A Data-Driven Adaptive Fault Diagnosis Method Based on Multi Evaluation Index Traction

In order to match suitable fault diagnosis methods that can meet performance evaluation index for diagnosing faults based on different system data, it is usually necessary to conduct comparative analysis and optimization research on different data processing, feature extraction, and fault diagnosis algorithm models. However, there are disadvantages such as relatively long development time and poor universality. Starting from the overall architecture of system data fault diagnosis and based on existing resource integration, this study proposes an adaptive fault diagnosis method with APR index as the optimization objective, which is a comprehensive performance evaluation index based on fault diagnosis accuracy, fault precision, and fault recall. Simply input discrete feature data or time continuous data, and typical time-domain and frequency-domain fusion feature extraction will be automatically performed. Based on the extracted feature data, adaptive learning methods will be used for data processing and fault diagnosis algorithm model combination encoding, and the optimal algorithm combination model will be selected through adaptive evaluation based on encoding decoding and optimization. Finally, the optimal algorithm model combination that meets the given APR threshold or reaches the iteration termination condition can be obtained, as well as the diagnostic results of system data from multiple measurement points, and the adaptability assessment of the input system data based on the obtained optimal APR index.