Predictive Maintenance Model and System Integration for Warehouse Equipment Based on Digital Twins and Deep Learning

作者：Ruizhong Kong · 发表于：2026 IEEE International Conference on Power, Electronics and Green Energy (ICPEGE) · 年份：2026 · DOI：10.1109/ICPEGE67691.2026.11451317

This paper addresses the current challenges of warehouse equipment operation, such as the difficulty in timely fault detection and the inefficiency and high cost of traditional scheduled maintenance. This paper proposes a predictive maintenance model and system integration approach based on digital twins (DT) and deep learning (DL). First, a digital twin of the warehouse equipment is constructed, mapping key operational data such as temperature, vibration, and current collected by sensors into a virtual model in real time, enabling dynamic interaction between the physical entity and the virtual space. Second, a deep learning architecture combining convolutional neural networks (CNNs) and long short-term memory (LSTM) is used to extract features and model health status from multi-source time series data. Transfer learning is then used to optimize the modeling process to address the sample shortage issue across different equipment. Finally, anomaly detection and remaining life prediction modules are integrated to achieve a closed-loop monitoring, analysis, and decision-making process at the system level. Experimental results, based on operational data from 1,000 units, show that failure predictions consistently approach or reach 72 hours in advance, with a prediction accuracy of 97%. The average maintenance cost is approximately 12.3k¥. This model not only improves the operational reliability and safety of warehouse equipment but also enables the implementation of predictive ma...