98 / 2025-04-16 18:12:45

Deep Dynamic Residual Space Contribution Based Fault Diagnosis

GRU,DPCA,RBC,fault diagnosis

In complex dynamic processes, fault information will propagate among various variables over time, making it difficult for traditional fault diagnosis methods to locate the root cause of the fault. To address this issue, this paper proposed a fault diagnosis method based on deep dynamic residual space (DDRS) contribution. The model used for fault information extraction mainly consists of a gate recurrent unit (GRU) based autoregressive network, and extracts the remaining process information from the predicted residuals through dynamic principal component analysis (DPCA). The final fault diagnosis result is achieved through the construction of reconstruction based contribution (RBC) on the trained DPCA. The effectiveness of the proposed fault diagnosis method was verified through a numerical example and experiments on multiphase flow processes.