基于模糊自动机的沉积环境判别方法研究

发布时间：2018-07-27 16:31

【摘要】：搬运介质、搬运方式、沉积环境和气候等因素控制着沉积物粒度参数的变化,因此,沉积物粒度分析对揭示气候变化和环境的演变具有重要意义。粒度参数中平均粒径、标准偏差、偏差和峰值是沉积物粒度的四个重要参数。不同粒度组分的形成与搬运均与沉积环境密切相关,处理、分析粒度数据有利于进一步确定沉积环境,这对于现代沉积学研究,乃至古代沉积物的沉积环境分析无疑都具有重要的理论和现实意义。模糊神经系统是这样一个模糊系统,它在处理数据样本过程中,使用某种学习算法去确定系统中的各项参数如模糊集和模糊规则,而该学习算法则由神经网络原理学习或激发出来的。利用模糊技术与神经网络的结合,我们可以充分发挥两者的优势,有效弥补各自的不足,这使得有关模糊神经系统的研究得到追捧,在包括自动控制、聚类分析和模式识别等领域中已得到成功应用,为人工智能的发展又增添了新的方向。本文是以诺姆·乔姆斯基提出的Chomsky体系文法为基础的,通过引入模糊自动机基本理论,简要概述了当前模糊自动机研究中有关神经网络技术的应用情况,这其中主要是以模糊有限状态自动机、模糊文法和神经网络这三者的关系为主线的,而通过对相关的神经网络进行训练,我们就可以抽取出所需的自动机,然后利用抽取的自动机进行文法的推导,以及对沉积环境进行判别分析。有关模糊神经系统的研究与应用已经具有一定的规模,但对于本论文来说,主要是研究基于神经网络的模糊正则文法推导算法以及模糊自动机的判别方法相关的问题。在模糊正则文法推导中,训练用于抽取模糊自动机的二阶反馈神经网络的算法主要有实时递归学习算法(RTRL)和实数编码基因遗传算法(RCGA),但是这两种算法存在通病：时间复杂度高,训练速度极慢。另外针对模糊有限态自动机(FSA)样本,两种算法都有不同种类的异常情况出现,也就是说泛化性很弱。最后,RTRL稳定性很弱；RCGA早熟现象经常出现。因而本文给出针对RCGA的改进算法：Levesbeg-Marguard Genetic Algorithm(LMGA),改进了训练速度以及解决了早熟的问题；针对RTRL算法,第七章给出了Levesbeg-Marguard Back Propagation(LMBP)算法,不仅加快了训练速度,增加了吞吐量,而且还能处理特殊情况下的字符串,比如超长串的情况。本文还给出了实验仿真与验证,并归纳了本文的结论。
[Abstract]:Factors such as transport medium, handling mode, sedimentary environment and climate control the change of grain size parameters. Therefore, sediment grain size analysis is of great significance to reveal climate change and environmental evolution. The average particle size, standard deviation, deviation and peak in grain size parameters are four important parameters of sediment granularity. The formation and transportation are closely related to the sedimentary environment. Processing and analyzing the grain size data are beneficial to further determine the sedimentary environment. This is of great theoretical and practical significance for modern sedimentology research and even the sedimentary environment analysis of ancient sediments. Fuzzy neural system is such a fuzzy system, which is dealing with data samples. In the process, some learning algorithms are used to determine the parameters of the system, such as fuzzy sets and fuzzy rules, and the learning algorithm is learned or excited by the principle of neural network. By combining the fuzzy technology with the neural network, we can make full use of the advantages of the two and make up for their own shortcomings, which makes the fuzzy God concerned. After systematic research, it has been successfully applied in the fields of automatic control, cluster analysis and pattern recognition. It has added a new direction to the development of artificial intelligence. This paper is based on the Chomsky system grammar proposed by Norm Chomsky, and briefly outlines the current theory of fuzzy automata by introducing the basic theory of fuzzy automata. In the study of fuzzy automata, the application of neural network technology is mainly based on the relationship between fuzzy finite state automata, fuzzy grammar and neural network, and by training the related neural networks, we can extract the required automata, and then use the extracted automata to carry out the text. The derivation of the method and the discriminant analysis of the sedimentary environment. The research and application of the fuzzy neural system have already had a certain scale. However, for this paper, it is mainly to study the fuzzy regular grammar derivation algorithm based on neural network and the related problems of the discriminant square method of fuzzy automata. The algorithms for training two order feedback neural networks for extracting fuzzy automata mainly include real time recursive learning algorithm (RTRL) and real coded genetic genetic algorithm (RCGA), but these two algorithms have common faults: time complexity is high, and training speed is very slow. In addition, the two algorithms have different kinds of algorithms for fuzzy finite state automata (FSA) samples. The exception of the class, that is to say, the generalization is very weak. Finally, the RTRL stability is very weak; the RCGA precocious phenomenon often appears. Therefore, this paper gives an improved algorithm for RCGA: Levesbeg-Marguard Genetic Algorithm (LMGA), improves the training speed and solves the problem of precocious. For RTRL algorithm, the seventh chapter gives Levesbeg-Ma The rguard Back Propagation (LMBP) algorithm not only speeds up the training speed, increases the throughput, but also can handle the strings in special cases, such as the super long string. This paper also gives the experimental simulation and verification, and concludes the conclusion of this paper.
【学位授予单位】：广东工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P512.2;TP301.1

【参考文献】