基于极限学习机的分类算法及在故障识别中的应用

Classification algorithm based on extreme learning machine and its application in fault identification of Tennessee Eastman process

利用极限学习机(ELM)分类器的结构特点重新设计面向多分类任务的ELM分类器,提出基于ELM的优化分类算法One-Class-PCA-ELM.该算法的实现过程如下:对故障数据进行主元分析(PCA)处理,降低数据维数,去除噪声与冗余信息;将训练数据集按类分割,建立各类对应的单分类模型,整合得到One-Class-PCA-ELM分类模型;将待分类数据输入One-Class-PCA-ELM分类模型,得到待分类数据的类标号,完成分类.仿真实验结果表明,该算法保持了极限学习机极快的训练速度,具有较高的分类准确率及较理想的分类稳定性.

A new extreme learning machine(ELM)classifier for multi-classification task was designed based on the structural features of the ELM classifier,and an improved classification algorithm based on ELM(One-Class-PCA-ELM)was purposed.The classification algorithm was realized as follows.PCA method was utilized to process the fault data for dimensionality reduction as well as removing noise and redundant information.Then the training data were allocated into different categories according to their respective class labels and the corresponding classification model was constructed for each training data category,obtaining One-Class-PCA-ELM model.An unclassified fault data was constructed into the trained One-Class-PCA-ELM model,getting its class label and making classification process completed.Experimental results show that the proposed algorithm maintains the fast training speed of ELM,and has high classification accuracy and ideal classification stability.