Abstract
Aviation engines, as vital aircraft components, encounter challenges in Condition Monitoring (CM) signal fault diagnosis, including low accuracy and poor real-time performance. To tackle these, by integrating an Auto-Encoder (AE) and Bidirectional Gated Recurrent Unit (BiGRU), this study proposes an AE-BiGRU-based fault signal feature extraction model for aviation engines. Then, an ISMA-SVM-based aviation engine CM signal fault diagnosis method is introduced, employing the Improved Slime Mould Algorithm (ISMA) to optimize Support Vector Machine (SVM) parameters. The findings reveal that in function optimization, the ISMA exhibits stronger variance reduction in both unimodal and multimodal functions, particularly in the latter. In terms of fault diagnosis, the model performs excellently in precision (0.90), recall (0.95), and F1 score (0.92), achieving the best performance in the C-MAPSS dataset prediction. Case applications show their advantages in extracting weak fault signals and identifying fault frequencies, enhancing aviation engine fault diagnosis, safeguarding health status, and aiding damage repair.