Study on Adaptive Fault Threshold of Hydraulic Actuation System
Model-based fault detection (FD) depends on its fault decision-making, in which the threshold plays a very important role because it is the basis to determine whether a system fails or not through comparing the discrepancy between the actual and the predicted system output of observer. Conventional FD utilizes the constant threshold to determine system fault while it is easy to lead to fail to report or false alarm. Considering the hydraulic actuation system is a typical nonlinear system, this paper investigates the influence factors that affect the residual error, presents a new adaptive threshold based on wavelet and neural network (ATWN), in which the wavelet technology can filter the influence of noise and disturbance around the system and the neural network has robustness to nonlinear and uncertain factors. Application indicates that ATWN can approach the variance of residual error under some disturbances and noise existing and the controller failure can be detected successively with ATWN.
adaptive fault threshold wavelet analysis neural network hydraulic actuation system
Wang Shaoping Liu Hongmei Shen Guoxun Cui Mingshan Mileta M. Tomovic
Department of Mechatronic Engineering, School of Automation, Beijing University of Aeronautics and A Department of Mechanical Engineering Technology, Purdue University, USA
国际会议
北戴河
英文
1039-1043
2007-06-06（万方平台首次上网日期，不代表论文的发表时间）