Comparative study of machine learning and deep learning techniques for fault diagnosis in suspension system

Comfort and stability are the prime reasons to own an automobile (car). Suspension system of an automobile plays a major role in providing comfort, stability and control. Over a period of time, internal components in the suspension system exhibit faults due to fatigue and wear. Hence, it is essential to perform fault diagnosis such that the performance of the suspension components is restored. However, high instrumentation cost, skilled labor requirement and expertise in the particular field of study are certain drawbacks of traditional fault diagnosis techniques. Such challenges have made industrialists and the research communities look for advanced fault diagnosis techniques. Advancements in machine learning and deep learning techniques can be used to fulfill the need of a high degree intelligent fault diagnosis system. In the current study, the performance of machine learning (ML) classifiers are compared with the performance of deep learning (DL) models and the best performing model among them is adopted to detect faults in the automobile suspension system. A total of eight test conditions, namely strut external damage, strut mount failure, ball joint worn out, control arm bush worn out, control arm ball joint worn out, strut worn out, low wheel pressure and good condition, were considered in the study. The vibration measurements were acquired for three load conditions. Among all the techniques considered for classification, the pre-trained VGG16 model outperformed other DL and ML models with an overall classification accuracy of 98.10%.