(A) study on model-based fault diagnosis and countermeasures of vehicle suspension system using machine learning technique

Abstract

This dissertation deals with model-based fault diagnosis algorithm and fault countermeasure of vehicle suspension system using machine learning technique. The suspension system of the vehicle consists of three sprung mass accelerometers, two unsprung mass accelerometers, suspensions and tires. This study covers suspension sensor and magnetorheological (MR) damper fault diagnosis and the countermeasure method for suspension sensor fault. In this dissertation, first, a model-based fault diagnosis method is used to generate the residuals in response to sensor fault. This residual reacts independently to the fault of the sprung mass accelerometer and unsprung mass accelerometer. To evaluate the residual generated by the model-based fault diagnosis method, we use a support vector machine, one of the machine learning techniques. By combining model-based fault diagnosis with machine learning, this study designs fault diagnosis algorithm without threshold tuning process. Next, based on the robust Takagi-Sugeno (T-S) fuzzy observer, the fault of MR damper is diagnosed. Using the T-S fuzzy modeling technique, the nonlinear characteristics of the MR damper is modeled and the MR damper fault can be diagnosis for variety current range. Finally, to overcome the limitation that conventional relative velocity estimation methods cannot be used in the event of unsprung mass sensor fault, the alternative relative velocity estimation method is proposed using wheelbase preview assumption. Using this fault countermeasure algorithm, the vehicle is able to keep the performance of suspension while the sensor fault occurs. In addition, in the appendix, indirect TPMS using the tire's frequency characteristics are designed using ABS sensor mounted on the vehicle wheel. By combining deep neural network and frequency analysis based indirect TPMS, the proposed method overcome the limitation of previous researches.