基于混合MPLS的多阶段过程质量预报方法

发布时间：2018-01-23 22:43

  本文关键词： 多阶段特性 多向偏最小二乘方法 费舍尔判据分析 贝叶斯原则 多阶段模型融合　出处：《山东大学学报(工学版)》2017年05期 　论文类型：期刊论文

【摘要】：针对传统的多向偏最小二乘方法(multi-way partial least squares,MPLS)在质量预报中存在着模型预测精度低、局部预报能力不足等问题,提出一种多MPLS模型融合方法来提高预报表现。利用高斯混合模型(Gauss mixture model,GM M)对每批次过程和质量数据组成的高维空间进行阶段识别。针对多批次同一子阶段长度不等问题,采用动态时间规整(dynamic time warping,DTW)算法依据最长持续时间同步为等长轨迹,并在子阶段中按变量展开方式建立MPLS模型。根据Fisher判据分析(Fisher discriminate analysis,FDA)最小化子阶段数据集间相关性,利用核密度方法估计子阶段数据集去相关后的概率密度分布来在线监测阶段切换。利用贝叶斯原则融合各子阶段MPLS模型进行质量预报。将该方法应用到工业青霉素发酵过程中,表明了所提方法具有更好的监控性能和预报能力。
[Abstract]:In view of the traditional multi-way partial least squares method, there is a low precision of model prediction in mass prediction. In order to improve the prediction performance, a multi-#en0# model fusion method is proposed to improve the prediction performance. The Gao Si mixed model is used to improve the performance of Gauss mixture model. GM M) is used to identify the high dimensional space of each batch process and quality data, aiming at the different length of the same sub-stage of multiple batches. The dynamic time warping time (DTW) algorithm is used to synchronize the longest duration to equal length trajectory. The MPLS model is built according to the variable expansion in the sub-stage, and the Fisher criterion is used to analyze the discriminate analysis. FDAs minimize the correlation between substage datasets. The kernel density method is used to estimate the dense-correlated probability density distribution of the sub-stage data set to monitor the phase switching online. The Bayesian principle is used to fuse the sub-stage MPLS models for quality prediction. The method is applied to work. In the process of penicillin fermentation. The results show that the proposed method has better monitoring performance and prediction ability.
【作者单位】： 杭州电子科技大学新型电子器件与应用研究所;湖州师范学院信息与控制技术研究所;
【基金】：国家自然科学基金资助项目(61573137)
【分类号】：TB114.2
【正文快照】： 融合的多阶段质量预报方法。首先,利用GMM模0引言型对每批次采集数据进行阶段识别。针对多批次同一子阶段长度不等问题,应用动态时间归整(DTW)在工业过程中,由于工艺和检测技术的限制,产算法[14-15]依据相似度最小和最长反应持续时间同品的质量指标难以在线直接测量,需要离线，

本文编号：1458429