摘要
将并行偏最小二乘(CPLS)算法引入到间歇过程监控中,提出一种多向并行偏最小二乘(MCPLS)监控方法。CPLS算法可以提取过程变量与质量变量的相关信息,也能对彼此不相关的信息进行主元提取。与基于PLS的监控方法不同,基于CPIS的过程监控方法提供了一个完整的监控框架,不仅能够监控过程变量,而且也能监控质量变量的信息,更好地反映了过程的运行状态。文中首先将间歇过程三维数据转换为二维数据,然后应用CPLS算法建立过程监控模型,构建T_c^2,T_x^2,Q_z,T_y^2,Q_y监控指标,并通过间歇过程批次间的统计特性计算出监控指标控制限,分别监控过程变量与质量变量的相关信息、彼此无关信息以及残差信号等。最后将MCPLS算法应用到青霉素发酵过程的监控中,应用结果表明了该方法在间歇过程监控中的有效性和优越性。
This paper introduces the concurrent projection to latent structures (CPLS) algorithm into the batch process monitoring to propose a novel batch process monitoring approach based on the multi-way concurrent projection to latent structures (MCPLS).The CPLS algorithm extracts both the relevant and irrelevant information between process variables and quality variables.The monitoring method based on CPLS supports a completed monitoring framework which can monitor not only the process variables' information but also the quality variables' information to reflect the running state of the process compared with the monitoring method based on PLS.This paper turns a three-dimensional batch process data into a two-dimension,then T2c,T2x,Qx,T2y,Qy are established to monitor the relevant information of process variables and quality variables and the irrelevant information and the residual signal based on the CPLS model and calculate the monitoring indicators control limits through the statistical property of the batch process.Finally,the MCPLS algorithm is used to a penicillin fermentation process and shows the effectiveness and superiority of the proposed method for the batch process monitoring.
出处
《江南大学学报(自然科学版)》
CAS
2014年第3期282-286,共5页
Joural of Jiangnan University (Natural Science Edition)
基金
江苏省自然科学基金项目(BK2010149)