An Advanced Fault Diagnosis Model for Complex Electromechanical Systems with Enhanced Feature Extraction

Complex electromechanical systems are widely used in modern industrial production. However, the numerous subsystems and complex coupling relationships often obscure fault characteristics, thereby increasing the difficulty of fault diagnosis. Accurate diagnosis of faults is essential for reducing the probability of unplanned plant downtime and ensuring production sustainability. To address this challenge, this paper introduces a Feature-Enhanced Transformer (FET) model, which integrates Gram Angle Field Transform (GAF) and multi-scale convolution to enhance and extract features, then the global semantics are modeled and fused with the help of the Transformer. Experimental results demonstrate that the proposed model outperforms other deep learning methods in diagnostic accuracy and stability, providing an effective solution for complex electromechanical system fault diagnosis.