Fault detection and fault-tolerant control of single-rod electrohydrostatic actuated system
Mirbeygi Moghaddam, Amirreza
Single-rod electrohydrostatic actuators present an efficient alternative to valve-controlled systems. However, such systems have significant nonlinearities and as is the case with hydraulic circuits, the possibility of faults exists in the system. Since these faults cannot be seen directly and have significant effects on the overall behaviour of the Electrohydrostatic Actuator (EHA), it is of crucial importance that they are detected and the performance of the EHA is restored. These behavioural complexities along with its asymmetrical dynamics, make the control task of the single-rod EHA challenging. In this thesis, by acquiring accurate signals from a single-rod EHA using a novel fuzzy denoising method a fault detection analysis is performed in a multi-fault environment. To do so Variance Fractal Dimension (VFD), Length Fractal Dimension (LFD) and wavelet detail coefficients are utilized. Also, the severity of the faults (bulk modulus and internal leakage) is associated with the aforementioned measures. Building upon this comprehensive analysis, a fault decision algorithm is developed to allocate a quantitative value to the internal leakage severity in the system if such fault occurs. With regards to control, first, a healthy condition and a faulty condition Fractional-Order Proportional Integral Derivative (FOPID) controller are developed and optimized using the Modified Nelder-Mead (MNM) algorithm. Using these two FOPID controllers, then, the fault-tolerant controller is designed which uses the generated degree of fault variable from the fault detection algorithm to assign a weight to each of the inputs from the two controllers through a fuzzy inference system. The effectiveness of the fault detection and control strategies is demonstrated in several experimental results.
fault-tolerant control, fault detection, electrohydrostatic actuator, fractional-order control, fuzzy control, fractal dimensions, denoising, wavelet transform