论文部分内容阅读
An algorithm of data selection kernel least squares regression(DS-KLSR)for fault isolation are proposed in this paper.KLSR can increase the distance be tween two different kinds of faults.To deal with the linearly non-separable problems,the nonlinear fault data sets are mapped into high-dimensional featur e space to implement the linear separability of different faults.To increase the distance of the different faults,pulling technique is used to drag the data sets along the opposi te directions of the different faults.At last,on the basis of KLSR model,the data selection algorithm is proposed to obtain an efficient framework for fault is olation.The proposed method is used in electro-fused magnesia furnace process to perform fault isolation,and the results show the effectiveness.