Abstract: Fault detection based on the basic SVDD (support vector data description) method is not good at the processing of large sample data， and the modeling and process monitoring is time-consuming. The sparse characteristics of the original data in high dimension space was studied， according to which the first k high dimensional distribution edge data points were selected to carry out the SVDD modeling. Through theoretical derivation and simulation analysis， it was showed that the modeling and detection speed could be effectively improved by the proposed method， and the large sample data could be modeled by using the selected small sample， which could handle the classification problems of SVDD method on solving large sample data; meanwhile， this method did not affect the accuracy of fault detection. The effectiveness of the proposed method was illustrated by applying it to the monitoring of TE process.

Key words: sparsity, SVDD, sparsity SVDD, fault detection