基于SVM和RBFN的汽车主动降噪系统传感器故障诊断

Sensor Fault Diagnosis of the Automobile Active Noise Control System Based on SVM and RBFN

汽车主动降噪系统的工作依赖于多个噪声传感器,一旦传感器发生故障,将严重影响降噪效果。为保证汽车主动降噪系统的性能,提出了由支持向量机(SVM)预测模型和径向基神经网络(RBFN)预测模型构成的传感器故障诊断系统,SVM模型判断是否发生传感器故障,RBFN模型则利用各传感器间的信息冗余关系定位故障传感器并对其信号进行重构。仿真结果表明,该诊断系统可有效实现汽车主动降噪系统中的传感器故障诊断及信号重构。与传统的汽车主动降噪系统相比,引入传感器故障诊断系统可保证更稳定的降噪性能。

The normal operation of the automobile active noise control system depends on multiple sensors. Once there is any sensor failure, it will severely affect the noise reduction effect. In order to guarantee the automobile active noise control system＇s performance,a sensor fault diagnosis system based on support vector machines（ SVM） and radial basis function networks（RBFN）is put forward. The SVM model monitors sensor fault,meanwhile the RBFN models locate the fault sensor and reconstruct its signal based on the information redundancy between each sensor. Simulation results prove that the proposed diagnosis system could effectively diagnose any sensor fault in the automobile active noise control system as well as reconstruct fault sensor＇s signal. Compared to the conventional automobile active noise control system,introducing the proposed diagnosis system provides higher reliability of noise reduction.