Banca de DEFESA: JOAO PABLO SANTOS DA SILVA

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
STUDENT : JOAO PABLO SANTOS DA SILVA
DATE: 27/08/2025
TIME: 14:30
LOCAL: Remota - Google Meet
TITLE:

K-Nearest Neighbors for Anomaly Detection and Predictive Maintenance in Water Supply Systems

 

KEY WORDS:

Neural Networks, Machine Learning, K-Nearest Neighbors, Hydraulic Anomaly Detection, Smart Cities.

 

 

PAGES: 11
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUBÁREA: Eletrônica Industrial, Sistemas e Controles Eletrônicos
SPECIALTY: Automação Eletrônica de Processos Elétricos e Industriais
SUMMARY:

 This work proposes the application of machine learning and neural network methodologies with the aim of optimizing the operation of water supply networks, with an emphasis on detecting hydraulic anomalies, such as leaks and failures in pumping systems. Given the growing need for efficient water resource management, aligned with the concepts of smart cities, aiming at sustainability, reduction of losses, and improvement in water quality, predictive models were developed to identify anomalies and propose solutions that increase operational efficiency, reducing failures and maintenance costs, based on the use of data from sensors installed in a supply network. The method used in this study employed the K-Nearest Neighbors (KNN) algorithm for correlational analysis and failure prediction. The algorithm was applied to predict pump failure based on sensor data and integrate different approaches to improve the accuracy of the results. The model emphasizes simplicity, maintaining high accuracy with the fewest possible input parameters. The results demonstrated that the approach enables early detection of anomalies,rapid decision-making based on real-time data, and the generation of previously unidentifiedoperational insights, as well as a reduction in the rate of operational failures, provingthe effectiveness of the system. It was concluded that the integration of machine learning and neural networks optimizes water network management, ensuring continuous supplyand water quality. The model developed can be adapted to different contexts, reinforcing its relevance for industrial applications and public managers.

COMMITTEE MEMBERS:
Presidente - 350693 - ANDRE LAURINDO MAITELLI
Interno - 347628 - ADRIAO DUARTE DORIA NETO
Interno - 1149567 - ANDRES ORTIZ SALAZAR
Externo à Instituição - JOÃO MARIA ARAÚJO DO NASCIMENTO - IFRN
Externo à Instituição - OSCAR GABRIEL FILHO - PETROBRAS