Study Of Spatial And Temporal Dimensions Dynamical System Modeling Based On Multi-polymerization Process Neural Networks

Dynamical system modeling has matured in solving the theories and methods of linear systems, and it also kept in mind in some complicated nonlinear systems. But many practical systems are complicated and time-variant, and even have many uncertain factors. It is difficult to describe the procedure by accurate mathematics models. It hard to building-up such model by traditional dynamical system modeling methods, and the accuracy of such model is very lower. MPPNN is a new artificial neural networks model which was the expanding of traditional neural networks. The input/output and connection weight function of the networks can be consecutive multivariate function which was varying by time(or procedure).MPPNN models are good at non-linearity, generalization, fault tolerance, and it has the features of massive parallel distribution processing and favorable regime of adaptive learning. It has good flexibility to using the MPPNN in the study of temporal dimensions dynamical system modeling.This thesis has analyzed the favorite methods of system modeling, combines the problem of temporal dimensions dynamical system modeling technology, study out the MPPNN models which are fit for such problems. This thesis has building up temporal dimensions MPPNN model. MPPNN has enormous calculated quantities and information transformation mechanisms of the MPPNN are very complicated, think of the problem above this thesis has come up with two kinds of MPPNN parallelization tactics which are based on the concurrency of structure and data. Moreover, it has come up with two kinds of MPPNN studying algorithms that are on account of Legendre orthogonal basis evolving and modified differential evolutionary algorithm. The speed and accuracy of the networks are improved much more. In the end, it has come up with a kind of reduced discretization MPPNN model, and the model is applied to four-dimensional dynamical system modeling of oil deposit. The validity of the model is proved by experiment.Studying of temporal dimensions dynamical system modeling technology based on MPPNN can use the MPPNN have the feature of dealing with temporal dimensions information and nonlinear transforming the information of input/output and intrinsic. The structure of the networks are flexible, it has excellent adaptability in solving problems in mechanisms. The MPPNN can also be used in some other fields, such as information retrivaling, information filtering and data mining, it has favourable actual using outlooks.