Batch Process Fault Monitoring Based on LPGD-kNN and Its Applications in Semiconductor Industry

摘要

The abstract should summarize the contents of the paper and should contain at least 70 and at most 150 words. It should be written using the abstract In order to address the high dimensionality and multiple conditions of batch process data, a method of LPGD-kNN is proposed in this article. Firstly, standardization of local neighborhood (LNS) is processed to overcome the pretreated data character of multiple conditions. Then, Locality Preserving Projection (LPP) which can extract adaptive transformation matrix of the high modal batch data to form a new modeling data is applied in this method. Different from the traditional k-Nearest Neighbor (kNN) which extracting similarity information by Euclidean distance, Geodesic distance based kNN method is proposed for fault detection with constructing statistical indicators. Improved Dijkstra (IDijkstra) algorithm is proposed to calculate the Geodesic distance between each training data, so as to characterize the shortest distance of the nonlinear data within local areas accurately. Finally, the improved LPGD-kNN algorithms is applied in semiconductor industry examples and the effectiveness of the proposed method has been verified by comparison.