模糊C均值聚类与K近邻算法的故障检测研究

基于K近邻的故障检测(FD-KNN)算法可以有效处理非线性、多模态的故障检测问题,但在过程故障检测中存在故障类型多、测量变量复杂等缺陷。将模糊C均值聚类(FCM)和K近邻(KNN)相结合,提出一种新的故障检测方法FCM-KNN。该方法与传统算法相比较,故障检测率有明显的提升。首先,应用FCM聚类将多模态训练集按模态聚类,同时根据样本与各聚类中心的距离比例来得到样本对于每个聚类中心的隶属度;再根据隶属度来判断样本所属模态,进而在各个模态下完成基于KNN的故障检测。通过多模态仿真实例进一步验证该方法的有效性。该方法具有检测率高、漏报和误报率低等优点,可有效提高检测效果。

Research on Fault Detection of Fuzzy C-Means Clustering and KNN Algorithm

The K-nearest neighbor-based fault detection(FD-KNN)algorithm can effectively deal with nonlinear and multi-modal fault detection problems,but there are many fault types and complex measurement variables in processing fault detection.Combining fuzzy C-means(FCM)clustering with K-nearest neighbor(KNN),a new fault detection method FCM-KNN is proposed.Compared with the traditional algorithm,the fault detection rate is significantly improved.Firstly,the C-means clustering is used to cluster the multi-modal training set according to the modality.At the same time,in terms of the distance ratio between the sample and each cluster center,the membership degree of each cluster center is obtained,and then the membership is judged according to the membership degree.The modality,in turn,completes KNN-based fault detection in each modality.The effectiveness of the proposed method is further verified by multi-modal simulation examples.The method has the advantages of high detection rate,low false negative rate and false positive rate,which can effectively improve the detection effect.