In the domain of complex equipment health management, persistent obstacles such as low identification accuracy, inefficiency, and exorbitant maintenance costs have long hindered the ability to optimize system performance. This study delves into the potential of combining Internet of Things technology with deep learning (DL) models, particularly Convolutional Neural Network (CNN), to develop a more efficient equipment health management system. The overarching objective is to extend the operational lifespan of intricate mechanical systems while minimizing maintenance expenditures. Addressing the challenges inherent in managing complex equipment health, this study refines the traditional Multi-Layer Perceptron (MLP) by incorporating Adagrad optimization parameters. This integration effectively mitigates the shortcomings of conventional MLP models. By capitalizing on the unsupervised feature extraction capabilities of Deep Belief Network (DBN) and the supervised classification proficiency of MLP, a robust fault diagnosis model grounded in DL is constructed to forecast the operational state of complex machinery with high precision. Rigorous performance evaluations conducted on the proposed system reveal that fault diagnosis accuracy for the advanced CNN, standard CNN, and Stacked Auto-Encoder (SAE) models all surpass 90%. Among these, the advanced CNN model achieves an average accuracy of 95.35%, outstripping both the standard CNN and SAE models by at least 4%. These results provide significant insights into the application of DL technology within complex equipment health management, presenting viable solutions to longstanding challenges in the field. The findings hold the potential for making a substantial and lasting impact on both academic research and engineering practice in related domains.
Zhao et al. (Sat,) studied this question.