. Therefore this work investigates the workload scheduling in geophysical methods.

 Three types of workload scheduling will be compared: centralized dynamic (CD), decentralized static (DS), and decentralized dynamic (DD).  The CD was implemented with the technique master-slave, where the master node is responsible for the distribution of the tasks to other nodes called slaves. The DS with an equal division of tasks. The DD implements the work-stealing method proposed by Assis et al. 2019, where an idle node can steal the tasks of an overloaded node.

The principal geophysics method used was Full waveform inversion (FWI) 2D with the acoustic wave. To analyze the performance of workload scheduling methods, we employ a synthetic velocity model and present speedup, efficiency, and load distribution plots generated with different model sizes and different quantity of nodes. The FWI code and the workload scheduling methods were implemented in C to distributed memory parallelization and using the message passing interface (MPI) library.

With the results of the workload scheduling methods in FWI, it was applied the work-stealing (because this technique presented the more effective performance) in another geophysics problem: Least-squares migration (LSM).  We used the LSM with DS implemented by Chauris et al. 2017 to compare with the LSM with work-stealing. The LSM code and DS used in this problem were implemented in Fortran and the work-stealing in C. The communication between the nodes was implemented using MPI.  To analyze the performance of workload scheduling methods in LSM, we used the marmousi velocity model.

The growth of data to be processed in the Oil & Gas industry matches the requirements imposed by evolving algorithms based on stencil computations, such as Full Waveform Inversion and Reverse Time Migration. Graphical processing units (GPUs) are an attractive architectural target for stencil computations because of its high degree of data parallelism. However, the rapid architectural and technological progression makes it difficult for even the most proficient programmers to remain up-to-date with the technological advances at a micro-architectural level. In this work, we present an extension for an open source compiler designed to produce highly optimized finite difference kernels for use in inversion methods named Devito c . We embed it with the Oxford Parallel Domain Specific Language (OP-DSL) in order to enable automatic code generation for GPU architectures from a high-level representation. We aim to enable users coding in a symbolic representation level to effortlessly get their implementations leveraged by the processing capacities of GPU architectures. The implemented backend is evaluated on a NVIDIA R GTX Titan Z, and on a NVIDIA R Tesla V100. in terms of operational intensity through the roofline model. Experimented with a 3D acoustic isotropic wave propagation stencil kernels for varying space-order discretization levels. It achieves approximately 63% of V100’s peak performance and 24% of Titan Z’s peak performance for stencil kernels over grids with 256 3 points. This study reveals that improving memory usage should be the most efficient strategy for leveraging the performance of the implemented solution on the evaluated architectures.

The motor activities of the human body, as well as those related to decision making and emotional and psychic issues, can be understood through the analysis of electrical signals from the brain, also known as electroencephalogram (EEG) signals. The study and application of electroencephalographic data has been growing within the scientific community. It is known that the use of these signals forms the basis of the development of the Computer Brain Interface (ICC), and that these represent the future of assistive technologies, especially those aimed at people who do not have motor control. However, the extraction of characteristics and patterns of these signals is still a complex process. Surveys involving ICC and EEG signals usually implement event related potentials (ERP) analyzes; being the main ones: the static potentials evoked visually (SSVEP) and the potential P300. In general, they are responses to external stimuli (visual, auditory, tactile), and are widely used to recognize patterns in EEG signals associated with changes in brain activity. The purpose of this paper is to analyze the signs of the neural activity of individuals who are exposed to external stimuli using the identification of potential SSVEP and P300. The project uses a low-cost, non-invasive EEG signal sensor with wireless, wireless technology. It is expected to extract EEG data to the point of making possible the correlation of these with characteristics that can be applied in control tools.

The growth of mobile internet access from fourth generation (4G) devices, combined with increasing usage of smartphones, the upcoming fifth generation (5G) and massive usage of multimedia services, make the demand for mobile traffic reach high levels and the need for bandwidth grows. However, the electromagnetic spectrum utilized by these applications is limited, creating scarcity in the face of demand, besides the high costs and bureaucracy for granting new bands. To overcome this problem, one of the solutions is to leverage the unlicensed spectrum, for it is free of charges, presents some of its portions with bandwidth higher than the licensed spectrum, and portions with underused profile, indicating less probability of interference between technologies. In this context rises the LTE-U and LTE-LAA technologies with modifications in the medium access mechanism of LTE for leveraging the unlicensed spectrum in the Industrial, Scientific and Medical (ISM) 5 GHz band. These technologies must coexist with the most successful and consolidated technology already using this portion of the spectrum, the Wi-Fi. However, each technology implements its access mechanism, then studies must be done to point out impacts that each of these technologies has when coexist. Besides the coexistence study, the application of machine learning techniques to automatically adjust the medium access parameters, controlling the generated impact of one technology into another must be realized.   Studies in such a scenario and with single-cell environments have already been explored in literature, remaining the challenge of new solutions targeting multi-cell environments. For all it has been exposed, this work has the following goals: (1) The coexistence study of LTE-U and Wi-Fi technologies in a multi-cell scenario, with co-channel and inter-RAT (same technology) interference; (2) The application of machine learning algorithms (reinforcement learning) to adjust the parameters targeting optimizing the medium access for one, or both technologies, and consequently reach improvements in the coexistence measured in the form of data rates and decreasing packet losses. 

Due to the increasing demands in the operation monitoring of industrial plants, methodologies for fault detect and diagnose in the operation of these processes are gaining more and more importance, because they can contribute to more assertive and even predictive repairs in the components that generated such disturbances to the proper functioning of the system. With the growth of data-oriented approaches, Artificial Neural Networks have become considerable allies in solving these problems, and Recurrent Neural Networks, in particular, have gained strength due to their affinity in dealing with series that have temporal links between their samples, which is the case of industrial process variables monitoring. Due to this relevance, this dissertation analyzes the performance of Long Short-Term Memory (LSTM) recursive neural network for the detection and classification of faults in a pilot-scaled level control process. For the performance evaluation, a methodology based on Monte Carlo statistical tests was used, in which the influence of the LSTM network hyperparameters, such as number of layers and size of the input and regressors, was analyzed. The accuracy was the metric chosen to quantify the fault classification performance. The data set obtained from the operation of the pilot plant contained 23 situations of disturbances in this process, which resulted from disturbances applied to components such as sensor, valves and the water tank itself. The adopted methodology proved to be quite efficient to analyze both the performance and the robustness of these neural networks for the fault classification activity, in addition to indicating the best network architecture configurations.

In recent years, the demand for high data rates in wireless communications has increased. The Enhanced Mobile Broadband (eMBB), one of the use cases of fifth generation of mobile communications systems (5G), requests 20 Gbps of minimum peak data throughput using carrier aggregation technique to achieve ultra-widebands. However, wideband and ultra-wideband signals are more susceptible to time and frequency spread degradation due to signal propagation over multiple paths and relative movement between transceivers, respectively. In this context, the channel sounding is of great relevance, since it allows to estimate the characteristics of the transmission medium, enabling the conception of solutions to combat the performance degradation imposed by the communication channel. From this perspective, this work develops a research on the main wideband channel sounding techniques, taking as main objective the conception in software of a channel sounding simulator and a post-processing and analysis tool for wireless wideband channel characterization. In addition, a comparative analysis of the direct cross-correlated soundings is performed, using 6 different sequences as sounding signal. Among them are the sequences PN, Frank-Zadoff-Chu and Golay.

Artificial Neural Networks (ANNs) are bio-inspired systems with a high level of parallelization and almost infinite applications. However, due to the associated high computational power requirements, most application demands powerful processing characteristics and consequently, high-costs and not-so-small form-factors. This work presents an implementation of a Multilayer Perceptrons (MLPs) for 8-bit microcontrollers in two different scenarios, embedded training, and inference. Analysis of training convergence, inference time duration, and program code occupation into the internal memories and a technique to optimize this implementation to fit bigger MLP architectures. The aim of this work is to provide an overview of the feasibility of ANNs on these low-cost, low-size-memory devices, known as microcontrollers

In this work, a methodology for the computation of robust controlled invariant polyhedra of fixed complexity, based on bilinear optimization, is proposed for discrete-time linear systems, subject to constraints on the states and control inputs and to bounded disturbances. In many real-life applications, certain variables in a system must comply with certain constraints. In general, these constraints are specified by linear inequalities that define limited and closed polyhedral sets. A set is robust controlled invariant if any state trajectory starting in this set can be maintained within it through a suitable control input, in spite of the disturbances. Thus, the calculation of a controlled invariant set is an important step in solving control problems under constraints. Conventional methods for robust invariant polyhedra computation can result in high complexity sets, defined by a large number of vertices. The methodology proposed in this work has bilinear invariance conditions and polyhedra represented by vertices, whose quantity is fixed in advance. The aim is also to maximize the volume of the robust controlled invariant polyhedra. Through numerical examples, the methodology is able to compute polyhedra with larger volumes than those obtained by recent methods which also seek for reduced complexity sets. In addition, a methodology is numerically efficient, applicable to larger systems than those treated by the methods available in the literature.

The use of relational databases is increasingly present in the industry. Applications in medicine, IoT and Industry 4.0 are examples of this. Despite the great capacity and efficiency in data storage and retrieval, this type of database requires technical knowledge in specific query languages to access this information, which distances these types of applications from the non-specialized public. In this work, we propose an application of recent models in natural language processing that uses mechanisms to translate natural language into English into SQL applied to a database that stores sensor data, focusing on the concept of Industry 4.0. Paired examples of natural language phrases were generated with their corresponding SQL query to be used for training and validation. The model was agnostic in relation to the database schema, so that it only handles the input and output sequences regardless of the database structure. The data comes from historians of the typical process used in industrial settings. When training the deep neural network, we obtained a language model with an accuracy of approximately 99 \% in the validation set.

Two sensors for the characterization of the dielectric constant of different types of materials have been implemented and one of them applied in the determination of the soil water content (θ), while the other is a bio inspired antenna based on the Shiso leaf. The proposed sensors consist of in: The shape of the antenna patch is inspired by the Shiso leaf, whose scientific name is Perilla Frutescens. A complementary interdigital capacitor (CIDC) is inserted into the geometry; A microstrip antenna with a circular patch in a container printed on ABS filament, based on the complementary split-ring resonator (CSRR) structure. The operating principle is based on the displacement of the element’s resonance frequencies when the relative permittivity of the material under test (MUT) is changed. Several simulations were performed, with the data obtained, two empirical models are proposed. To validate its effectiveness, measurements were made with dielectric substrates that have the dielectric constant known in the literature. The second sensor can be applied to different types of materials, in this work it was used to determine the percentage of water contained in different types of soils. The prototypes were built and the results obtained in the measurements were compared with the results of other studies, thus validating the effectiveness of the proposed sensors.

The k-server problem in a weighted graph (or metric space) is defined by the need to efficiently move k servers to fulfill a sequence of requests that arise online at each graph node. This is perhaps the most influential online computation problem whose solution remains open, serving as an abstraction for a variety of applications, as buying and selling of currencies, reassign processes in a parallel processing for load balancing, online transportation service, probe management of oil production rigs, among others. Its conceptual simplicity contrasts with its computational complexity that grows exponentially with the increasing number of nodes and servers. Prior to this work, the Q-learning algorithm was used to solve small instances of the k-server problem. The solution was restricted to small dimensions of the problem because its storage structure grows exponentially with the increase in the number of nodes and servers. This problem, known as the curse of dimensionality, makes the algorithm inefficient or even impossible to execute for certain instances of the problem. To handle with larger dimensions, Q-learning together with the greedy algorithm were applied to a small number of nodes separated into different clusters (hierarchical approach). The local policy obtained from each cluster, together with
greedy policy, were used to form a global policy satisfactorily addressing large instances of the problem. The results were compared to important algorithms in the literature, as the Work function, Harmonic and greedy. The solutions proposed so far emphasize the increase in the number of nodes, but if we analyze the growth of the storage structure defined by Cn;k ' O(nk) It can be seen that the increase in the number of servers can be quickly limited by the problem of the curse of dimensionality. To circumvent this barrier, the k-server problem was modeled as a deep reinforcement learning task whose state-action value function was defined by a multilayer perceptron neural network capable of extracting environmental information from images that encode the dynamics of the problem. The applicability of the proposed algorithm was illustrated in a case study in which different problem configurations were considered. The behavior of the agents was analyzed during the training phase and their performance was evaluated from performance tests that quantified the quality of the displacement policies of the servers generated. The results provide a promising insight into its use as an alternative solution to the k-servers problem.

With the growth of distance education (EAD) in Brazil in its levels of education, be it medium, technical or superior and the increase in the offer of online courses made available through the Virtual Learning Environments (AVA), there is a concern with the development Of this student during the course and their learning. In order to associate the resources and activities offered in AVA with the Learning Styles classes, this work proposes to create an algorithm that can identify the behavior patterns of the EAD students through the use of tools available in the AVA and associates them with a classification of Learning Styles in order to assist the teacher in the preparation of his / her lesson plans, suggesting resources and evaluation activities arranged in the AVA that will favor the learning of its students according to the Learning Style characteristic of each one of them.

In mobile systems, information signals must be transmitted with high rates of data transmission and with high reliability. One of the possible solutions to achieve these desirable requirements is the use of systems with multiple transmitting and/or receiving antennas, featuring a MIMO system (Multiple Input-Multiple Output). However, signal processing techniques in MIMO systems depend on the Gaussianity of the transmission channel, decreasing their efficiency in non-Gaussian communication scenarios. To goal a robustness of this transmission technique, it will become imperative or to study this transmission technique in scenarios in which the channel is not exhibited by Gaussianity. In this work, a new signal detection technique in MIMO systems is proposed in scenarios characterized by non-Gaussian noise. A proposed reception technique is called Maximal Correntropy Detector (MCD) and it has been experimentally proven to treat a generalization of the MLD detector (Maximum Likelihood Detector) using the use of the correntropy complex. The MCD detector is analyzed on Gaussian and non-Gaussian channels and its performance is superior to classic detectors, without a significant increase in computational complexity. Together, this work also presents a new method of spectral sensing suitable for detecting signals that show non-linear phase variations over time. The proposed method is based on the use of theory of cyclostationary signals in waiting time, or which transformation employs signals to be a sensor in order to mitigate or effect non-linear variation of the phase. A proposed architecture is evaluated without a BPSK signal sensor and is compared to a temporal cyclostationary sensing technique. The results of the simulation tested prove the efficiency of the architecture proposal, presenting detection rates of primary users with an order of 8 dB.

Health information systems in Brazil have been designed and developed in a heterogeneous manner based on local regional characteristics, resulting in a lack of health information integrity. In this context, the Brazilian Ministry of Health pointed out the need for interoperability solutions of health information systems, noting the importance of integration with national databases and alignment with Brazilian data protection laws as well its application in education to aid with continuing education for health professionals. Therefore, this work presents Sabiá, a platform for authentication, authorization and data delivery based on user consent for health information systems in Brazil and currently applied in the context of health educational ecosystems. Sabiá's architecture is designed to achieve the following requirements: R1) Provide a Federated Identity; R2) Be a Federated Resource Manager; R3) Collect user data from different information systems and; R4) Deliver user data to systems based on user consent. Sabiá consists of three main components: 1) Sabiá Authorization Server, responsible for implementing Open Authentication; 2) Sabiá Collector, responsible for collecting data from different information systems and; 3) Sabiá Resource Server, responsible for delivering data previously authorized by the user to the systems. After analyzing historical data, R4 functionality was selected to be submitted to performance testing because it is the process that most affects overall system performance. The tests aimed at analyzing Sabiá's behavior in the heaviest scenario based on historical data. The results showed no flaws and indicated system stability and consistency, in which the user perceives a system reaction instantaneous, whose response time averages remained below 100 ms.

In this work we propose a methodology for estimating the performance and comparing control systems for autonomous sailboats. One of the biggest problem in mobile robotics research is the absence of standard tools and methodology for distribution/verification of implementations and the performance of the employed techiniques. Our approach consists of an framework for performance evaluation divided in three components: scope, metrics and scenarios. The scope define the parts of the system that are comparable and also presents a compreesive approach to define the basis navigation system for autonomous robot. The scenario represent situations that happen during sailboat missions such as keep the sailboat close to a desired waypoint (virutal anchoring), docking and tracking. Quantitative metrics are extracted by experiments in the scenario and are used to evaluate characteristics expected from control systems such as stability, energy consumption, robustness and efficiency, resulting in an initial classification of the control system. An initial implementation of the system is made using the Ardupilot framework, where some control techinique and metrics were implemented for first analisys. The level of analysis can be gradually improved by adding more control techinique and metrics.

The modern distribution system widely uses renewable energy sources (RESs) as a distributed generation that introduces problems like unintentional islanding, protection concerns, and reverse power flow. The RES intermittent characteristics can cause frequency deviations that could result in systems instabilities. In general, hierarchical structures implement these microgrids' power flow control management, composed of primary, secondary, and tertiary control layers. The primary control layer consists of inner current and voltage control loops, typically implemented via droop control loops, for the purpose of maintaining a stabilization of frequency and voltage amplitudes, as well as power-sharing. The principal drawbacks of droop control approaches are steady-state frequency and voltage deviations. For that reason, the secondary control layer is introduced to mitigate these deviations. The accurate knowledge of microgrid frequency deviations permits the implementation of proper frequency restoration. Besides, those microgrids could operate in the connected or islanding modes. Islanding or fault occurrences must be detected and treated to ensure system stability. This way, the continuous wavelet transform is employed for developing a hybrid islanding detection and a frequency monitor to fulfill the microgrid's requirements related to a secure islanding detection and a frequency restoration. In this Thesis, both systems' development is carried out, with the effectiveness of proposed methods evaluated and validated through experimental results.

This work is concerned with design of output feedback controllers for constrained linear discrete-time systems via set-invariance techniques. In this regard, Output-Feedback Controlled-Invariant (OFCI) polyhedra are used to ensure that state and input constraints are satisfied all time even in the presence of additive disturbances and measurement noise. Necessary and sufficient conditions for a polyhedral set to be OFCI are presented, which can be checked by the solution of a set of Linear Programming (LP) problems. Then, a dynamic output-feedback compensator (possibly nonlinear) is proposed, through the construction of an OFCI set, from a pair composed by a conditioned-invariant and a controlled-invariant polyhedron. Based on the available measurements and on the state of the compensator, which constitutes an estimate of the system state, a suitable control sequence can be computed to enforce the constraints. The uncertainty on the state is progressively reduced using information about the contraction of the conditioned-invariant set. An LP problem is formulated to compute a control action that enforces state and control constraints and minimizes, one step ahead, a guaranteed distance from the admissible states to the origin. The problem of tracking a constant reference signal in the presence of constant disturbances is also considered for which the conception of the tracking controller is motivated from the stabilizing controller. With the current approach, as illustrated through numerical examples, by embedding the estimator in the compensator structure and using the OFCI concept, it is possible to obtain solutions with larger sets
of admissible initial states and admissible initial estimation errors, compared to other approaches available in the literature.

The constant growth in the potential for generating energy from wind sources has been highlighted on the world stage. Global behavioral changes in search of sustainable development and consumption, replacing fossil fuels aiming at improving the quality of life and preserving the planet, are important factors for the continuation of this growth, due to the increasing need for renewable energy sources. The wind energy conversion systems that use the Doubly Fed Induction Generator - DFIG, are the most popular today. Advantages such as reduced mechanical load, simple picth control, control of active and reactive power, lower power converters that mean less switching losses, reduced converter costs, less harmonic injection into the network, put it ahead of turbine topologies with synchronous generators or induction generators with squirrel cage rotors. This thesis aims to present and demonstrate the feasibility of using a new approach on voltage control, active and reactive power control and power factor control through direct control of the voltage applied to the DFIG’s rotor, by the Rotor Side Converter - RSC. This voltage is determined analytically, solving a restricted optimization process in real time. A system of equations derived from the equivalent circuit is adopted to represent the desired operating point of the DFIG, and the solution of this system defines the values of the rotor supply voltage coordinates, in steady state. A new control strategy is proposed to reach the rotor voltage reference values, without violating the limits existing in other variables. This strategy is designed so that the speed of evolution of the voltage applied to the rotor, is governed by the dynamic evolution of the mechanical speed of the rotor. An optimization process was formulated to minimize the time of convergence of the rotor speed, restricting transitory variations in the net power generated, in order to accelerate the machine, without exceeding the current limits. Following recent trends in solutions with a reduced number of sensors, only measurements obtained from the stator sensors (voltages and currents) are used. In this way, the angular velocity and rotor currents are estimated in real time. An algorithm for estimating inductance is also included, preventing deviations from the nominal value could lead to false reference voltages or changes in reactive control. In addition, a method for defining the pitch angle and the reference speed is proposed, using the Newton-Raphson numerical solution method.

Advances in Internet of Things (IoT) applications have provided the development of new paradigms in this area, such as the Internet of Underground Things (IoUT). Research shows the potential of IoUT systems, mainly in precision agriculture, in which data it’s collected by different devices interconnected in a communication network. These data can facilitate decision-making about the management of agricultural activities. An important component in the architecture of IoUT is Underground Things (UT) responsible for sensing soil parameters, such as temperature and moisture. In view of this, the Radio Frequency Identification (RFID) technology has been shown promising for the development of UTs, mostly due to the ability of RFID tags to operate as sensors. However, the adaptation of wireless communication in IoUT systems for application in underground conditions represents a challenger yet. In this context, the proposal of the work is to develop a UT based on RFID technology for sensing soil moisture. The UT proposed is formed by a rectangular patch microstrip antenna designed to operate at Ultra High Frequency (UHF), connected to a passive UHF RFID tag chip. The RFID tag in buried in the soil and functions as a moisture sensor. To reduce the effects of soil moisture on the RFID tag, the microstrip antenna designed in the systems was encapsulated in a Dielectric Soil Attenuator (DSA). The DSA is formed by multilayers made of natural carnauba wax and applied over the antenna. Simulated and experimental results were analyzed to verify the effectiveness of the DSA on the antenna parameters at different levels of moisture and soil types. Measurements of the Received Signal Strength Indicator (RSSI) were also performed using a UHF RFID reader, and thus, finding a relationship with the moisture present in the soil. The obtained value shows that the DSA reduces variations in the antenna parameters such as gain and reflection coefficient (S11) caused by soil moisture. Statistical analyzes of the RSSI measurements showed that the proposed UT can work as a soil moisture sensor.

We propose an application of electromagnetic bandgap (EBG) for suppression or reduction of higher order resonant modes in frequency selective surfaces (FSS), not yet studied in the literature.

            The study consists of an application of a rectangular periodic array of circular air holes in FSS dielectric substrate to create rejection bands and to suppress or reduce specific resonant frequency modes. Another type of unconventional air hole used was the square shaped holes, aiming to reduce the computational effort during the simulations.

            Optimizations with genetic algorithms were implemented to optimize the physical parameters of the holes, aiming at the suppression in the desired frequency range.

            In order to prove that the prohibited bands occurred in the desired frequency bands, a Brillouin dispersion diagram was constructed verifying that the structures acted as EBG

Seven prototypes were manufactured and the results measured compared to the simulated results obtained in ANSYS HFSS. A good agreement between the results is observed. Simulations and measurements show suppression levels of up to 6 dB. The simulation time of the FSS with EBG of square holes showed a reduction of 70% in relation to the simulation time of the FSS with EBG of cylindrical holes.

The Industrial Internet of Things (IIoT) is often presented as a concept that is significantly changing the technological landscape of industries, through automation procedures and identification of relevant objects. For this, however, reliability and performance issues must be considered when providing anticipated communication services. By employing Radio Frequency Identification (RFID) in the IIoT context, several previous studies have worked to improve the efficiency of RFID communications systems, generally defining mathematical models for planning and evaluating quality. However, such models are designed based on error-free communications, which is in fact unrealistic when considering the error-prone nature of wireless communications in industrial plants. Therefore, this thesis proposes a new anti-collision algorithm for RFID together with a formal model based on Generalized Stochastic Petri Nets (GSPN) to evaluate RFID communications, modeling different possibilities of errors between readers and RFID tags. Since this proposal uses the EPCglobal UHF Class 1 Gen2 parameters as a reference, which are already adopted by the Dynamic Frame Slotted Aloha anti-collision protocol for passive RFID systems, this model can be explored to assess the performance of different RFID access protocols when assuming noisy channels, supporting better comparisons between different algorithms and protocols. The results showed that the proposed algorithm is able to present a better performability in relation to the other evaluated protocols, mainly in the presence of noisy channels if a large number of tags to be read. Simulation scenarios are defined to provide reliability and performance results when evaluating RFID tag readings, which are valuable when designing and maintaining IIoT applications.

The k-medoids model is one of the most popular clustering methods. In this work, we propose the Convex Fuzzy k-Medoids Problem (CFKM), which not only allows one object to be assigned to multiple clusters, but also allows a cluster to be represented by multiple medoids. The proposed model is convex and thus is robust to initialization. To evaluate the importance of CFKM, we compare it with another two fuzzy k-medoids models: the Fuzzy k-Medoids Problem (FKM) and the Fuzzy clustering with Multi-Medoids Problem (FMMdd), both solved by heuristics due to their computational complexity. Experiments with both synthetic and real-world data, along with an user survey, show that CFKM is not only more robust to the choice of parameters of fuzzy models, but also is the only able to reveal important aspects of inherently fuzzy data.

Vocal pathologies negatively affect the social and professional life of the sick, some of which may even lead to death if they are not treated quickly. Among the main procedures for the diagnosis of vocal disorders we can cite: laryngoscopy, perceptual-auditory evaluation, acoustic analysis of the voice, aerodynamic evaluation and self-evaluation of the voice by the patient. However, these procedures are usually invasive or inaccurate. Therefore, digital signal processing techniques have been used in the design of non-invasive systems to aid in the diagnosis of vocal tract pathologies. In this work a system of detection and classification of vocal pathologies is presented, using a classification technique based on descriptors obtained through the Correntropy Spectral Density (CSD) function, defined as the Fourier transform of the autocorrelation function. The descriptors obtained have information of second-order and higher-order statistical moments of the speech signal, through which vocal pathologies can be efficiently detected and classified. The classification is made by a neural network multilayer perceptron (MLP), performing a binary classification between normal and pathological voices and then between pathologies (edema and nodule). The classifier was evaluated by computer simulation, and the results indicate a high hit rate of detection and classification among pathologies.

Industrial plants are composed of processes that add up to thousands of variables. To ensure safety and quality of operation, these processes are monitored and alarms are configured to indicate a possible malfunction. Among the most common problems associated with industrial alarms we can mention: high occurrence of false alarms, missed alarms and chattering alarms, operator overload and alarm flooding. These problems are related to the process of selection of the monitored variables, the techniques of activation and deactivation of alarms, among other characteristics of the process and the alarm system. This work focuses on defining an approach to configure efficient and significant alarms for the operator. The approach proposed here is inspired by the workflow of a data scientist who initially needs to identify the characteristics of the databases used to then apply transformations that make the data more suitable allowing the extraction of valuable information. Many times the scientist is interested in creating a model that describes the data or makes predictions possible. This is a very similar task of alarm configuration where it is necessary to select the relevant variables and to configure the settings of each alarm in order to classify the operation of the process as appropriate or not and to help identify the fault. The approach proposed here consists of four parts: description of data, selection of variables, tuning and performance evaluation. During the description step, relevant information about the data is obtained, such as the presence of events, the number of different events, the duration of events, etc. In the selection stage, the relevant variables for detection of abnormalities are defined. The tuning of alarms is similar to a training process, where a model is built to describe the behavior of the data. Finally during the evaluation, the settings found are applied to a process history to asses whether the settings behave in a way that meets security and quality constraints. As a case study, an industrial alarm configuration was obtained for the Tennessee Eastman Process which is a simulator widely used by the academic community.

In view of the evolution of the standards of wireless communication systems, where there is a need for performance in several frequency bands, it is indispensable to design devices capable of meeting the requirements that this progress demands. The antennas are elements that play a fundamental role in these systems, in which planar microstrip antennas have been demonstrated with high evidence due to their low profile, miniaturized devices, low manufacturing cost and are easily integrated with the devices. With the objective of optimizing the occupation of physical space in the transceivers and in the end devices of the users, this work presents the proposal of a microstrip antenna based on a planar monopole with ground plane with slits for operation in multiple bands to meet new systems of wireless communications such as IEEE 802.11ax and IEEE 802.11ah. The process of design formation and design of the structure will be presented, considering the most recent research in the specialized literature. For the purposes of analysis, simulations were performed by the Finite Element method implemented through the commercial software Ansoft HFSS. Some initial prototypes were fabricated for purposes of validation of the preliminary results of this work. The experimental characterization of these prototypes was performed through a vector network analyzer. In view of the comparisons between simulated theoretical results and the preliminary experiments, it can be affirmed that a good agreement between them is observed corroborating with the studies previously developed.

This work describes the analysis of Bioinspired Frequency Selective Surfaces (BFSS) for applications in wireless systems, operating in the C band, Ku band and UWB (ultra-wideband). Simple and coupled BFSS structures are considered. The BFSSs have array elements with the four-leaf clover-shaped and maple leaf-shaped, and presented dual-band response, with the operating frequencies in the C and Ku bands. For the development of BFSSs with the ultra-wideband, a cascade structure was developed, in which a FSS with elements with patches with the shape of square loops was coupled to the BFSSs. In addition, the four-leaf clover BFSS synthesis was developed using an artificial neural network with a cascade feedforward architecture and Bayesian regularization training algorithm to obtain the specifications of resonance frequency and respective desired bandwidths. The numerical values obtained by simulations for the developed prototypes were obtained by the ANSYS HFSS software. Prototypes were manufactured and experimentally characterized. The measured results were compared with the simulated ones and a good agreement was observed.

During the productive life of most gas wells there is an accumulation of liquid at the bottom of the well which causes a pressure contrary to the flow of the reservoir, reducing the production of the well. In this way, one of the possible artificial lift methods to solve this problem is represented by the Plunger Lift. It consists of a low-cost installation and maintenance technique that uses a piston to, among other uses, increase the efficiency of liquid removal from the system. However, this method requires a well-adjusted controller to appropriately define the opening and closing periods of a motor valve installed in the wellhead for production control. Predictive control is a technique that can be applied to perform such task, performing predictions of the future behavior of the plant in order to obtain its optimum performance. In this work, the implementation and application of a linear predictive controller to a simulated well operated by Plunger Lift was performed. The controller was applied under different conditions of draining ability, with and without, of the gas present in the annular space. Results of MPC without drainage were compared with conventional versions of Plunger Lift control in the oil industry through graphical analysis and observation of performance parameters. This controller obtained better results regarding the control of the average speed of plunger ascent and accumulated production. In the comparison of MPC with and without drainage, the version capable of manipulating the gases from the annular space was able to cause the system to operate at speeds of piston ascent in safer ranges.

In this work is presented a low power instrumentation amplifier using 0.6 μm CMOS technology. It is initially shown the theoretical background about the chosen application: treatment of biopotentials. Next, we present the architectures of instrumentation amplifiers present in the literature focused on biopotentials, as well as low power and low noise configurations. A study is also carried out on the types of pseudo-resistors present in the literature. Once this is done, all the methodology used to perform this work is presented, such as the choice of the architecture used, how the system design is performed and what types of simulations are used to evaluate the performance of the system. Finally, it can be concluded that using integrated circuits with CMOS technology can enable numerous low-power portable applications that use the acquisition of biopotentials, thus justifying such design.

Partial shading decreases the power generated by solar panel arrangements. In this condition, some panels end up receiving solar radiation different from others, causing multiple peaks in P-V and I-V, resulting in losses. This work aims to reduce the losses caused by partial shading from reconfigurations of the panel arrangements, in conjunction with an adaptive P & O MPPT algorithm. The modeling of the solar panels is carried out in a Matlab & Simulink simulation environment. The modeling of the adaptive MPPT P&O method is also performed. The tested arrangements are of the Total Cross-Tied (TCT), Total Cross-Tied (SP-TCT), Bridge-Linked - Total Cross-Tied (BL-TCT), Bridge-Linked-Honey-Comb (BL-HC), plus a Magic Square (MS) equivalent for each of the arrangements presented. Magic Square is a way to rearrange the solar panels in different positions, in order to mitigate the losses caused by shading. The performances of all configurations are tested for different shading situations and demonstrated from V-P graphics. In addition to the MS, in order to make the system more efficient, it implements the adaptive MPPT P&O that has a faster response time than the standard P&O.

The Authorship Attribution, the science of inferring an author for a particular text based on their writing characteristics is a problem with a long history. In this work, it is being proposed the study of the problem of attribution of authorship in order to make it a tool of use in the distance teaching platform of the Ministry of Health (MS), AVASUS, and will be presented the techniques of analysis of text and authors' stylistic characteristics that allow authorship to be determined in significantly better indexes, in which the texts are greater than 140 characters. This proposal targets AVASUS, where students take the courses of the platform, share their interests and thoughts in the form of messages in the forums and do activities that require writing on certain topics in the area of health, these written productions are the focus of the application of attribution of authorship. The techniques studied as a proposal are a two-stage process, where in the first stage, stylometric information is extracted from the collected data set and in the second stage different classification algorithms are trained and lexical analysis techniques are applied to predict the authors of the texts. The effort is to maximize the accuracy of predictions with optimal amount of data and users under consideration. 

In this work is proposed a real-time digital simulator of power systems using a low-cost custom platform based on FPGA (field-programmable gate array) proper to perform real-time validation of traveling-wave-based transmission line protections. The operational view of the simulator is introduced by means the modeling, implementation, and simulation steps of a transmission system, which is used to highlight the simulator capability to represent high-frequency transient phenomenon taking place transmission lines. Hence, at first, the mathematical models of the power system used in the case study are presented as well as the solver design, which is developed based on the electromagnetic transients program (EMTP) proposed by Dommel. Then, the simulator characteristics, such as the hardware architecture, development software, communication strategies, graphical interface, input/output, and data export, are introduced, as well as the implementation stages of the test system. Moreover, it addresses the implementation of a relay prototype using a hardware based on DSP (digital signal processor), running a traveling-wave-based protection scheme, besides its closed-loop integration with the simulation.  A GUI (graphical user interface) is developed to set the simulation parameters, including the conditions for applying an electrical fault, and to monitor the dynamic of power system used as a case study. Off-line simulations obtained from Matlab/Simulink are used to validate the real-time results.

The world energy demand is expected to rise in the coming decades. Renewable Energy Source Distributed Generation (RES-DG) is being used to meet this increased demand. RES-DG systems can serve local loads on an isolated system (islanded operation) or connected to the power grid (grid-connected operation). Ideally, RES-DG systems should have the ability to transition smoothly between modes so that a continuous supply of power is achieved regardless of the grid state. This work proposes a communication-less control strategy for a low-voltage RES-DG system capable of operating in islanded as well in grid-connected mode and of achieving a smooth transfer between modes. An energy storage system is not required for operation in islanded mode since the energy balance is maintained by acting in the power generated by the RES. The proposed control strategy is divided in three parts: (a) the VSI control, which constitutes the inner control loops of the Voltage Source Inverter (VSI) (current, voltage and power control loops), (b) mode-related control loops (DC link, voltage at the PCC and DG system angular frequency control loops) and (c) the synchronization and islanding detection methods. The control strategy is validated through simulation of a three-phase RES-DG system. The proposed strategy presents a smooth transition between modes, with little to no perturbation on the PCC voltage, and a fast synchronization process from islanded to grid-connected mode. In addition, it also presents a fast DC link voltage control response time that is capable of maintaining the transitory small, even for large power imbalances during mode transition.

This work presents a study on Selective Frequency Surfaces (FSS) 2,5 D and 3D, for applications in modern wireless systems. Initially, 2.5D, patch-like structures are analyzed. The simple FSS are formed by elements in square and spiral turns based on the Minkowski fractal. The use of the fractal curve resulted in a miniaturization of the circuit, the appearance of a new transmission band and a decrease in bandwidth. The 2,5D FSS are formed from the simple structures that, for the formation of their elements, the coils were sectioned in eight parts, where four were printed on the upper face of the dielectric, four on the underside of the dielectric and were connected with cylindrical jumpers of 0.5 mm in diameter. The 2.5D surfaces provided miniaturization of the circuit without the appearance of new transmission bands and without decreasing the relative bandwidth. All HSS were stable with respect to the angle of incidence of the plane wave on the surface of the circuit. The simulated numerical results for the projected prototypes were obtained by Ansoft HFSS software. The prototypes were constructed and the experimental measurements of the transmission coefficients were performed, the values obtained were compared and discussed with the results of the simulations, which showed good agreement.

This work proposes the insertion of Complementary Split-ring Resonators (CSRR) in the planar filter ground. The intention is to improve some filter parameters found in the literature that have practical application, considering the growing demand for higher performance, smaller size, lower weight and cost of the devices that integrate the communication systems. The filters chosen were a bandpass filter that uses matrioska geometry for applications in the 2.45 GHz Industrial Scientific and Medical (ISM) band. The other filter was a low pass filter for digital TV applications. The idea is to insert the CSRR into the structure ground plane to increase the free attenuation of the matrioska band-reject filter and to reduce the size of the low-pass filter. In the work the numerical and experimental results are presented, for the purpose of validation of the proposed technique. A good agreement between the results is observed, these results show that the proposals presented were achieved.

The estimate of power quality indices has as primary goal provides to the monitoring systems (measurement, control, protection, etc.) information about proper levels of voltage and power quality. Besides power quality monitoring strategies, the knowledge of the grid impedance is fundamental importance to the understanding of the deviation of the electric power quality indices, providing relevant information to control and stability of the power system. In this context, in this thesis, it proposes the application of the stationary discrete wavelet packet transform (SDWPT) to estimate effective voltages and currents, as well as primary power quantities (active, total apparent, nonactive power, and power factor), and distortion power. Additionally, the analysis tools of signal processing developed in this work allowed the development of a method for estimating the grid impedance. The SDWPT provides a uniform frequency band and possesses the timeinvariance property, which is ideal for power estimation and grid impedance estimation in the real-time. The method of estimation of power quality indices was assessed and compared to the IEEE Standard 1459-2010 under different operational scenarios, where stationary synthetic waveforms and with both stationary and nonstationary power quality disturbances were emulating in an experimental setup. Regarding the grid impedance estimation, the current study was developing to identify the impedance of low voltage electric grids interconnected with distributed power generation systems. The primary goal is providing parameters to delimit the threshold of system instability in terms of the penetration level and identify islanding conditions. The validation of the proposed method was accomplished in a microgrid whose distributed generation is implemented through the photovoltaic plant of 8 kWp connected to the point of common coupling of a three-phase feeder composed by a transformer of 15 kVA. 

This work presents a methodology of how to acquire and represent knowledge through automatic logic rules for a simulated industrial plant. The initial knowledge about an industrial process can be acquired through a specialist who interprets situations present in the plant and can describe what is occurring. In the work, a way of acquiring statistical knowledge of the plant during the execution of its processes is presented, using a method of online clustering known as TEDA-Cloud being modified to improve performance. The representation of knowledge is described through the manipulation of a neural network known as CILP and its own symbology is described to represent the logical variables drawn from the process signals. The results show an efficiency in interpreting the rules and acceleration in the process of clustering and classifications of the standards that define the rules.

In this work the design of a piecewise affine proportional integral (PWA-PI) controller algorithm based on invariant set and multiparametric programming for constrained systems is proposed. We implemented the algorithm in a programmable logic controller (PLC) to control an industrial constrained level plant and analyze its behavior. Structured text routines were programmed and validated while controlling two systems with PLC. The results show that the constraints on the error, integral of the error, system output and control action are respected because PWA-PI controllers are tuned from the solution of an optimization problem. The evaluated performance indexes (such as mean square error, Goodhart,  overshoot and settling time) show that PWA-PI can be adjusted for better performance than proportional integral (PI) controller tuned by Ziegler–Nichols (Z–N) rules. In the analyzed cases, a settling time of 108 s was obtained, whereas PI controller with Z–N rules presented a 179 s settling time. All of the analyzed performance indexes that we used to evaluate both controllers show PWA-PI as a better controller for constrained systems.

This work aims to present new proposals for equalization and beamforming associated with modern digital communication systems. Three proposals are presented in which the first is characterized by an innovative neural equalizer structure called here the Butterfly Neural Equalizer (BNE). The BNE can be applied to linear and non-linear problems involved in intensity-modulated and direct-detection (IM-DD). The second proposal is characterized by an adaptive beamforming scheme applied to Orthogonal Frequency-Division Multiplexing (OFDM) systems. The OFDM technique and variants are being used in various communication systems such as LTE, digital TV, and others. Finally, the third proposal presents a neural beamforming structure called Butterfly Neural Beamforming (BNB) that can be applied to several communication systems. Results associated with the three proposals are presented and show the feasibility of the techniques in several channel scenarios and digital modulation schemes.