Research On LPV Modeling Method For Batch Process

With the development of market in modern society,the high demand for multi variety and multi specification products makes batch process become the more and more important mode of production.Batch process has been widely used in fine chemicals,bio pharmaceuticals,food,maquillage and other industries.An accurate and efficient model is very important for the control,optimization and monitoring of batch process.This dissertation considers the multi-stages and multi-batches of batch process,and mainly studies LPV modeling method which is based on batch process.The main results are following as:(1)Firstly we have the research on single batch of batch process.The linear variable parameters(LPV)model based on multi model method is used to deal with the batch process.First,the key variable(s)can be taken as the scheduling variable(s)and the main operating point is determined based on the actual production process.Then,the local models are constructed around each operating point,Meanwhile,to indicate whether the output measurement is missing at each sampling instant,a binary variable is defined through which the availability of the output measurement is denoted.Expectation-Maximization algorithm(EM)is introduced to estimate their parameters.Finally,the local models are combined according to an exponential weighting function.A simulated numerical example and the continuous stirred tank reactor(CSTR)are employed to demonstrate the effectiveness of the proposed method.(2)The model developed by the measurements within a single batch only can describe the dynamics of a single one rather than all the batch runs.Therefore,a LPV model based on multiple batches data is proposed to describe the batch process.First,the measurements of batch process are reduced to a plurality of two dimensional batch matrices.Then all the matrices are combined into a set of two-dimensional data.The proposed multi batch fusion LPV method divides the two-dimensional data into different stages according to the scheduling variables.In this model,the data of multiple batches are integrated into the model parameters by EM algorithm,which can avoid the influence of the initial value fluctuation of the batch process.The proposed method is evaluated through a simulated multi-batches penicillin fermentation process.(3)In the actual process,large data of batch process brings a great challenge to the multi batch fusion modeling method.Therefore the multiway-LPV(MLPV)model is proposed.The MLPV method transforms the multiway measurements to a large two-dimensional matrix along the time direction.The time slice matrices are divided into different stages according to the scheduling variables.At each stage,a sub model is built based on the “extended” input and “extended” output.Finally,all the sub models are combined into a global model.The applications of the MLPV method in several cases illustrate its advantages.