A novel knowledge representation model based on factor state space

Abstract

Knowledge representation is the foundation of artificial intelligence, which is one of the hottest fields of artificial intelligence research. To solve the defect of tradition knowledge representation that a large amount of semantic heterogeneous information results in the difficulty of sharing and exchanging information, first, a framework of the knowledge description is given in the paper, in which all kinds of things and their properties can be expressed and described. Second, concepts of factor and factor space are presented and their formal description methods are proposed. Uncertainty reasoning based on factor state space is analyzed. Finally, oil and gas reservoir protection expert system was developed and verified correctness and practicality of the proposed model. Application of the expert system shows that the proposed model can solve uncertainty reasoning problem, the reasoning results are more correct, and provide the theoretical support for expert system. Research on knowledge representation on factor state space can guide relevant research and application on intelligent engineering.

Introduction

Knowledge representation is used to represent the facts and relations in domain knowledge by knowledge engineers. With the continuous development of society, artificial intelligence is making high progress. So far, a lot of knowledge representations have been put forward. First order predicate logic presentation is the earliest knowledge representation method, which is applied to solve problems in theorem method. But it is difficult to express heuristic knowledge and meta knowledge, not easy to realize non monotonic and inexact reasoning and it is easy to appear “combinatorial explosion” when there are more facts in reasoning process, and the efficiency is low. Production rule is natural, flexible and has the merits of good modularity and high generality. Currently, it is the most widely used in knowledge presentation technology. However, it has the disadvantages of low efficiency and poor expression ability. Frame-based representation is a good method for knowledge expression which can not only describe the details of things from the simple to the profound, but also detect conflict and achieve efficient reasoning. Petri net is another knowledge representation method, which can deal with parallel reasoning, but can’t handle Backtracking reasoning, backward reasoning and so on.

By the analysis of the above knowledge representations, they have many great differences among them, but they have a common characteristic that they can describe all complicated things as a set of “known” and set up a structured system. The reasons for doing this are that knowledge and intelligence can be expressed easily and be implemented easily in computers as well. Refs. [1], [2] gave a factor description method of knowledge and proposed unified framework which can describe concepts and relationships between concepts, the semantic concept from the relationship between the concepts. Refs. [3], [4] presented a simulation-type factor neural network model to forecast the potential risks of geological disasters in geological disasters defense expert system. Refs. [5], [6] put forward a new factor-oriented formal description model in SCADA network attacking and defense. A factor knowledge representation method was proposed to describe equipment information classification in [7], [8], [9], [10].

Uncertainty in artificial intelligence (AI) is the common form of intelligent reason; structured representation of uncertainty reasoning is a knowledge representation for uncertainty reasoning. In the paper, a knowledge representation method based factor state space was proposed to provide a unified framework of knowledge description and a structured representation for uncertainty reasoning. The rest of the paper is organized as follows: factor state space model is described in Section 2, uncertainty reasoning based on factor state space is introduced in Section 3, reservoir protection expert system is developed and the results are analyzed in Sections 4 State space representation of the oil and gas formation damage factors, 5 Uncertainty reasoning based factor neural network in formation damage expert system, and the study is concluded in Section 6.