论文部分内容阅读
针对间歇过程存在的非线性、噪声强、多工况等特点,提出一种基于改进多向核熵成分分析 (Improved Multi-way Kernel Entropy Component Analysis, IMKECA) 的非线性故障检测方法.该方法引入小波变换对过程数据进行去噪处理,抑制噪声对统计建模造成的影响,并对降噪后数据进行非线性统计分析,建立基于Renyi熵的MKECA模型.为更好的适应多工况数据的变化,进一步构建基于 k 近邻马氏距离的 M 监控统计量,代替传统 T2监控统计量,以提高故障检测效果.最后所提方法以盘尼西林发酵过程为监控对象进行仿真研究,结果验证了IMKECA方法能够比传统MKECA方法更为有效的监控过程故障.“,”Aiming at the nonlinearity, strong noise and multi-modes characteristics of batch process, an improved multi-way kernel entropy component analysis (IMKECA) is proposed. This method uses wavelet transform to denoise the process data in order to prevent the influence of noise during the statistical modeling, then the denoised data is used to build the MKECA model according to the Renyi entropy. To better monitor the multi-mode data, this method further applies the kNN based Mahalanobis distance to build the M monitoring statistic, which replaces the traditional T2monitoring statistic to improve the effects of monitoring. The proposed method is applied to the Penicillin Fermentation process, and the simulation results show that the IMKECA method can monitor faults more effectively than the traditional MKECA.