prof CLAM-Net: A Fault Diagnosis Model for Mechanical Systems under Complex Operating Conditions

Abstract

With the rapid development of intelligent manufacturing and industrial automation, the operating environments of mechanical systems have become increasingly complex, imposing higher demands on the accuracy and generalization capability of fault diagnosis. Existing fault diagnosis methods typically focus on local features or single-sequence modeling, lacking mechanisms to simultaneously extract multi-scale spatial features and capture long-term temporal dependencies, which limits their ability to effectively represent complex dynamic fault characteristics in mechanical vibration signals. To address this issue, this paper proposes a novel architecture—CLAM-Net, which integrates convolutional neural networks (CNN), long short-term memory networks (LSTM), attention mechanisms, and meta-learning. The model leverages CNNs to extract multi-scale time–frequency features, employs LSTMs to capture temporal dependencies, and incorporates an attention mechanism to adaptively focus on critical fault-related features. Furthermore, a meta-learning framework is adopted to enable rapid adaptation and knowledge transfer across varying operating conditions and equipment states. Extensive experiments on real-world mechanical datasets demonstrate that CLAM-Net achieves an accuracy of 98.5%, a recall of 98.2%, and an F1 score of 98.3%, outperforming the state-of-the-art GRU–Attention model by approximately 3.5%, 4.0%, and 3.7%, respectively. These results confirm that CLAM-Net exhibits superior diagnostic performance and robustness under complex operating conditions, providing an efficient and scalable solution for intelligent mechanical fault diagnosis.