Kernel Principal Component Analysis Algorithm Based On Parameterization PSO For Penicillin Fermentation Process Monitoring

Performance monitoring of industrial production process is a hot field of intelligent information processing and application.On line or off-line monitoring of industrial production operation status,detection of production interference and abnormal conditions,and identification of fault types and fault causes when faults occur can ensure the safety and maintainability of industrial production process,and improve the production quality.With the rapid development of process control technology and the wide application of DCS(Distributed Control System),various intelligent instrument and control equipment in the industrial process,large volumes of data are sampled and collected.It is one of the most active research area in the field of process control that how to fully utilize these collected data and mine deep-level information about process operation to improve the performance of process monitoring.The process monitoring method based on PCA(Principal Component Analysis)is a data-driven process monitoring algorithm.It can realize process monitoring not by the precise mathematical model but by the process data.So it obtained rapid development in recent several dozen years.The process monitoring method based on PCA is one of the most research and application in the multivariate statistical monitoring methods,however,presently the theoretical system is not perfect.Kernel learning method is based on structural risk minimization theory,which has good generalization ability for training set,has advantage for nonlinear process modeling.Kernel PCA(KPCA)is a process monitoring method.It maps nonlinear process data to high-dimensional feature space through kernel function,and then makes principal component analysis on characteristic variables.KPCA just calculate the eigenvalue instead of nonlinear optimization problem.Kernel parameter can severely impact the performance of process monitoring.So this thesis has presented the particle swarm optimization with dynamic accelerating constants(CPSO)to optimize kernel parameter through on-line monitoring fed-batch penicillin production(Pensim v2.0).The main research contents and conclusions are as follows:1.This thesis introduces the rationale of Particle Swarm Optimization(PSO),presented CPSO to optimize nuclear parameter.The simulation test shows: acceleration constant set as a function with evolution generations changes,which can speed up convergence and avoid trapping into local minima in the searching.2.Cross validation is complicated calculations and inefficiency.So this thesis presented a method which applies Parameterization PSO into Process Monitoring.In Fisher Discriminant Analysis(FDA),similar models dispersion will be as small as possible,distinct models dispersion will be as big as possible.With that mechanism we formulate the target function to optimize kernel parameter and improve KPCA performance.At last,using public beta set that IRIS proves the validity of the presented method.3.Apply Parameterization PSO into Process Monitoring of the Penicillin Fermentation Simulation.It turned out that the method can reduce dimension of feature vector and then get a few feature vector which is more representative from fed-batch penicillin process.So the method can effectively reduce the computational complexity,enhance the effectiveness of the fault monitoring,improve the efficiency of processing data and the result of fault recognition.At last the conclusions came out and the future work that is theory about process monitoring was schemed out.