An Feature Data Extraction Method Based on BPCA-SPCA

With the increasing integration of systems and the rapid development of sensor monitoring technology, the types of monitoring data are more diverse and the number of samples are larger in flight test and operation than before. Due to the large volume and high complexity of monitoring data, the system fault diagnosis and prediction based on monitoring data is inefficient. In order to solve the problem of monitoring data redundancy, a feature extraction algorithm based on BPCA-SPCA is proposed. A system with high failure rate is selected as an example to extract the features of the system monitoring data and realize the verification of data sparseness. The experimental results show that based on BPCA-SPCA the principal components extracted monitoring data can effectively express the features of the original data, and the feature sparse vector can provide effective input information for subsequent fault diagnosis and indirectly improve the efficiency of fault diagnosis.