Research On Statistical Performance Monitoring And Control Of Batch Processes Based On The Mixed-Kernel Function

Batch processes, as an important industry production, which have been widely applied to bio-pharmaceuticals, food and fine chemical etc. To monitor the intermittent process to detect the abnormal situation in the process of production, it’s great significance to ensure safe and reliable operation of the production. However, intermittent production process data has many batches, multistage, nonlinear, dynamic and other complex characteristics, So it’s more difficulty to performer process monitoring and controlling on batch processes.The intermittent process modeling has become an important research topic.The intermittent production process monitoring often is based on three-dimensional data, according to the statistical principle often require large amounts of data samples. Based on the analysis of batch process data of nonlinear dynamic characteristics, with the gradually mature dynamic and nonlinear technology as a fundamental, the kernel technique is introduced into the DPCA method and a DKPCA-based statistical performance monitoring in a single batch data for modeling data for the nonlinearities and dynamics of batch processes. The method with a batch of data as the object, extension of time series and the nonlinear mapping of input space to feature space, and in the feature space to compute statistical characteristic, to describe the batch modeling data structure characteristics of the data. This method is good to be single batch data modeling, overcome the problem of on-line monitoring data of unequal length and does not need to estimate the future data. But need to determine the effective core function and parameters.According to DKPCA kernel function selection problem, this paper proposes a combination kernel function method which can well describe the nonlinear process. The selection of kernel function is a combination of gauss kernel and polynomial kernel function, and apply optimization to select the appropriate kernel function and its parameters. Through the analysis of the principle of statistical monitoring, identified in the SPE control subject chi-square distribution for the first goal and minimum of the SPE statistic average of sliding windows and SPE statistic average ratio of accumulative for the second goal for set up the optimization model. According to the model of nonlinear which is difficult to mathematical expression, using the particle swarm algorithm parameters and weight coefficient of mixed function optimization.In penicillin fermentation process, a lot of control parameters have important effects on the fermentation process. The changes of these parameters monitor online can make people understand the whole process of fermentation condition, which can help to make the corresponding adjustment, to ensure the stable operation of the fermentation process. Though analyzing the important factors penicillin fermentation process,10 variables are selected as monitoring variables, they are aeration rate, agitator power, temperature and so on. The aeration rate, agitator power, substrate feed rate of three step fault simulation results show that, the proposed method is effective.