基于粒子群优化KFCM的风电齿轮箱故障诊断

发布时间：2018-02-02 05:02

本文关键词： 模糊核聚类粒子群优化算法风电机组齿轮箱故障诊断　出处：《振动.测试与诊断》2017年03期 　论文类型：期刊论文

【摘要】：针对基于有监督学习的方法无法识别未知类别故障,提出了一种基于粒子群优化模糊核聚类(kernel fuzzy c-means clustering,简称KFCM)的风电机组齿轮箱故障诊断方法。首先,建立以训练样本分类错误率为目标的聚类模型,利用KFCM对训练样本进行分类;然后,以初始聚类中心和核函数参数作为优化变量,利用粒子群优化算法求解聚类模型,获得最优分类结果下每个类的类心;最后,根据新样本与各类心之间的核空间样本相似度判断新样本属于已知故障或者未知故障。以某风电机组齿轮箱为例,对提出方法的有效性进行试验验证。结果表明,与传统基于有监督学习的神经网络方法相比,该方法能有效诊断已知和未知类别的故障。
[Abstract]:The method based on supervised learning can not identify unknown class faults. A fuzzy kernel clustering method based on particle swarm optimization (PSO) is proposed for kernel fuzzy c-means clustering. Firstly, a clustering model aiming at the classification error rate of training samples is established, and KFCM is used to classify the training samples. Then, using the initial clustering center and kernel function parameters as the optimization variables, the particle swarm optimization algorithm is used to solve the clustering model, and the cluster center of each class is obtained under the optimal classification results. Finally, according to the similarity between the new samples and the core space samples, the new samples are identified as known or unknown faults. A wind turbine gearbox is taken as an example. Experimental results show that the proposed method can effectively diagnose known and unknown types of faults compared with the traditional neural network method based on supervised learning.
【作者单位】：华北电力大学能源动力与机械工程学院;北京华能新锐控制技术有限公司;
【基金】：国家自然科学基金资助项目(51305135) 中国华能集团科技资助项目(HNKJ13-H20-05) 中央高校基本科研业务费专项资金资助项目(2014XS15)
【分类号】：TH132.41;TM614
【正文快照】： 引言齿轮箱是风电机组传动链的关键设备,由于风电机组运行环境恶劣,导致齿轮箱故障频繁发生,不仅事后维修工作难度大,而且费用高。因此,对风电机组齿轮箱进行有效的监测和故障诊断,保障风电机组齿轮箱安全可靠运行具有重要意义。故障诊断的本质是对故障信息的模式识别与分类[1

【相似文献】