摘要
针对间歇过程的非线性和动态性,提出了全局—局部正则化高斯混合模型(GLRGMM)算法。首先引入邻域保持嵌入算法提取局部流形结构,通过寻求一种低维投影对非线性过程进行全局结构保持,同时最大限度地保留局部流形特征;然后通过对高斯混合模型引入正则项来在线监控更新高斯模型,获取非线性数据流形结构,解决数据动态性问题;最后集成全局—局部监控指标实现在线监控。通过青霉素发酵过程进行了验证,结果表明所提算法比DPCA、GLNPE具有更好的在线监控效果。
Aiming at nonlinear and dynamic characteristics of batch process,this paper proposed a global-local regularization Gaussian mixture model(GLRGMM)algorithm.At first,this algorithm introduced neighborhood preserving embedding algorithm to extract the local manifold structure.It preserved global structure of nonlinear process by seeking a low-dimensional projection while preserved local manifold features to the maximum extent at the same time.Then it introduced the regularization term into GMM to online monitor and updated Gaussian model.It obtained the manifold structure of nonlinear data and solved the data dynamic characteristic problem simultaneously.Finally,it used the integration of global-local monitoring indicators to effectively achieve online monitoring.The results verified by the penicillin fermentation process show that the proposed algorithm has better online monitoring effect than DPCA and GLNPE algorithms.
作者
赵小强
周文伟
Zhao Xiaoqiang;Zhou Wenwei(College of Electrical&Information Engineering,Lanzhou University of Technology,Lanzhou 730050,China;National Experimental Teaching Center of Electrical&Control Engineering,Lanzhou University of Technology,Lanzhou 730050,China;Key Laboratory of Gansu Advanced Control for Industrial Processes,Lanzhou 730050,China)
出处
《计算机应用研究》
CSCD
北大核心
2020年第1期127-130,152,共5页
Application Research of Computers
基金
国家自然科学基金资助项目(61763029).
