Fault Diagnosis Using Improved Discrimination Locality Preserving Projections Method Based on Cosine Distance for Industrial Process

Nowadays, with the continuous complexity of the scale of the process industry, the safety problems in industrial production have attracted numerous people's attention. However, the current industrial data is characterized by high dimensionality, nonlinearity and strong coupling, and it is difficult for traditional data dimensionality reduction methods to extract important information from them. Aiming at this problem, a fault diagnosis method based on Cosine Distance Improved Discrimination Locality Preserving Projections (DLPP-C) is proposed in this paper. This method uses cosine distance instead of Euclidean distance, which solves the problem that Euclidean distance is easily affected when calculating high-dimensional data. Firstly, the proposed method is used to extract important information of high-dimensional data. Secondly, data of dimensionality reduction is classified by AdaBoost classifier. Finally, the Tennessee Eastman Process (TEP) is used for simulation experiment. The experimental results show that the effectiveness of the proposed method.