基于MLFDA的化工过程故障模式分类方法

发布时间：2018-10-07 19:50

【摘要】：Fisher判别分析(FDA)是一种有效的化工过程故障模式分类方法,但是其忽视了数据局部结构信息的挖掘。针对该问题,提出一种多块局部Fisher判别分析(MLFDA)方法,以更有效地识别化工过程故障。从变量和样本两个维度来分析数据的局部结构特性。针对变量维度的局部信息挖掘问题,设计了一种基于变量与数据集主元空间的相关度的变量分块方法,将全局过程变量划分为多个局部变量块。进一步考虑到样本维度的局部结构特性,应用基于局部权重因子的局部Fisher判别分析(LFDA)为每个局部变量块构建分类器。提出一种基于分类性能加权的多分类器集成方法,以融合不同分类器的决策结果。在Tennessee Eastman过程上的仿真结果说明,M LFDA方法具有比传统的FDA和LFDA方法更低的故障误分类率。
[Abstract]:Fisher discriminant analysis (FDA) is an effective method for classification of fault patterns in chemical processes, but it ignores the mining of local structure information. To solve this problem, a multi-block local Fisher discriminant analysis (MLFDA) method is proposed to identify chemical process faults more effectively. The local structure of data is analyzed from two dimensions: variable and sample. To solve the problem of local information mining of variable dimension, a variable block method based on the correlation between variables and the principal component space of the dataset is designed. The global process variables are divided into several local variable blocks. Considering the local structure of the sample dimension, the local Fisher discriminant analysis (LFDA) based on the local weight factor is used to construct the classifier for each local variable block. A multi-classifier ensemble method based on classification performance weighting is proposed to fuse the decision results of different classifiers. The simulation results on the Tennessee Eastman process show that the M LFDA method has a lower fault error classification rate than the traditional FDA and LFDA methods.
【作者单位】：中国石油大学(华东)信息与控制工程学院;
【基金】：国家自然科学基金资助项目(61403418,61273160,21606256) 山东省自然科学基金资助项目(ZR2014FL016,ZR2016FQ21,ZR2016BQ14) 中央高校基本科研业务费专项资金资助项目(17CX02054) 研究生创新工程资助项目(YCX2017058)
【分类号】：TQ050.7

【相似文献】