To enhance the effectiveness of predictive maintenance for industrial production equipment, this work explores a cloud platform-based predictive maintenance system. Moreover, it designs an equipment fault diagnosis model using the One Dimensional Deep Residual Shrinkage Network (1DDRSN). The performance of the 1DDRSN-based equipment fault diagnosis model and the proposed predictive maintenance system is validated through bearing fault detection experiments. The results demonstrate that the 1DDRSN model significantly outperforms other models in equipment fault diagnosis, achieving an accuracy, precision, recall, and F1 score of 0.9886, 0.9796, 0.9684, and 0.974, respectively. Compared to other models, these metrics represent improvements of at least 0.66%, 0.56%, 0.76%, and 0.62%, respectively. This indicates that the 1DDRSN model offers higher robustness and better predictive performance for complex industrial equipment fault diagnosis tasks. Additionally, performance testing of the cloud platform-based predictive maintenance system demonstrates superior response time, system throughput, and data processing efficiency compared to traditional systems. This suggests the proposed system’s ability to better support real-time maintenance needs in complex industrial environments. The findings of this work provide technical support for intelligent maintenance in industrial production and lay the foundation for future developments in the field of smart manufacturing.