基于SVM预测器的传感器故障诊断与信号恢复研究

Research on Sensor Fault Detection and Signal Recovery Based on SVM

支持向量机(SVM)是一种新兴的基于统计学习理论的机器学习方法。简要介绍了SVM回归原理,据此建立了基于SVM的时间预测器并用于传感器的故障诊断和信号恢复,阐述了具体的实现方法和步骤。仿真结果表明:SVM预测器有效地克服了神经网络的不足,能准确预测和跟踪传感器的输出信号,并在传感器发生故障后一定的时间段内能较精确的估计传感器的正常输出。

Support vector machine (SVM) is a new machine learning method based on statistic learning theory, it has excellent performance compared with other non-linear regression, such as neural network. We introduced the fundamental theory of SVM regression, proposed a predictor based-on SVM regression for sensor fault detection and signal recovery, and presented the principle of the predictor and its on-line algorithm. The simulation results show that this method can detect sensor fault and recover sensor signal successfully.