深度学习算法在工业设备故障诊断应用研究  

Research on Application of Deep Learning Algorithm in Fault Diagnosis of Industrial Equipment

作　　者：范慧鹏 李瑞华[1] 李福林[1] 房哲续 彭六保 

机构地区：[1]华能沁北发电有限责任公司,河南 济源 [2]航天智控(北京)监测技术有限公司,北京

出　　处：《传感器技术与应用》2021年第4期193-200,共8页Journal of Sensor Technology and Application

摘　　要：故障诊断分类技术在工业上已经被广泛的使用,在工业设备维护起到了关键性作用,但是自动故障诊断分类技术目前还存在不足,要求精确地对设备机械进行自动诊断,准确地分析出设备故障产生的原因,从而确定故障发生的部位。针对工业上旋转机械设备的特殊性和复杂性,引入了深度学习算法来提高设备故障分类的准确率。首先对旋转机械设备建立数据集,通过深度学习算法对数据进行特征提取,由多个网络层迭代学习设备故障特征,最终优化深度学习算法模型输出不同设备故障类型,提高系统分类的准确率。本文还对故障诊断分类等技术进行总结与分析,然后重点分析了深度学习故障诊断技术在工业上机械旋转类的应用;最后提出了现有深度学习故障诊断分类技术研发方法的不足,希望深度学习领域在故障诊断技术有很好的发展。Fault diagnosis classificationtechnology has been widely used in the industry and has played a key role inthe maintenance of industrial equipment. However, the automatic fault diagnosisclassification technology currently has shortcomings. It requires accurateautomatic diagnosis of equipment and machinery, and accurate analysis ofequipment failures, the cause of the occurrence so as to determine the locationof the fault. Aiming at the particularity and complexity of industrial rotatingmachinery and equipment, deep learning algorithms are introduced to improve theaccuracy of equipment fault classification. Firstly, a data set is establishedfor rotating machinery and equipment, and features are extracted through deeplearning algorithms. Multiple network layers are used to iteratively learnequipment fault features. Finally, the deep learning algorithm model isoptimized to output different types of equipment faults to improve the accuracyof system classification. This paper also summarizes and analyzes faultdiagnosis and classification technologies, and then focuses on the applicationof deep learning fault diagnosis technology in industrial mechanical rotation.Finally, it puts forward the shortcomings of the existing deep learning faultdiagnosis classification technology research and development methods, and hopesthe deep learning field has a very good development in fault diagnosistechnology.

关 键 词：故障诊断分类 深度学习 设备预知性维护 特征提取 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]