基于5G边缘计算的压缩机PLC数据异常检测方法

Anomaly Detection Method for Compressor PLC Data Based on 5G Edge Computing

摘要:

工业压缩机运行状态复杂，其故障易引发生产中断和安全风险。针对集中式云计算异常检测在实时性和通信效率方面的不足，对基于5G边缘计算的压缩机PLC数据异常检测方法进行了研究。构建了边缘—云协同的监测架构，在边缘节点部署轻量化深度学习模型，实现多源时序数据的实时分析。采用LSTM自编码器对压缩机正常运行模式进行建模，并结合模型压缩策略降低计算与存储开销，同时针对典型运行参数设计了相应的数据预处理与特征工程方法。基于实际工业场景采集的数据进行了实验测试，结果表明，该方法的异常检测精度达到96.3%，平均推理延迟为45ms，能够有效识别渐进性退化与突发性故障。实际应用结果表明，该方法满足工业现场对低时延、高可靠性设备状态监测的应用需求。

Abstract:

Industrial compressors operate under complex conditions, and their failures can lead to production interruptions and safety risks. To address the limitations of centralized cloud-based anomaly detection in terms of real-time performance and communication efficiency, an anomaly detection method for compressor PLC data based on 5G edge computing is investigated. An edge–cloud collaborative monitoring architecture is established, in which lightweight deep learning models are deployed on edge nodes to enable real-time analysis of multivariate time-series data. An LSTM autoencoder is employed to model normal operating patterns of compressors, combined with model compression techniques to reduce computational and storage overhead. In addition, data preprocessing and feature engineering methods tailored to typical operating parameters are designed. Experimental evaluation using real industrial data demonstrates that the proposed method achieves an anomaly detection accuracy of 96.3% with an average inference latency of 45ms, enabling effective identification of both gradual degradation and sudden faults. Practical application results indicate that the method meets the requirements of low-latency and high-reliability condition monitoring in industrial environments.