Classificação automatizada de falhas tribológicas de sistemas alternativos com o uso de redes neurais artificiais não supervisionadas

摘要

Preventing, anticipating, avoiding failures in electromechanical systems are demands thathave challenged researchers and engineering professionals for decades. Electromechanicalsystems present tribological processes that result in fatigue of materials and consequent lossof efficiency or even usefulness of machines and equipment. Several techniques are used in anattempt to minimize the inherent losses of these systems through the analysis of signals fromthe equipment studied and the consequences of these wastes at unexpected moments, such asan aircraft in flight or a drilling rig in an oil well. Among them we can mention vibrationanalysis, acoustic pressure measurement, temperature monitoring, particle analysis oflubricating oil etc. However, electromechanical systems are complex and may exhibitunexpected behavior. Reliability-centric maintenance requires ever faster, more efficient androbust technological resources to ensure its efficiency and effectiveness. Failure Mode EffectAnalysis (FMEA) techniques in equipment are used to increase the reliability of preventiveand predictive maintenance system. Artificial neural networks (ANNs) are computationaltools that find applicability in several segments of the research and signal analysis, where it isnecessary to handle large amounts of data, associating statistics and computation in theoptimization of dynamic processes and a high degree of reliability. They are artificialintelligence systems that have the ability to learn, are robust to failures, and can deliver realtimeresults. This work aims at the use of artificial neural networks to treat signals from themonitoring of tribological parameters through the use of a test bench to simulate contactfailures in an air compressor in order to create an automated fault detection and classificationsystem, unsupervised, with the use of self-organized maps, or SOM, applied to the preventiveand predictive maintenance of electromechanical processes.