Quality Prediction of Multiple Time-varying Batch Production Process Based on MW-MKEPLS
周文伟 1孙步功 1石林榕1
扫码查看
点击上方二维码区域，可以放大扫码查看
作者信息
1. 甘肃农业大学机电工程学院,兰州 730070
折叠
摘要
间歇生产过程的多重时变特性和非线性使得质量预测问题变得复杂.为了提高间歇过程质量预测精度,提出了滑动窗多向核熵偏最小二乘(moving window multiway kernel entropy partial least squares,MW-MKEPLS)方法.首先采用滑动窗进行数据的动态更新获取,构建了滑动窗多重时变模型;然后在滑动窗多重时变模型下通过核函数将数据映射到高维特征空间,采用Renyi熵贡献度进行数据特征提取,更好地获取数据的信息熵和非线性;最后在KECA处理后的高维特征空间进行质量预测.通过青霉素生产发酵过程进行了实验验证,并与MKPLS和MKEPLS进行对比分析,结果表明所提方法的质量预测精度更高.
Abstract
The multiple time-varying characteristic and nonlinearity of batch production process complicate the problem of quality prediction.In order to improve the accuracy of batch process quality prediction,a moving window multiway kernel entropy partial least squares(MW-MKEPLS)method was proposed.Firstly,the moving window was used to dy-namically update and obtain the data,and the multiple time-varying model of the moving window was constructed.Then,under the moving window of multiple time-varying model,the data was mapped to the high-dimensional feature space through the kernel function,and the Renyi entropy contribution was used to extract the data features to better obtain the information entropy and nonlinearity of the data.Finally,the quality prediction was carried out in the high-dimensional feature space processed by KECA.The Experimental verification was carried out through the fer-mentation process of penicillin production,and comparative analysis was conducted with MKPLS and MKEPLS.The results showed that the quality prediction accuracy of the proposed method was higher.
关键词
间歇过程/多重时变特性/核熵成分分析/偏最小二乘/质量预测
Key words
batch process/multiple time-varying characteristic/kernel entropy component analysis/partial least squares(PLS)/quality prediction
维普
万方数据