Research On The Representation Method Of Knowledge Based On Petri Network

As one of the core issues of knowledge engineering, knowledge representation is mainly research to solve problems with effective method and convenient for computer processing. A lot of research and practice have proved that knowledge representation not only has a certain impact on knowledge acquisition and learning mechanism, but also has great impact on the efficiency and the application range of the knowledge processing. Reasonable and effective knowledge representation method is the key to the success of knowledge system.Petri nets has intuitionistic model, rigorous mathematical theory and effective model analytical methods, which is an effective tool to study parallel, asynchronism, uncertain and distributed systems. Description and process knowledge are intergrated by knowledge representation approach based on Petri nets, knowledge reasoning can be reaized in this way. The merits of this approach include graphical representation of structure and knowledge, reasoning and search rapidly and reasoning implemented by math. On the ground of the developed fruit of Petri nets, intensive analysis and study with regard to key issues of fault diagnosis based on Petri nets such as modeling and reasoning algorithm, verification of knowledge base, fuzzy and uncertain knowledge representation and reasoning and self-learning ability are carried out in the paper. The artificial intelligence theory system is enriched and artificial intelligence technology is promoted by the theory and approach studied in the paper. It has, therefore, great theory meaning and application value. The main work and fruit of this thesis as follows:1. Knowledge representation modeling and reasoning algorithm approach based on Petri nets is put forward. Traditional knowledge representation approach based on Petri nets is using basic Petri nets to describe and reasoning. The limitations such as conflict, collision and memoryless are analyzed firstly. Then knowledge Petri nets model and its formal reasoning algorithm which are more appropriat for the description and analysis of knowledge are presented in the paper.2. Verification of diagnositic knowledge base based on Petri nets is proposed. Forward knowledge base verification based on production rules is not adequate to the inconsistency verification between rules. The approaches of redundancy, inconsistency, circularity and imcompleteness rules of knowledge base based on Petri nets are proposed. The validity of the approach is validated by an instance.3. Fuzzy and uncertain knowledge representation and reasoning mechanism based on a new model named weighted fuzzy Petri nets are proposed. Though fuzzy Petri nets model has the ability of fuzzy knowledge representation and reasoning, it does not consider the different influence degree of proposition to conclusion. In view of the limitation of the representation based on fuzzy Petri nets, the more accurate, reasonable and effective model named weighted fault Petri nets is presented. Traditional reasoning algorithms are based on graphical, which have low reasoning effiency and improper to be implemented by computer. Formal reasoning algorithm is presented, which has high reasoning effiency and can be implemented by computer convenietly.4. The self-learning ability of fuzzy Petri nets is studied. Lacking of self-learning ability is a main deficiency of fuzzy Petri nets. Artificial neural network has strong self-learning and self-adaption ability. The self-learning algorithm of artificial neural network is incorporated into fuzzy Petri nets model, by which making fuzzy Petri nets has self-learning ability.