Russia, Ufa, Ufa State Aviation Technical University

E-mail: vasilyev(aMizi.ugatu.ac.ru

Abstract

One of approaches to joint application of different kinds of knowledge in hybrid expert systems, functioning in the conditions of uncertainty, is considered in the paper. The technique of construction of similar systems for the problems of diagnostics and checking of gas- turbine engine parameters is offered.

Аннотация

В докладе предлагается один из подходов к совместному применению разных видов знаний в гибридных экспертных системах, функционирующих в условиях неопределенности. Предлагается методика построения подобных систем для задач диагностики и контроля параметров ГТД.

At present, different knowledge sets representation and application problem takes a central position in the general hybrid expert system design. The use of different knowledge sets is set by heterogeneous nature of data an expert applies performing technical and scientific applications. There are various types of knowledge are represented in the hybrid expert system, such as, conceptual, expert, experimental, etc. knowledge sets as well as appropriate algorithms to process it. The hybrid expert system is required to be able to represent a problem area model and to perform some logical reasoning with it. Depends on application area a data- base (including fuzzy databases) access necessity can appear. The main purpose of hybrid system design consists in the optimal combination of knowledge representation patterns and processing algorithms within unified expert system information area. In the other hands, the main purpose is the investigation of optimal different knowledge processing routine combination in order to improve the expert system performance for gas turbine engine diagnostic and control procedures within indeterminacy.

The mobility of an expert system is defined by the knowledge base mobility and its ability to progress and to gain experience from different information sources (databases, expert knowledge bases, etc.) as well inference procedures. Expert knowledge can be divided into exact and non-exact, complete and uncompleted, static and dynamic single-valued и multiple- valued, etc. Besides, the expert knowledge usually is imprecise and has a subjective character. Inaccuracy and ambiguity of knowledge cause expert system to deal with several alternative areas.

Depends on knowledge representation form and logic inference algorithm hybrid expert systems can be of the following type:

1) Inductive system. The knowledge sets are represented by examples of actual events and appropriate decisions (reactions). There is an inductive knowledge synthesis is used.

2) Productive rule system. There are forward, backward and combined inference algorithms are used.

3) Logical system. The knowledge sets are represented by functional calculus formula forms. The knowledge base is considered to be a one level theory, therefore expert system inference is a statement proving.

4) Object-oriented system. The knowledge sets are represented by the object and interrelations between it. There are various inference algorithms are used.

5) Hybrid system. There is a cooperative use of different representation of knowledge sets and appropriate inference mechanism both.

The knowledge set can be conceptual, constructive, procedural or factual. Depends on rigidity of expert system knowledge base the knowledge set can be expert (slight and soft) and conceptual (rigid and fundamental).

The conceptual knowledge set is a set of concept, features and interrelations between concepts a human being uses to process a diagnostic problems, for example. This type of knowledge set is elaborated in fundamental science branches as well as in different theoretical branches of various application and engineering sciences.

The constructive knowledge set is a knowledge about intended structure and interaction between various objects which an expert deals with. The knowledge set consists a basic content of the most technical sciences. It is close to the conceptual knowledge set and therefore can be formulated in terms of concept.

Procedural knowledge set is a set of methods and algorithms an expert dealt with and had got into its processing.

Factual knowledge set are qualitative and numerical characteristics of objects, events and its elements. This type of knowledge set can be formulated in the form of table, reference and database.

All the above-mentioned knowledge types are usually bound up otherwise they cannot be used. The expert knowledge plays an important role in the slightly structured problem areas where the formal models are absent. Nevertheless, the role of expert knowledge is still considerable in areas where the formal models are applicable, but, in this case, some experience is required in order to make a decision. Thus, the proper performance of expert system is provided by suitable representation of knowledge which is rather heterogeneous depending on problem area and user purposes. The problem appears is the problem of knowledge entry definition (what to represent) and the problem of representing algorithm selection (how to represent). So, the last one can be divided into two independent problems: how to organize the knowledge; how to represent the knowledge by selected means of definition.

To consider the approach to cooperative various knowledge set usage in unified diagnostic hybrid expert system within indeterminacy. To develop a technique for its implementation.

The modem expert system is aimed to solve a vast number of technical and scientific applications, for example, a gas turbine engine diagnostic application requires empirical knowledge and database access. This type of expert system may contain the following subsystems:

expert knowledge subsystem; object-oriented (conceptual) knowledge subsystem; database subsystem which performs database access by inference engine or expert system kernel.

The expert knowledge subsystem is an inductive instrumental expert system connected with the kernel via messages. The design of expert system dedicated to solving some technical and scientific applications involves expert and conceptual knowledge bases adjunction as well as database adjunction that comprises computational task models to be solved during consultations. The expert can represent an expert knowledge set dedicated to gas turbine diagnosis as a set of patterns. The internal form of expert knowledge representation is a set of decision paths (i.e. decision tree). The set of patterns is described by attributes and contents samples with the same structure . The set of examples may be connected by logic passes. In this case, the appropriate decision path is united in the way where the terminal path of decision tree includes another decision tree.

The attribute of the sample structure definition can be of four types. This type defines how the system acts during interaction with expert system: selection attribute (the sample value is a condition (including fuzzy condition) which user can select during expert system consultation); sequence attribute (the value is the expert inference engine input sequence to be processed); comparative attribute (the value is the condition related with database and computational model objects); consequence attribute (the value is a final decision or logic pass in the expert system).

An interaction between expert knowledge subsystem and inference engine is organized through sentence attributes and comparison attributes of expert system as well as special conceptual and procedural knowledge sets.

Database expert system is a database management system extended with the knowledge set and logic inference mechanisms to provide knowledge-based applications. The creation of. database expert system is based on the following general principles: to identify database structure and queries to data a special conceptual knowledge set is used; expert knowledge set defined in database allows to tune expert system for solving a gas turbine engine parameter diagnostic and control problem within indeterminacy. This knowledge set is used to specialize the selection criteria for data queries. The data query is considered to be a usual computative task. The result may be used for consecutive iteration or for expert system management task. The expert knowledge files can be automatically formed from database files by converter (a special programme aimed at data importing and exporting).

A computational model of expert system and database aimed at diagnostic problem solving within indeterminacy can be formulated in the following general form:

where A - knowledge base and database attribute set; - domains (knowledge base and database attribute values); В - functionality set defined upon attribute set; - type definition set used in; - set of fuzzy relation for A.

Performing malfunction diagnosis the hybrid expert system provides a fault seeking in depth and width. In this case, the goal is divided into a subset of goals among which the optimal goal is selected to match the most detailed level of problem description. So, performing gas turbine engine diagnosis based on typical failure set, the hybrid expert system will specialize the appropriate fault symptoms until the supposal is not refused. While seeking in width the expert system analyzes all symptoms, which are in the same level of state space even they are related with the different failures, then it goes to the next detail level symptoms.

The complex dynamic object diagnostic problem solving (gas turbine engine diagnosis, for example) by hybrid expert systems (C-PRIZ, Clips 6.04 and RTworks3.5) includes the following stages: conceptual design; formulation; programme code generation; solution; abstraction; specification; analysis; data input.

In order to solve this problem by hybrid expert system (C-PRIZ, Clips 6.04 and RTworks3.5) the reference gas turbine engine model is stored in knowledge base and is corrected as expert system gains an experience. The real pattern is formed in database connecting with reference model through user queries (in automatic or semiautomatic mode). The automatic program code synthesis within the above-mentioned hybrid expert system allows to generate a complete program in the automatically mode or in the mode of user queries which are being executed provide the final task solutions.

Relational database management system for gas turbine engine diagnostic and control purposes provides an user by the following subset of functions: the selection based on logic conditions; relationship projection; creation of a new relation matches the certain database structure which includes fuzzy attributes.

In the hybrid expert system, a conceptual gas turbine engine model can be represented as a graph nodes of which are objects and edges are relations.

During the gas turbine diagnosis and control process, hybrid expert system provides the following things: the diagnostic efficiency for all nominal and non-nominal gas turbine engine states during operational development, for example; the ability of working gas turbine engine diagnosis (on test bed or in flight); the on-line diagnosis ability due to high computer system performance (RICS-processors); the automated diagnostic information processing and gas turbine engine diagnosis ability; information flow distribution and functional division among database, knowledge base, expert knowledge base, expert system kernel, its components and computer system resources during complex object diagnosis and control processes.

The proposed approach to different knowledge base organization within hybrid expert system can be extended by the various network technologies (client-server API for distributed database and knowledge base). However it is necessary to consider the certain requirement of knowledge representation the different hybrid expert systems request. Moreover, as far as knowledge set is heterogeneous (frame, semantic net, database, concept of knowledge base, neural net, fuzzy logic, genetic algorithm, etc.), it is difficult to combine the set within unified information space of hybrid expert system. The use of object-oriented approach (i.e. conceptual approach) allows partially eliminate this drawback and to increase the mobility of hybrid expert system.

Here is a methodology of gas turbine engine diagnostic and control hybrid expert system different knowledge set design: 1. Conceptual model design (formalization of problem area). 2. Selection and adaptation of diagnostic algorithm. 3. Representation of gas turbine engine diagnosis model as a set of separated concepts (knowledge sets) in the knowledge base. 4. Creation of knowledge base including rule base as a intelligence kernel control component. 5. Description of gas turbine engine diagnostic and control knowledge set in the various hybrid expert system components (database, knowledge base, expert knowledge base, graphical data- base, etc.). 6. Fuzzy logical software design. 7. Distribution of information flows between ex- pert system and its various components. 8. Testing of expert system within user queries. 9. Testing of hybrid expert system within indeterminacy.

The proposed approach to different knowledge set hybrid system design within indeterminacy allows the following: to facilitate the adaptation of the object under diagnostic by means of object-oriented technology; to apply different knowledge sets (conceptual, constructive, procedural, factual knowledge sets, rule base including membership functions, rules and fuzzy rules in database, knowledge base, expert knowledge base) including different inference algorithm for efficient diagnostic problem solving; to generalize and to improve the conceptual representation model of different knowledge sets in relational databases (Access, FoxPro, Informix) combined with a kernel of basic hybrid expert system (C-PRIZ, Clips6.04 and RTworks 3.5); to solve the information flow optimization and distribution problem of various expert system components within indeterminacy.

The proposed approach to the different knowledge set hybrid expert system design within indeterminacy has completely improved the lightness of designed methodology and its applicable for such intelligence system creation for gas-turbine engine diagnostic and control problem processing.