An approximation of predictive maintenance for a plane propeller through sensorization, vibration analysis and Machine Learning

Abstract

In industrial machinery, failure will eventually occur, either in the short term due to manufacturing defects, or in the long run due to accumulations of debris, deterioration of internal parts, or purely from wear and tear. Machinery, systems, and industrial lines force factory operators to deal with seemingly endless maintenance and repair cycles, particularly when undetected faults in machinery cause catastrophic failure. Maintenance and repair schedules can be more efficiently planned if we study the behaviour of the machine before the failure occurs. That it's known with the name of Predictive Maintenance. With it, it's possible to save significant amounts of money in an industrial environment by less production stops and less useless maintenance mechanics work hours. Predictive maintenance has more and more name in the production lines and is also used in the Aeronautical Industry to be able to foresee what failures are going to arise in the aircraft parts. With this, it's possible to proceed to carry out necessary maintenance activities before failures occur and extend the life of the machinery. One way to study this maintenance is by analysing machinery vibrations. These mechanical parameters can determine abnormal behaviour and thus alert failures. The longer the behaviour of a machine is studied, the easier it is to draw conclusions due to the large amount of data collected. The objective is to know the machine, create behaviour models, detect anomalies to carry out maintenance and avoid failures.