Research of Sensor Fault Detection and Diagnosis for EMB System Based on CSA-SVM Model

Abstract

Electro Mechanical Brake (EMB) is a high efficiency pure electric vehicle braking system which is based on the technology of Electronic, machinery, communication network. Because of the expensive cost and immature key technology in safety and reliability, the products cannot be mass-produced on the market at present. Electronic control of the EMB system needs a variety of sensors information feedback, therefore, how to correctly detect and diagnose the faults of the sensors is one of the important problems for the development of pure electric vehicles. Based on three-loop control architecture model of EMB system, the sensor fault detection model is established on the basis of Support Vector Regression (SVR), and the sensor fault diagnosis model is established on the basis of Support Vector Classification (SVC). In order to further improve the Support Vector Machine (SVM) for fault classification accuracy and fault detection reliability, the parameters of SVM can be optimized by using Clonal Selection Algorithm (CSA) and the modified CSA-SVM model is established. Simulated result of experiment indicates that the proposed CSA - SVM fault detection rate is increased than the traditional SVM by 62.5% and the fault classification accuracy is increased by 10% which laids a solid foundation of fault-tolerant control technology for the EMB system.