Intelligent Diagnosis of Gear Transmission Systems of Robots Based on a Digital Model

Abstract

In a highly competitive industrial context, the maintenance field has relied in recent years on various tools such as artificial intelligence algorithms (AI algorithms), in order to achieve its main objective, which is the continuous availability of machines. Previous applications of these algorithms in machine monitoring were mainly relying on historical operating datasets, which restricted the reliability and accuracy of diagnosis. This is mainly because the collected databases that will be exploited in AI algorithms are often difficult to obtain, especially for complex machines like robots. For this reason, this chapter proposes a new approach for monitoring of gear systems inside robots using AI algorithms, but which avoids the requirements of experimental data and replaces them with datasets generated by a numerical model. In order to realize this approach, a robot model able to simulate the joints vibration behavior without and with the presence of gears defects, is developed. Then, from the numerical simulations, fault indicators well known in the literature of gear systems diagnosis, are extracted. Finally, three common classifiers, SVM (multiple kernel support vector machine), DT (decision trees) and KNN (k-nearest neighbor algorithm) are driven by the most relevant simulated features and subsequently validated by experimental datasets.