Diagnosis functions are commonly used within modern vehicles. These functions are applied to subsystems as the engine, exhaust system or electric system. An important component regarding safety as well as driving characteristics i s the damper. However, damper defect diagnosis is not performed in production cars. Al so in literature, there are only very few approaches that address damper diagnosis during driving. This paper shows an approach to diagnosing damper defects with machine lea rning. A support vector machine (SVM) architecture is used to classify measureme nt data with different settings of the damper functionality. Therefore, test drives we re conducted with a vehi- cle, equipped with semi-active dampers and measurement data of longitudinal and lat- eral accelerations as well as wheel speed signals was gathered. Several signal features are calculated to characterize the measurement data of each sensor. The most meaning- ful features are explained in this paper. Feature reduction is required to maximize the classification result. This is done by three different methods and the performance of each technique is studied. Finally, a linear SVM classifier is trained that allows the detection as well as the isolation of the vehicle’s damper stat es.      «

Diagnosis functions are commonly used within modern vehicles. These functions are applied to subsystems as the engine, exhaust system or electric system. An important component regarding safety as well as driving characteristics i s the damper. However, damper defect diagnosis is not performed in production cars. Al so in literature, there are only very few approaches that address damper diagnosis during driving. This paper shows an approach to diagnosing damper defects with machine...     »