摘要
针对传统转子故障诊断方法存在精度低、适应性差、难以满足识别复杂多变故障类型的需求问题,提出一种基于卷积神经网络的故障检测方法,实现对转子系统故障的诊断与研究。通过HZXT-009滑动轴承故障模拟综合试验台进行转子不对中、不平衡、碰摩等故障试验,获取数据并分析时域振动信号;搭建卷积神经网络故障诊断模型;将该方法与深度信念网络(DBN)算法进行对比。该方法准确率能够达到99.16%,高于深度置信网络(DBN)算法方法,该方法有望在实际生产中得到广泛应用,提高转子系统故障诊断效率和可靠性。
Aiming at the traditional rotor fault diagnosis method,which has low precision,poor adaptability,and is difficult to meet the demand for identifying complex and variable fault types,a fault detection method based on convolutional neural network is proposed to realize the diagnosis and research of rotor system faults.Through the HZXT-009 sliding bearing fault simulation comprehensive test rig to carry out rotor misalignment,unbalance,touching and other fault tests,to obtain data and analyze the time-domain vibration signals;to build a convolutional neural network fault diagnosis model;and to compare the method with the deep belief network(DBN)algorithm.The accuracy of this method can reach 99.16%,which is higher than that of the Deep Belief Network(DBN)algorithm,and this method is expected to be widely used in actual production to improve the efficiency and reliability of rotor system fault diagnosis.
作者
潘宏刚
李员禄
郭宝仁
刘昱伯
PAN Hong-gang;LI Yuan-lu;GUO Bao-ren;LIU Yi-bo(Shenyang School of Energy and Power,Shenyang Institute of Engineering,Shenyang 110136,China;Northeast Branch of Huadian Electric Power Research Institute,Shenyang 110180,China;Guizhou Power Grid Co.,Ltd.,Anshun Power Supply Bureau,Anshun 561000,China)
出处
《汽轮机技术》
北大核心
2024年第2期145-148,共4页
Turbine Technology
基金
中国博士后科学基金(2019M661125)。
