Statistics Modeling Of The State Space And Monitoring Of Chemical Process Systems

The process monitoring is an important constituent of computer integration process system. It is necessary to the process safety and the quality stabilization. The process modeling is core of the process monitoring, the feasibility of modeling and the validity of model is contradictory each other. Based on first principle of process and according to lumped strategy, a new concept, the process statistics state variable is put forward in this paper. Determination criterion of the variables is presented too. Combining Blind signal processing algorithm with frame of process state space, the paper realized to the reasonable and effective process monitor and the contradiction process modeling has been unification. The concept and the method which propounded in this paper are confirmed by the representative case -- Tennessee Eastman Challenge Process. The model has better validity and the rationality than data-driven model.The state space theory is the foundation of modern control theory. The state space modeling has a higher validity. This paper point out that a process system may be described by certain state variables, those state variables are lumped parameter and accord with dimension relations of process system. The criterion is put forward to determine those variables. Combining state space theory with process system, A primary process state space theory frame is constructed.Consistency between state space modeling and multivariate statistical modeling is pointed out from the contrast of the two modeling methods. A visual process description is given according with process state space frame and multivariate statistical algorithm. This paper has carried on the discussions about some key questions of multivariate statistical modeling, such as choice of statistical variable, the system linear approximation and the noise influence and so on.Blind signal processing (BSP) is introduced into process system engineering research. Using the independent component analysis(ICA) which is a self-organizing multivariate BSP algorithm, this paper get and identify those process statistics state variables, the process statistical state space model also been established. Competed to model of principal components analysis (PCA), the model of ICA is more valid.Based on the statistics state variable process, the algorithm and frame which monitor process state is established ,the frame include the process modeling, the model verification, the state monitoring and the fault diagnosis. Comparing to existing multivariate statistics monitoring frame, this frame have more improvements, those improvements include reconstruct of system, identifiability and separability of fault. The application of the frame is carried to TE process, all disturbances provided by the case has been accurately monitored. It is been realized that the process state estimation and the fault diagnosis.