一种基于局部模型的非线性多工况过程监测方法

针对复杂工业过程中的非线性、非高斯特性以及多工况问题, 提出了一种基于局部模型的在线统计监测新方法. 首先利用局部最小二乘支持向量机回归 (Least square support vector regression, LSSVR) 模型对过程输出进行预测, 与真实的输出相比较构成残差序列. 然后利用 ICA-PCA 两步特征提取策略, 完整地提取残差的高斯和非高斯信息, 最后用三个统计量 (I2、T2 和 SPE) 对过程进行监测, 建立了一种具有非线性、非高斯特性的多工况过程在线监测算法. 通过对 TE (Tennessee Eastman) 过程的仿真研究, 验证提出的方法是可行、有效的, 并显示出了一定的故障检测能力.

Local Model Based Monitoring for Nonlinear Multiple Mode Process

In order to handle the multiple operating mode problem and the widely existing non-Gaussian information of nonlinear processes,a new online local model based process monitoring method is proposed.First,least square support vector regression (LSSVR) local model is used to predict process outputs.Residuals are made between real outputs and predicted outputs.Then a two step ICA-PCA information extraction is carried out in order to extract both Gaussian and non-Gaussian information from the residuals.Finally,three statistic quantities are built to monitor the process.An online monitoring algorithm for nonlinear multiple mode processes with non-Gaussian information is also developed.Simulation of TE process shows the feasibility and efficiency of the new method.Besides,the new method also shows fault detection ability.