Autonomous Driving evolves from future vision to reality. One major challenge is transitioning the driver’s responsibility for condition monitoring to an automated system which keeps track of the system’s health. Chassis components have a huge impact on the vehicle’s stability. Therefore, monitoring and diagnosing any defects is a prerequisite for automation levels 4 to 5. We observe increased data availability in modern vehicles as well as advances in the field of machine learning. Many data-driven fault detection and isolation (FDI) systems use machine learning algorithms with hand-crafted features. Instead, Convolutional Neural Networks (CNN) can replace the cumbersome and error-prone process of feature engineering. This thesis investigates the general applicability and influencing factors of CNN for the automated diagnosis of chassis components, i.e. dampers. It is shown that CNN can replace feature engineering with their feature learning capabilities. Additionally, the proposed approach shows improved fault classification results.
«
Autonomous Driving evolves from future vision to reality. One major challenge is transitioning the driver’s responsibility for condition monitoring to an automated system which keeps track of the system’s health. Chassis components have a huge impact on the vehicle’s stability. Therefore, monitoring and diagnosing any defects is a prerequisite for automation levels 4 to 5. We observe increased data availability in modern vehicles as well as advances in the field of machine learning. Many data-dr...
»