Fault detection of nonlinear process based on adaptive KLPP of MEWMA
For the small disturbance problem in nonlinear dynamic process,a fault detection of nonlinear process algo-rithm based on the adaptive kernel locality preserving projections(KLPP)of multivariate exponentially weighted moving average(MEWMA)is proposed in this paper.Firstly,a data matrix with dynamic characteristics is constructed,and the kernel function is introduced to execute the KLPP algorithm.Secondly,the feature vectors extracted by the KLPP are whitened.The MEWMA is used to predict the mean shifts in nonlinear dynamic process.Finally,an adaptive monitoring statistic is constructed by combining the estimated mean shift with the whitened feature vectors,and its control limit is determined by using the kernel density estimation.The proposed monitoring scheme is applied to a nonlinear numerical example and the TE process for simulation analysis.The simulation results show that the method is feasible and superior.
万方数据
维普
