Banca de QUALIFICAÇÃO: THIAGO MEDEIROS DE FARIAS

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : THIAGO MEDEIROS DE FARIAS
DATE: 02/06/2026
TIME: 14:00
LOCAL: Videoconferência
TITLE:

USE OF COMPUTATIONAL INTELLIGENCE FOR
FAILURE PREDICTION IN GAS-LIFT
COMPRESSORS USING HISTORICAL AND
REAL-TIME INDUSTRIAL DATA

 

KEY WORDS:

computational intelligence; failure prediction; gas-lift compressor; historical
industrial data; predictive maintenance; PIMS.

 

PAGES: 52
BIG AREA: Ciências Exatas e da Terra
AREA: Ciência da Computação
SUMMARY:

The oil and gas sector plays an important role in energy supply, job creation, and the
support of several productive chains. In this context, production continuity depends on
the safe operation of large-scale equipment, which is subject to high maintenance costs,
the replacement of often imported components, and complex logistics for intervention in
critical assets. Among this equipment, the gas-lift compressor stands out. Gas lift is an
artificial lift method in which pressurized gas is injected into the well to reduce the weight
of the fluid column and help the oil rise to the surface. In this process, the compressor is
responsible for supplying the gas under the conditions required for injection. Failure of this
equipment may reduce production, cause unplanned shutdowns, and increase operational
risk. At the same time, PIMS systems, used to store and organize industrial process data
over time, gather information from variables such as pressure, temperature, flow rate,
and vibration. These historical data can support more complete analyses of equipment
behavior and serve as a basis for failure prediction models.
The research aims to evaluate computational intelligence techniques capable of
identifying, in advance, signs associated with failures in gas-lift compressors, using historical
industrial data as the main source of analysis. The study involves a literature review,
definition of the most relevant variables, organization of the data for analysis, comparison
between models developed from data and approaches built on the technical knowledge
of specialists regarding equipment operation and failure, as well as validation with these
professionals. As partial results, the research problem, the theoretical review on predictive
maintenance and computational intelligence, the characterization of the studied equipment,
and the initial definition of the industrial variables that may compose the modeling
database have already been consolidated. In this way, the study is expected to increase the
lead time of alerts, reduce unplanned shutdowns, and support safer and better-grounded
maintenance decisions, contributing to asset reliability.

 

COMMITTEE MEMBERS:
Presidente - 2180207 - ITAMIR DE MORAIS BARROCA FILHO
Interno - 2668551 - ANDRE MORAIS GURGEL
Interno - 3374361 - JEAN MARIO MOREIRA DE LIMA