A novel fault diagnosis technique based on model and computational intelligence applied to vehicle active suspension systems

Abstract

This paper introduces a novel approach to design fault detection and diagnosis<br><br>systems by using computational intelligence techniques for industrial<br><br>plants and processes. According to the increasing development of machine<br><br>learning algorithms, selecting an appropriate algorithm for identifying and<br><br>estimating the size of faults in dynamic systems can turn in to a challenge.<br><br>A notable point in this regard is that although the available intelligent algorithms<br><br>have considerable advantages and strengths, they are also vitiated by<br><br>some weaknesses too. Therefore, using a proper and intelligent structure<br><br>through combining common algorithms can be considered as a new and<br><br>appropriate strategy to increase the efficiency and accuracy of fault detection<br><br>and diagnosis systems. Given the importance of improving safety and reliability<br><br>in vehicles, the proposed approach based on dynamic behavior is implemented<br><br>on a simulated nonlinear active suspension system model. The<br><br>results obtained from multiple tests indicate that the proposed method with<br><br>a novel architecture has managed to increase the accuracy and reduce the<br><br>announcement of false alarm and computational load, compared with other<br><br>conventional methods. This achievement leads to improvement of diagnostic<br><br>systems' performance.