Abstract: Aiming at the problem of bad fault detection effect of batch process because of its non-linearity and mixed distribution of Gaussian and non-Gaussian,Multi-way Differencial Neighborhood Preserving Embedding-Weighted and Differencial Independent Component Analysis (MDNPE-WDICA) algorithm for fault detection of batch process was proposed.The original data space was divided into Gaussian and non-Gaussian subspaces by Jarque-Bera testmethod (J-B test).In Gaussian subspace,MDNPE algorithm was proposed by combining differential strategy with NPE algorithm to preserve the local structure invariant and deal with the nonlinearity of data while the dimension of data was reduced,which could overcome the computational complexity caused by the introduction of the kernel function.In non-Gaussian subspace,WDICA algorithm was proposed by combining weighted differential strategy with ICA algorithm to solve the nonlinearity of data while the non-Gaussian information of data was fully extracted,and the local information of data was effectively used.A new monitoring statistic was established by Bayesian inference to realize fault detection for the whole batch process.The simulation results of penicillin production process demonstrated that the proposed algorithm was feasible and effective.

Key words: batch process, fault detection, nonlinear, Gaussian and non-Gaussian, differential strategy

摘要： 针对间歇过程数据的非线性和高斯与非高斯混合分布特性导致故障检测效果不佳的问题,提出了基于多向差分邻域保持嵌入—加权差分独立元分析(MDNPE-WDICA)的间歇过程故障检测算法。首先采用Jarque-Bera检验方法(J-B test)将原始数据空间划分为高斯和非高斯子空间;然后,在高斯子空间,将差分策略与NPE算法结合提出MDNPE算法,对高斯空间数据进行维数约简,在保持其局部结构不变的同时处理非线性,并克服传统非线性处理方法由于引入核函数带来的计算复杂的问题;在非高斯子空间,将加权差分策略与ICA算法结合提出WDICA算法,在充分提取数据非高斯信息的同时解决其非线性,并有效利用数据的局部信息;最后,通过贝叶斯推断构建一个新的监测统计量,实现整个间歇过程数据的故障检测。通过青霉素生产过程仿真结果验证了所提算法的可行性和有效性。

关键词: 间歇过程, 故障检测, 非线性, 高斯与非高斯混合分布, 差分策略