Internal Leakage Detection in Electrohydrostatic Actuators Using Multiscale Analysis of Experimental Data

Abstract

One of the main faults that may happen in electrohydrostatic systems is the actuator internal leakage that occurs due to wearing in the piston seal. This paper focuses on detecting the internal leakage using multiscale analysis of experimental data measured from an electrohydrostatic actuator (EHA) test rig. Multiscale techniques are the strong tools in analysis of time series, as they are able to extract more useful information about dynamical systems as compared with single-scale methods. In this paper, several multiscale measures are obtained from the actuator pressure signal of an EHA testbed in both healthy and faulty operating modes. The measures are correlation fractal dimension, variance fractal dimension, maximal Lyapunov exponent, average value of correlation entropy, and wavelet detailed and approximation coefficients. Sensitivity of each measure to the effect of the internal leakage is quantified by calculating the percentage of change of faulty measures with respect to those of the healthy operating mode. The percentage of change in the mean value of correlation entropy and level five wavelet detailed coefficient indicated that these two measures are appropriate indicators to detect different levels of actuator internal leakage in EHA systems. In contrast, the correlation fractal dimension, the variance fractal dimension, and maximal Lyapunov exponent did not exhibit reliable sensitivities to the internal leakage.