基于SVM和Kalman预测的非线性系统故障预报

针对Kalman预测在非线性系统故障预报中预测误差较大的问题.提出一种基于支持向量机预测新息的Kalman预测方法.根据未知非线性系统的典型变量分析子空间模型进行Kalman预测.采用支持向量机时间序列预测算法预测未来时刻的新息,利用新息进行Kalman单步和多步预报.在连续搅拌反应器上的仿真研究表明:所提出方法能准确地预测较长时间段内故障过程的劣化趋势,预知可能发生的故障,使操作人员有时间采取必要措施消除故障隐患.

Nonlinear system fault prognosis based on SVM and Kalman predictor

For Kalman predictor performing worse in nonlinear system fault prognosis,a Kalman predictor with innovation predicted by support vector machine is presented.Kalman predictor is conducted with canonical variate analysis subspace model.Time series support vector machine is trained to predict future innovation.Then,one step ahead and multi-step ahead fault prognosis are implemented by Kalman predictor with future innovation. The simulation results on CSTR indicate that the proposed method can predict the dete...