Fault Detection in an Electro-Hydrostatic Actuator Using Polyscale Complexity Measures and Bayesian Classification

This article presents a novel approach for fault detection in a hydraulic actuation system. The fault of interest is the internal leakage of the actuator, which may often be caused by the wearing down of the piston seal. Bayesian classification and polyscale complexity measures are used in this article. Bayesian inference provides a probabilistic framework for classification that combines prior knowledge with observed data to update the probability distribution of the classification parameters. It results in a posterior distribution that reflects the updated knowledge. This allows for more accurate and reliable fault detection, especially in cases where the available data are uncertain or noisy. In order to extract features from the acquired signals, a polyscale measure known as variance fractal dimension (VFD) is employed. VFD measures are employed as features for Bayesian classification, allowing for distinguishing faulty conditions. The efficacy of the proposed method is demonstrated using experimental data, achieving an accuracy of 93.75%. Consequently, the proposed method is considered to be promising for fault detection in fluid power applications.