What Are Rule-Based Expert Systems?

An expert system is an intelligent computer program system that contains a large amount of expert-level knowledge and experience in a certain field. It can use the knowledge of human experts and problem-solving methods to deal with problems in this field. In other words, an expert system is a program system with a large amount of expertise and experience. It uses artificial intelligence and computer technology to make inferences and judgments based on the knowledge and experience provided by one or more experts in a certain field to simulate human experts. Decision-making process in order to solve the complex problems that need to be handled by human experts, in short, the expert system is a computer program system that simulates human experts to solve domain problems.

Chinese name: expert system
Foreign name: Expert system

Solid: Intelligent computer program system
Features: Simulate human experts to solve domain problems

Expert System Basic Information

Expert System Introduction

Expert system is the most important and most active application field in artificial intelligence. It has realized a major breakthrough in artificial intelligence from theoretical research to practical application, from the discussion of general reasoning strategies to the application of specialized knowledge. Expert system is an important branch of early artificial intelligence. It can be regarded as a type of computer intelligent program system with specialized knowledge and experience. Generally, the knowledge representation and knowledge inference technology in artificial intelligence are used to simulate what can usually be solved by domain experts. Complex issues.

Origin and Development of Expert System

In the early 1960s, some common uses of logic and simulated mental activity emerged.

expert system Problem solving programs that can prove theorems and perform logical reasoning. However, these general methods cannot solve large practical problems, it is difficult to transform the actual problems into a form suitable for computer solving, and it is difficult to deal with the huge search space required for solving problems. In 1965, EA Feigenbaum et al. Developed the world's first expert system dendral on the basis of summing up the successes and failures of general problem-solving systems, combined with expertise in the field of chemistry, to infer chemical molecular structures. For more than 20 years, the research of knowledge engineering, the theory and technology of expert systems have continued to develop, and applications have penetrated into almost all fields, including chemistry, mathematics, physics, biology, medicine, agriculture, meteorology, geological exploration, military, engineering technology, law, In the fields of commerce, space technology, automatic control, computer design and manufacturing, we have developed thousands of expert systems, many of which have reached the level of functions, even exceeded the level of human experts in the same field, and have been produced in practical applications. Great economic benefits.

The development of the expert system has gone through three stages, and it is transitioning and developing towards the fourth generation. The first generation of expert systems (dendral, macsyma, etc.) are characterized by a high degree of specialization and a strong ability to solve specialized problems. However, there are defects in the integrity, portability, transparency, and flexibility of the system, and the ability to solve problems is weak. The second-generation expert system (mycin, casnet, prospector, hearsay, etc.) is a single-disciplinary professional and application-type system. Its architecture is relatively complete, and its portability has also been improved. In addition, the system's man-machine interface, interpretation mechanism, and knowledge There have been improvements in acquisition techniques, uncertain reasoning techniques, enhanced knowledge representation of expert systems, and heuristics and generality of inference methods. The third-generation expert system is a multi-disciplinary comprehensive system, which uses multiple artificial intelligence languages, comprehensively uses various knowledge representation methods, multiple inference mechanisms and control strategies, and begins to use various knowledge engineering languages, skeleton systems, and expert system development Tools and environment to develop large integrated expert systems. After summing up the design methods and implementation technologies of the previous three generations of expert systems, large multi-expert collaboration systems, multiple knowledge representations, comprehensive knowledge bases, self-organized problem solving mechanisms, multidisciplinary collaborative problem solving and parallel reasoning, and experts have begun to be adopted. The latest artificial intelligence technologies such as system tools and environment, artificial neural network knowledge acquisition, and learning mechanism are used to implement a fourth-generation expert system with multiple knowledge bases and multiple agents.

Expert System Construction

Expert system usually consists of 6 parts: human-computer interaction interface, knowledge base, reasoning machine, interpreter, comprehensive database, and knowledge acquisition. Among them, the knowledge base and the inference engine are separated from each other, which is unique. The architecture of the expert system varies with the type, function, and scale of the expert system.

In order for computers to use expert domain knowledge, knowledge must be expressed in a certain way. Currently commonly used knowledge representation methods include production rules, semantic network, framework, state space, logical mode, script, process, and object-oriented. The rule-based production system is the most basic method to realize knowledge utilization. The production system is composed of three main parts: a comprehensive database, a knowledge base, and an inference engine. The comprehensive database contains facts and assertions from around the world that solve problems. The knowledge base contains all knowledge rules expressed in the form of "if: <premise>, then: <result>". The task of an inference engine (also known as a rule interpreter) is to use control strategies to find applicable rules.

Expert System Knowledge Base

The knowledge base is used to store knowledge provided by experts. The problem solving process of the expert system is to simulate the expert's thinking mode through the knowledge in the knowledge base. Therefore, the knowledge base is the key to the quality of the expert system. That is, the quality and quantity of the knowledge in the knowledge base determine the quality of the expert system. Level. Generally speaking, the knowledge base and expert system programs in the expert system are independent of each other. Users can improve the performance of the expert system by changing and improving the knowledge content in the knowledge base.

Knowledge representation in artificial intelligence includes production, framework, semantic network, etc., and the knowledge that is more commonly used in expert systems is production rules. Production rules appear in the form of IF ... THEN ..., just like conditional statements in programming languages such as BASIC. IF is followed by a condition (previous) and THEN is followed by a conclusion (post). Conditions and conclusions can be Compound by logical operations AND, OR, NOT. Here, the understanding of production rules is very simple: if the prerequisites are met, a corresponding action or conclusion is produced.

Expert system inference engine

The inference engine repeatedly matches the rules in the knowledge base against the conditions or known information of the current problem, and obtains new conclusions to obtain the result of the problem solution. Here, there can be two types of reasoning: forward and backward.

The strategy of the forward chain is to find out those rules whose premises can match the facts or assertions in the database, and use the conflict elimination strategy to select an execution from these satisfying rules, thereby changing the content of the original database. This is repeated until the facts of the database are consistent with the goal to find a solution, or stop when no rules can be matched.

The strategy of the reverse chain is to start from the selected goal and find the rules that can achieve the results after execution. If the premise of this rule matches the facts in the database, the problem will be solved; otherwise, the premise of this rule will be new. Sub-goals, and find new rules that can be applied, and execute the premise of the reverse sequence, until the premise of the last rule applied can match the facts in the database, or until no rules can be applied, the system uses dialogue The form asks the user to answer and enter the necessary facts.

It can be seen that the reasoning machine is like the expert's way of thinking in solving problems, and the knowledge base realizes its value through the reasoning machine.

Other parts of the expert system

The man-machine interface is the interface when the system communicates with the user. Through this interface, users enter basic information, answer related questions raised by the system, and output inference results and related explanations.

The comprehensive database is used to store the original data, intermediate results and final conclusions needed in the reasoning process, which is often used as a temporary storage area. The interpreter can explain the conclusion and solution process according to the user's question, thus making the expert system more humane.

Knowledge acquisition is the key to the superiority of the expert system knowledge base, and it is also the "bottleneck" of the expert system design. Through knowledge acquisition, the content in the knowledge base can be expanded and modified, and automatic learning functions can be realized.

Expert system implementation

Early expert systems used universal programming languages (such as fortran, pascal, basic, etc.) and artificial intelligence languages (such as lisp, prolog, smalltalk, etc.), and were directly implemented through the cooperation of artificial intelligence experts and domain experts. Its development cycle is long and difficult, but it is flexible and practical, and it is still used by artificial intelligence experts. Most of the expert system development work has been implemented using the expert system development environment or expert system development tools. Domain experts can choose appropriate tools to develop their own expert systems, which greatly shortens the development cycle of the expert system, thereby providing expert systems in various fields. Provide conditions for a wide range of applications.

Expert system working process

The basic structure of the expert system is shown in the figure, where the direction of the arrow is the direction of data flow.

Expert system structure diagram

The basic work flow of the expert system is that the user answers the system's questions through the human-machine interface, the inference engine matches the information entered by the user with the conditions of each rule in the knowledge base, and stores the conclusions of the matched rules in a comprehensive database. Finally, the expert system will present the final conclusion to the user.

Here, the expert system can also explain the following questions to the user through the interpreter: Why does the system ask the user (Why)? How did the computer come to a final conclusion?

The domain expert or knowledge engineer can acquire knowledge in the expert system through special software tools or programming, and constantly enrich and improve the knowledge in the knowledge base.

Main development tools: Gensym G2, CLIPS, Prolog, Jess, MQL 4.

Expert system performance

Expert system functions

By definition, an expert system should have the following functions:

Store the knowledge needed for problem solving.

Store the initial data for solving specific problems and various information involved in the reasoning process, such as intermediate results, goals, alphabets, and assumptions.

(3) According to the current input data, using the existing knowledge and according to a certain reasoning strategy, to solve the current problem, and can control and coordinate the entire system.

Be able to make necessary explanations about the reasoning process, conclusions or the system's own behavior, such as problem-solving steps, processing strategies, reasons for choosing a processing method, the system's ability to solve certain problems, and how the system organizes and manages its own knowledge. This is not only easy for users to understand and accept, but also convenient for system maintenance.

Provide maintenance methods such as knowledge acquisition, machine learning and knowledge base modification, expansion and improvement. Only in this way can the problem solving ability and accuracy of the system be improved more effectively.

Provide a user interface that is not only convenient for users, but also easy to analyze and understand the various requirements and requests of users.

It is emphasized here that storing knowledge and applying knowledge for problem solving are the two most basic functions of expert systems.

Expert System Features

Expert system is a knowledge-based system, which uses the expertise provided by human experts to simulate the thinking process of human experts and solve problems that are quite difficult for human experts. Generally speaking, a high-performance expert system should have the following characteristics:

Inspiring. Not only logic knowledge but also heuristic knowledge can be used. It uses normative expertise and intuitive judgment knowledge to make judgments, inferences and associations to achieve problem solving.

Transparency. It enables users to communicate with each other without knowing the structure of the expert system, and to understand the content of knowledge and reasoning ideas. The system can also answer some questions about the behavior of the system itself.

(3) Flexibility. The separation of the knowledge of the expert system from the inference mechanism enables the system to continuously accept new knowledge, thereby ensuring that the knowledge in the system is constantly growing to meet the needs of business and research.

Expert System Classification

According to knowledge, technology can be divided into: logic-based expert systems, rule-based expert systems, semantic network-based expert systems, and frame-based expert systems.

According to the type of task:

Interpretive: It can be used to analyze symbolic data and explain the actual meaning of these data.

Predictive: Infer the future evolution of an object based on its past and present conditions.

Diagnostic type: find faults and defects of objects based on input information.

Debugging type: Give yourself a troubleshooting plan for the fault.

Maintenance: Specify and implement a plan to correct certain types of failures.

Planning: Develop an action plan based on a given goal.

Design type: Form required schemes and drawings according to given requirements.

Monitoring type: complete real-time monitoring tasks.

Control type: complete the implementation of control tasks.

Educational: A combination of diagnostic and debugging for teaching and training.

Expert system application

Areas of Expert System Application

The original expert system was an application of artificial intelligence, but due to its importance and the rapid development of related application systems, it is already a specific type of information system. The term expert system comes from the "knowledge-based expert system". This system uses human knowledge stored in a computer to solve problems that usually require experts to handle. It can mimic The reasoning process of human experts in solving specific problems can be used by non-experts to improve their ability to solve problems. At the same time, experts can also consider it as an assistant with professional knowledge. Because human resources are scarce in human society, with the expert system, this precious expert knowledge can be universally applied.

In recent years, expert system technology has gradually matured, and has been widely used in engineering, science, medicine, military, and commerce, etc., and the results are quite fruitful. Even in some application fields, it exceeds the intelligence and judgment of human experts. Its functional application areas are summarized as follows:

Interpretation-such as testing lung tests (such as PUFF).

Prediction-such as predicting maize losses (such as PLAN) that may be caused by black moths.

Diagnosis-such as the diagnosis of a bacterial infection in the blood (MYCIN). Another example is the CATS system for diagnosing the cause of automobile diesel engine failure.

Fault Isolation-such as telephone troubleshooting system ACE.

Design-such as MOTORBRUSHDESIGNER, an expert system for the design of small motor springs and carbon brushes.

Planning (Planning)-the well-known and assisted planning of the IBM computer main architecture layout, re-installation and re-arrangement of expert system CSS, and PlanPower expert system to assist property management.

Monitoring-such as monitoring YES / MVS of the IBM MVS operating system.

Debugging-such as BUGGY to investigate the cause of student subtraction arithmetic errors.

Repair-such as the expert system SECOFOR for repairing crude oil storage tanks.

Scheduling-such as the expert system ISA of manufacturing and transportation. Another example is a work shop manufacturing step arrangement system.

Instructionsuch as a TVC expert system that teaches users to learn operating systems.

Control-Control system PRANANS that helps Digital Corporation computer manufacture and distribution.

Analysis-such as the expert system DIPMETER for analyzing the well storage capacity and the DENDRAL system for analyzing the possible structure of organic molecules. It was the earliest expert system and one of the most successful.

Maintenance-after analyzing the cause of the telephone exchange failure, and COMPASS, an expert system that can advise humans how to repair.

Architecture design (Configuration)-such as the expert system XCON designing the VAX computer architecture and the expert system VT designing the new elevator architecture.

Targeting-for example how calibrated weapons work

Expert system example

Here, we take a simple "animal identification expert system" as an example to get a preliminary understanding of the working mechanism and system characteristics of the expert system. The system's knowledge base is a collection of production rules, and Figure 2 shows two of them. In principle, the rules are independent of each other, and the "predecessors" of any two rules are generally not repeatable and do not have an inclusion relationship. The rules of a small expert system can be several dozens, and the rules of a large expert system can reach thousands, like this animal expert system knowledge base, there are only six rules.

Expert system trend prediction

At this stage, the application of expert systems at home and abroad stays in a relatively narrow sense based on rule reasoning. The application is more targeted at laboratory research and some lightweight applications, which are far from meeting the needs of large-scale commercial applications and achieve real-time intelligence. The need for reasoning and big data processing.

The next step in the development of expert systems will be model-based reasoning, supplemented by rule-based reasoning, and meet the needs of business applications, meeting the needs for real-time and large-data processing.

At the same time, the expert system will develop in a more specialized direction, providing targeted models and products for specific directional needs, such as a fault diagnosis model based on the causal directed graph CDG and a process processing model.