基于CLPSO-IDBN的风电机组轴承故障诊断  

CLPSO-IDBN Based Bearing Fault Dagnosis Method for Wind Turbine

作　　者：谢东东 沈艳霞[1] XIE Dongdong;SHEN Yanxia(School of Internet of Things Engineering,Jiangnan University,Wuxi 210000,China)

机构地区：[1]江南大学物联网工程学院,无锡214122

出　　处：《组合机床与自动化加工技术》2023年第10期109-113,119,共6页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家重点研发计划项目(2020YFB1711102)。

摘　　要：针对风电机组轴承故障诊断时的数据特征复杂难以提取,故障诊断准确率低,耗费时间长等问题,提出一种综合型学习粒子群算法(comprehensive learning particle swarm optimization,CLPSO)与改进深度置信网络(improved deep belief network,IDBN)相结合的故障诊断方法。首先在DBN内部添加了迭代误差阈值优化策略构建IDBN,大大减少了训练时间;然后利用CLPSO算法优选IDBN网络结构,运用具有最优结构的IDBN模型从原始信号中提取故障特征,识别轴承的故障类型。仿真实验结果表明,CLPSO-IDBN算法模型具有更高的准确率以及在相同情况时更少的训练时间,在训练集和测试集上的诊断准确率分别达到了98.28%与97%,并且可以平均节省约30%的训练时间,与4种其他方法相比较,证实了新方法的有效性与准确性。To solve the problems of complex data features,low accuracy and long time consumption in fault diagnosis of wind turbine bearing,a fault diagnosis method combining comprehensive learning particle swarm optimization(CLPSO)and improved deep trust network(IDBN)is proposed.Firstly,the iterative error threshold optimization strategy is added to the DBN to construct the IDBN,which greatly reduces the training time.Then the CLPSO algorithm is used to optimize the IDBN network structure,and the IDBN model with the optimal structure is used to extract fault features from the original signal to identify the fault type of the bearing.The simulation results show that the CLPSO-IDBN algorithm model has higher accuracy and less training time under the same conditions.The diagnostic accuracy rate on the training set and the test set has reached 98.28%and 97%respectively,and the average training time can be saved by about 30%.Compared with four other methods,the effectiveness and accuracy of this method is confirmed.

关 键 词：深度置信网络(DBN) 综合学习粒子群算法(CLPSO) 滚动轴承 故障诊断 风电机组 

分 类 号：TH133.3[机械工程—机械制造及自动化] TG65[金属学及工艺—金属切削加工及机床]