Diesel engine fault diagnosis data enhancement method based on decomposition-generation-reconstruction generation strategy

The invention discloses a diesel engine fault diagnosis data enhancement method based on a decomposition-generation-reconstruction generation strategy, and is suitable for the field of fault diagnosis of diesel engine equipment. According to the method, the generation task of the complex multi-source impact signal is simplified, and the purpose is to improve the quality and diversity of the generated signal from more detail and local levels. Firstly, standardized single-impact sub-signals of different fault types are effectively extracted from a multi-source impact signal through a shared window variational time domain decomposition method, and the comparability between the sub-signals is ensured. Then, a multi-condition variational adversarial auto-encoder network is constructed, fault types and window labels are used as condition information, the center of potential space vectors is accurately constructed, and therefore multi-label type high-quality single-impact sub-signals are generated; finally, the gene