Data-driven fault diagnosis for an automobile suspension system by using a clustering based method

This paper concerns the issues of fault diagnosis and monitoring for an automobile suspension system where only accelerator sensors in the four corners of the car body are available. A clustering based method is proposed to detect the fault happened in the spring, and the Fisher discriminant analysis is applied to isolate the root factor for the fault. Different from most of the existing approaches, the pure data-driven characteristic enables this method to serve as an on-line fault diagnosis and monitoring tool without suspension model or fault features known as a prior. Moreover, this method can classify different reductions in the spring coefficient into one fault rather than different faults. The effectiveness of the proposed method is finally illustrated on an automobile suspension benchmark.