Diagnóstico de fallos en cajas de engranajes mediante la aplicación de diferentes técnicas de inteligencia artificial

Laburpena

Gearboxes are vital in the transmission of movement in industrial machinery, an adequate diagnosis is highly demanded by the influence in the economy of the company, to reduce operational costs, support maintenance decisions, and improve the safety level. In the present work the diagnosis was made in gearbox failures based analysis of vibration signals through the application of different techniques of artificial intelligence. For this purpose, four databases were established, three databases of vibration signals were acquired in the laboratory and a fourth public database; two bases were in spur gears, and one in helical gears, the fourth database combines spur and helical gears. Subsequently to each signal of the databases the attributes were extracted in the domains of time, frequency and time-frequency; then three diagnostic systems were developed: system one, evaluated through statistical tests the best classifier between four neural networks and random forest, system two, evaluated the best classifier of system one with the classifier of vector support machines and a neural network, the system three evaluated as attributes and domain influences the result of the random forest classifier. After the execution of the tests in the diagnostic systems it was determined that random forest was the technique of artificial intelligence that had better performance for the classification of faults in the gearboxes.