基于独立成分和支持向量数据描述(IC-SVDD)的工业过程故障检测  

Fault monitoring of industrial process based on independent component and support vector data description(IC-SVDD)

作　　者：衷路生[1] 侯昌瑞 

机构地区：[1]华东交通大学,江西南昌330013

出　　处：《计算机与应用化学》2017年第4期285-290,共6页Computers and Applied Chemistry

基　　金：国家自然科学基金资助项目(61263010;60904049);江西省自然科学基金资助项目(20114BAB211014)

摘　　要：针对复杂工业过程数据分布复杂的情况,提出了一种基于IC-SVDD(independent component-support vector data description)的工业过程故障监测方法。由于实际工业过程数据存在非线性和非高斯性问题,为了解决这两个同时存在的问题,采用IC-SVDD算法对数据进行处理。首先,利用独立成分分析算法对工业过程数据进行ICA分解,寻找一个分离矩阵W,实现分离原始数据,通过W的线性变化,可以将独立主元从混合信号中分离出来。然后,把提取出来的数据利用SVDD算法进行数据重构,进而构建新的统计量和统计限。最后,对Tennessee Eastman(TE)过程进行仿真,实验结果验证了该方法的可行性和有效性。Aiming at the complex data distribution of complex industrial process, an industrial process fault monitoring method based on IC-SVDD （independent component-support vector data description independent） is proposed. In view of the practical industrial process data has nonlinear and non-Gauss problems, in order to solve these two problems, the IC-SVDD algorithm is used to process the data. Firstly, using independent component analysis algorithm to the ICA decomposition of industrial process data, looking for a separation matrix W, achieved the separation of raw data, through the linear change of W, independent principal component can be separated from the mixed signal. Then, The extracted data utilizing SVDD algorithm to reconstruct the data, and then new statistics and statistical limits are established. Finally, the feasibility and effectiveness of the proposed fault detection algorithm are demonstrated by the simulation of Eastman Tennessee （TE） process, and the detection results are improved.

关 键 词：ICA 支持向量数据描述 故障检测 非高斯 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]