基于深度学习的故障诊断与预测方法综述

Summary of fault diagnosis and prediction methods based on deep learning

智能制造背景下,机械设备趋于复杂庞大,海量、多源、高维度、非结构的工业数据给系统管理监测带来更大难度,设备的故障诊断与预测更显重要。传统故障诊断与预测方法难以建立准确的数据模型,在设备故障诊断预测应用方面受到很大局限,深度学习以其强大的自主学习非线性数据表示和模式识别的能力在许多领域都有重大突破,在工业设备的故障诊断与预测领域也得到广泛关注。文中对四类经典的深度学习模型:深度置信网络、卷积神经网络、自动编码器及其变体、循环神经网络的网络结构和模型思想作详细介绍,阐述并总结了这四类深度学习模型在故障诊断与预测领域的研究成果,讨论了基于深度学习的故障诊断与预测方法的优势与不足,对未来可能的研究方向作了展望。

In the context of intelligent manufacturing,the mechanical equipments tend to be complex and huge,and the massive,multi⁃source,high⁃dimensional and non⁃structured industrial data have been bringing greater difficulty to system management and monitoring,so the equipment fault diagnosis and prediction have become more important.However,it is difficult for the traditional fault diagnosis and prediction methods to establish the accurate data models,so the traditional methods are greatly limited in equipment fault diagnosis and prediction.Deep learning has made major breakthroughs in many fields with its powerful ability of autonomically learning non⁃linear data representation and pattern recognition,and has also received widespread attention in the field of fault diagnosis and prediction of industrial equipment.Therefore,the network structures and model ideas of the four types of classic deep learning models,i.e.,deep belief network(DBN),convolutional neural network(CNN),auto⁃encoder(AE)and its variant,and recurrent neural network(RNN),are introduced in detail.The research achievements about the four types of models in the field of fault diagnosis and prediction are expounded and summarized,the advantages and disadvantages of fault diagnosis and prediction methods based on deep learning are discussed,and possible future research directions are prospected.