Chemical Process Quality Prediction Based On Improved Partial Least Square Method

With the improvement of people's living standards,the requirements for product quality are getting higher and higher.In order to improve product quality,the chemical production process introduces more complicated control systems.On the other hand,with the development of sensors and measurement technology,more and more detailed data information about the production process can be obtained.however,there are still some important quality variables that cannot be obtained by direct measurement.Therefore,how to find useful information from massive data to serve product quality control has become an urgent problem to be solved.In this context,data-driven quality prediction methods have attracted widespread attention.The traditional quality prediction method is mainly based on partial least squares(PLS),which can extract the main information of process data,establish the relationship model between process input data and quality data,and then use the model to predict product quality.As the complexity of the chemical production process increases,the relationship between process variables becomes more and more complex,and the nonlinear relationship becomes stronger.The traditional PLS method cannot establish an accurate prediction model for such a system.To solve this problem,the research has improved the traditional PLS method to meet the needs of the actual production process.The main research contents are as follows:1.Aiming at the fact that the traditional PLS method preprocesses the data rough and can't deal with the nonlinear process,the OSC-SVM-PLS quality prediction method is proposed,which uses the orthogonal signal correction(OSC)algorithm instead of the traditional method.The preprocessing method processes the process data and then introduces a support vector machine(SVM)algorithm into the PLS to have the ability to deal with nonlinear data.The proposed OSC-SVM-PLS algorithm is then compared with the PLS algorithm and the kernel partial least squares(KPLS)algorithm,which are used in a numerical simulation example and the Tennessee-Eastman(TE)process.The root mean square error(RMSE)value quantifies the prediction performance of different methods.By analyzing the prediction results and comparing the RMSE values,it is proved that the proposed method is better than the PLS algorithm and the KPLS algorithm.2.For the noise contained in the chemical process data set,the data is processed by wavelet denoising(WD)method,and the WDOSC-SVM-PLS quality prediction method is proposed.The method first uses wavelet denoising to remove the noise contained in the data set,and then uses the orthogonal signal correction(OSC)algorithm to remove the part of the input data that is not related to the quality data.In order to enable the model to process the nonlinear process,a support vector machine(SVM)algorithm is introduced.The proposed WDOSC-SVM-PLS algorithm is compared with the PLS algorithm and the KPLS algorithm,and they are used in a numerical simulation example and penicillin fermentation process.By comparing the prediction results and the RMSE values,it is proved that under the same conditions,the proposed algorithm has better prediction performance than the PLS algorithm and the KPLS algorithm.