论文部分内容阅读
针对间歇过程不同批次原材料属性不同,不同反应阶段过程变量设定值不同的问题,提出基于核矩阵的多方向多阶段全潜结构投影法(Kernel based Multi-way Multi-stage Total Projection to Latent Structure,KMMT-PLS)用于其质量相关故障的在线监测。该方法首先将间歇过程的三维立体数据集平面化处理;然后将原始空间中呈强非线性关系的历史数据映射至高维特征空间,实现非线性问题的线性化;最后在特征空间建立核矩阵与输出的线性T-PLS模型,实现间歇过程质量相关故障的在线监测。基于Pensim V 2.0的青霉素发酵过程仿真实验表明所提方法的有效性。
Aiming at the problem that the raw materials have different properties in different batches and the process variables in different reaction stages have different set values, a Kernel based Multi-way Multi-way Total Projection to Latent Structure , KMMT-PLS) for online monitoring of its quality related failures. The method first planarizes the three-dimensional data set of the batch process, then maps the historical data with strong non-linear relationship in the original space to the high-dimensional feature space to linearize the nonlinear problem, and finally establishes the kernel matrix and Output of the linear T-PLS model to achieve intermittent quality-related online monitoring process failures. Pensim V 2.0 based penicillin fermentation process simulation experiments show the effectiveness of the proposed method.