Automatic learning algorithm for troubleshooting in hydraulic machinery

Abstract

In Peru, there are many companies linked to the category of heavy machinery maintenance, in which, on the one hand, although it is true they generate a record of events linked to equipment maintenance indicators, on the other hand they do not make efficient use of these data generating operational patterns, through machine learning, that contribute to the improvement of processes linked to the service. In this sense, the objective of this article is to generate a tool based on automatic learning algorithms that allows predicting the location of faults in hydraulic excavators, in order to improve the management of the maintenance service. When developing the research, it was obtained that the algorithm that assembles bagged trees presents an accuracy of 97.15%, showing a level of specificity of 99.04%, an accuracy of 98.56% and a sensitivity of 97.12%. Therefore, the predictive model using the ensemble bagged trees algorithm shows significant performance in locating the system where failures occur in hydraulic excavator fleets. It is concluded then that it was possible to improve aspects associated with the planning and availability of supplies or components of the maintenance service, also optimizing the continuity and response capacity in the maintenance process.