Research and Implementation of Deep Learning-based Anomaly Detection System for Industrial Manufacturing

Aiming at the problem of industrial equipment fault prediction, a deep learning-based equipment fault prediction and health management system is designed and developed. The system collects equipment operating data from multiple sources, establishes an equipment health evaluation model, and achieves fault prediction and diagnosis. The test results show that the system’s fault identification accuracy reaches $\mathbf{9 5. 8 \%}$, and the warning accuracy is $\mathbf{9 4. 3 \%}$. After six months of application in a petrochemical enterprise, the average fault interval time of the equipment is increased by 44%, and the maintenance cost is reduced by $\mathbf{3 2. 6 \%}$, achieving significant economic benefits.