基于EasyEnsemble的化工过程故障诊断性能改进

发布时间：2018-11-17 16:47

【摘要】：化工过程故障诊断中样本数据分布不均衡现象普遍存在.在使用不均衡样本作为训练集建立各类故障诊断分类器时,易出现分类器的识别率偏置于多数类样本的结果,由此产生虽正常状态易识别,但更受关注的故障状态却难以被诊断的现象.针对该问题,本文提出一种基于Easy Ensemble思想的主元分析支持向量机(Easy Ensemble based principle component analysis support vector machine,EEPS)故障诊断算法,通过欠采样方法抽取多数类样本子集组建多个新的均衡数据样本集,使用主元分析(principle component analysis,PCA)进行特征提取并使用支持向量机(support vector machine,SVM)算法进行训练,得到多个基于SVM的故障诊断分类器,然后使用Adaboost算法集成最终的分类,从而提高故障诊断准确性.所提方法被用于TE(Tenessee Eastman)化工过程,实验结果表明,EEPS算法能够有效提高分类器在不均衡数据集上的诊断性能和预报能力.
[Abstract]:The uneven distribution of sample data is common in fault diagnosis of chemical process. When using unbalanced samples as training sets to establish various fault diagnosis classifiers, the recognition rate of the classifiers is prone to be biased to the results of most class samples, which results in the fact that the normal state of the classifier is easy to recognize. But the more concerned fault state is difficult to diagnose the phenomenon. In order to solve this problem, a principal component analysis (Easy Ensemble based principle component analysis) support vector machine,EEPS fault diagnosis algorithm based on Easy Ensemble is proposed in this paper. Several new balanced data samples are constructed by extracting subsets of most class samples by under-sampling method, feature extraction is performed by principal component analysis (principle component analysis,PCA) and training is carried out using support vector machine (support vector machine,SVM) algorithm. Several fault diagnosis classifiers based on SVM are obtained, and the final classification is integrated with Adaboost algorithm to improve the accuracy of fault diagnosis. The proposed method is used in TE (Tenessee Eastman) chemical process. Experimental results show that the EEPS algorithm can effectively improve the diagnostic performance and prediction ability of the classifier on unbalanced data sets.
【作者单位】：上海理工大学管理学院;华中科技大学自动化学院;
【基金】：上海高校青年教师培养资助计划(10 15 303 808)资助~~
【分类号】：TQ050.7

【相似文献】