注塑过程子时段动态非线性质量预测

针对注塑间歇过程多阶段、缓慢时变、非线性和质量变量测量值不能在线获得等特点，提出子时段滑动窗口广义回归神经网络质量预测方法，首先，采用分类算法对三维数据矩阵的时间片PCA负载矩阵进行分析，根据相关性分析把注塑过程划分为几个子时段，然后确定与重量密切相关的阶段，在确定的阶段内采用滑动窗口建立GRNN多模型，解决常规MPLS在工业应用过程中存在的几个潜在问题：（1）静态单一模型；（2）模型失配问题；（3）MPLS线性方法不能充分有效压缩和抽取非线性过程信息；（4）估计未来测量变量所引进的模型偏差。所提方法与子时段滑动窗口MPLS方法进行仿真比较。结果证明了所提方法的有效性。

Sub-stage Dynamic and Nonlinear Quality Prediction for Injection Molding Processes

For multistage,time-variant,nonlinear characteristic and the unavailable on-line product qualities of injection molding batch process,a sub-stage moving window generalized regression neural network(GRNN) method was proposed.Using an clustering arithmetic,PCA P-loading matrices of time-slice matrices was clustered according to relevance,and injection molding process was divided into several operation stages,the most relevant stage to the quality variable was defined,and applying moving windows to un-fold stage data according to time,sub-stage GRNN models were developed for every windows for on-line quality prediction.The proposed method easily handles the following problems:(1) static single model;(2) process and its model do not match;(3) linear method may not be efficient in compressing and extracting nonlinear process data;(4) errors are added by estimating the future trajectory of the ongoing batch.For comparison purposes,a sub-MPLS quality model of every moving window was established.The results demonstrate the effectiveness of the proposed method.