Identification Method Study Of Sedimentary Environment Based On Fuzzy Automata

The change of sedimentary size parameters is dominated by some factors, such as transportation medium and method, sedimentary environment and climate, therefore, the analysis of sedimentary size parameters is much important to reveal climatic change and environmental evolution. Size parameters, such as mean size, standard deviation, deviation and peak value are usually deemed as the four sedimentary size factors. The formation of different size components is closely related to the sedimentary environment and transportation. We can further determine the sedimentary environment by processing and analyzing the data of size, thus, it is of momentous significance and realistic importance undoubtedly to the study of contemporary sedimentology, even to the sedimentary environment of ancient sediment.A fuzzy neural system is a fuzzy system that uses a learning algorithm derived from or inspired by neural network theory to determine its parameters (fuzzy sets and fuzzy rules) by processing data samples. By combining the fuzzy technique and the neural network, we can make full use of the advantages of them both, and fill the gap between them, effectively, which also makes these searches on fuzzy neural system much more famous. Furthermore, the fuzzy neural system has been successfully applied to the fields, such as automatic control, cluster analysis, pattern recognition and so on, which creates a new direction to the development of artificial intelligence.It is Chomsky Hierarchical Grammar that the foundation of this thesis is, which is proposed by Noam Chomsky. By introducing the basic theories of fuzzy automata, we briefly review some applications of neural network technology in the current fuzzy automata study, which clarifies the relationship among fuzzy finite automata (FSA), fuzzy grammar and neural network in a certain degree.Through the training of neural network, we are able to extract the automata that we need, then with this automata, we can get a way to carry out fuzzy grammatical inference and identify the corresponding sedimentary environment. There are a large numbers of studies and applications on fuzzy neural network, therefore, in this thesis, we mainly focus on the fuzzy grammatical inference based on fuzzy neural network and identification method study of fuzzy automata. On the aspect of fuzzy regular grammatical inference, there are two algorithms available, namely real-time recurrent learning algorithm(RTRL) and real-coded genetic algorithm(RCGA), being used to train second order recurrent neural network, both algorithms above have the same shortcomings: high time complexity and extremely slow training rate. When it comes to the samples of fuzzy finite state automata, there are different kinds of exceptions coming out, that is to say, both of them have seriously weak generative ability. At length, RTRL algorithm is unstable while RCGA algorithm may easily trap in prematurity. As a result, we introduce some improved algorithms, such as Levesbeg-Marguard Genetic Algorithm(LMGA) to RCGA and Levesbeg-Marguard Back Propagation(LMBP) to RTRL, where the former aims to solve the problems of the rate of training and prematurity, and the latter aims to not only rate,but also throughput and some specific cases, such as long long strings.Furthermore, we have shown experimental simulations and verifications together with summarizing our conclusions in this thesis.