Expert system for parametric modeling.

Pekshujev, Deniss ; Smirnov, Anton ; Kramarenko, Sergei 等

1. INTRODUCTION AND OBJECTIVES

Due to the steady industry growth the speed of the manufacturing processes should be increased and the quality of the products should be improved. At the first sight the computerization of both production and design process simplifies the problem. However actually along with simplification of a problem caused by many available software applications, which accelerate work of an engineer (computing programs, CAD/CAM/CAE PLM systems, etc) the qualified labor is required to work efficiently with all these applications (Jackson, 1998). Today professional must not be only the expert in own field, but also should be able to work with all required applications. The theoretical skills can be received from the educational institutions, but the applied knowledge that is required for fulfillment of tasks, can be achieved only by practice. It forces enterprises to make significant investments into education and training of labor. It takes up to 3 years before the required professional level could be achieved. Enterprises today are dependent upon the expert knowledge. The reassignment of trained labor will entail inevitable losses for enterprise (Tausend & Foht, 1990).

1.1 Expert systems

Manufacturing enterprise problems can be solved through the use of expert systems. The expert system is the software which uses the knowledge of experts in order to provide effective support of decisions made by users. Expert systems differ from regular applications, due to more complicated architecture. It consists from the knowledge database, problem solver and support component. Support component simplifies user work with main application (Giarratano & Riley, 2004).

Expert system ensures that the knowledge is obvious and accessible. It distinguishes expert systems from traditional applications. The main features of expert systems are defined as:

* Decision support possibility. An expert system stores experience of qualified experts. It enables to perform creative and effective decisions in selected field with possibility of tracking the reason behind proposed decision.

* Forecasting availability thanks to which expert systems is able to give solution accordingly to forecasted situation. If the solution is changed it is possible to discover what changes are responsible for the changed solution.

* Support of the institutional memory feature or knowledge base developed in cooperation with organization which represents the current working group policy. This set of knowledge becomes the database qualified solutions which is permanently updated by the best strategy and the methods used by the field experts. This enables to retain the knowledge when leading experts leave the enterprise.

* Training feature of the Expert systems. System improves the quality of personal training. New employees are given extensive baggage of experience and strategy, based on which it is possible to study the recommended policy and methods (Subbotin, 2008).

2. EXPERT SYSTEM DEVELOPMENT

The experts, engineers, knowledge, and tools for expert system design and development are required for the new system building. Frequently the experts used are employees of the enterprise. Due to the lack of information it is often believed by executives that development of the expert system is a complex, long and expensive process, and it is difficult to access the returns. Those are the main reasons why enterprises refuse from such system development such systems (Gavrilova & Horoshevskij, 2000).

The purpose of given work is to present that it is possible to develop expert systems by the use of resources available (such as Excel and Inventor applications). The engineers who bring knowledge to the Expert systems might be employees of the enterprise, since more qualified specialists in system design and programming usually lack for competence in enterprise activities, and because it saves more money. One of the most acceptable tools in modeling expert systems is MS Excel, which has found a great popularity in industry, especially with Visual Basic application.

Theoretically expert system possesses some properties of artificial intelligence which allows find the solutions that were not in the system before. However, such systems are very complicated in practice. Virtually, we need a system able to find a solution from the earlier experience (data base) according to the set of given parameters. Such a system looks more available.

Let's take as an example the production process of air-cooling air-cooled heat exchangers. There are about 8000 standardized air-cooled heat exchangers versions in petro-chemical industry. The description of their varieties represents a combination of 9 variables. However, in practice the combination of variables may increase considerably. For instance, a detail thickness which works under pressure is influenced by a number of parameters as: material physical properties, temperature of the working environment, pressure, character of exploitation, and so forth. Therefore, even the standard version of air-cooled heat exchangers has many modifications, not influencing the operation process, but causing changes in the construction. That is the case when an expert system could facilitate an autonomous modeling of the air-cooled heat exchangers with new or existing parameters for a given purposes.

Initially, a basic model needs to be build for an expert system in order to make any modifications. The modeling software used for that is Inventor, which is able to create parametric dependencies. The basic model has a sketch, which is crucial in determining these dependencies (Fig. 1). Then all the dimensions from the parametric table are tied with the results of expert system.

[FIGURE 1 OMITTED]

When the sketch is ready, the air-cooled heat exchangers details and units are modeled by a certain sketch, as an imported element, and finally produced. This order helps avoiding geometric mistakes when we change basic parameters. Within a detailed modeling all possible dependencies appear and are set in equation form. Basic parameters affected by calculation results are specified in a literal form (Fig. 2).

[FIGURE 2 OMITTED]

After creating the data base and the expert system all the results are attached to the model and fixed to the literal form as variables. This operation is performed by a function of button "Link". Finally we get a complete model--basis for output results of the expert system. Now just using a function "up- date", the expert system models the new air-cooled heat exchangers (Fig. 3).

[FIGURE 3 OMITTED]

2.1 The process of expert system development

The expert system has to follow two conditions:

* data input need to be performed avoiding incorrectness of data input;

* engineers may not use any additional sources (books, manuals, etc) for data input.

For that purpose the expert system should contain all the available methodologies, calculation characteristics. To export these data a function "Index" is used. If there is exist a standardized row of dimensions (for example diameter of ventilator), then we use function "Validation/List".

For the completeness of our expert system an option of hints and constraints has to be created. A command "Conditional formatting" is used for the text color changing. Additional notes could appear in information cells: if(dt<0; "increase tube thickness"; ""). "Data validation" creates pop-up hint windows. The data input is organized on a simple and intuitive level. In our expert system a "one-button" approach is used. It means that a user has just to select the descriptive code of seeking model to be designed and press the button "Design". The system will calculate the needed parameters itself (Fig. 4).

[FIGURE 4 OMITTED]

3. CONCLUSION

Our paper demonstrates that the development process of an expert system seems not complicated in Excel (Waterman, 1989). It is important to enrich the system with different option for intuitive input and rigorous outputs. Duplicate parts modelling and different variations making allows decreasing time of design, usability by engineers without outstanding skills and experience. In the near future our expert system is going to be enhanced to support manufacturing processes of other products, also being easily used by mid-skilled workers to reach the sufficient results.

4. ACKNOWLEDGEMENT

Hereby we would like to thank the Estonian Ministry of Education and Research for targeted financing program "DoRa" that enabled us to carry out this work.

5. REFERENCES

Gavrilova, T. & Horoshevskij, V. (2000). Knowledge base in expert systems. Textbook, St.Petersburg

Giarratano, J. & Riley G. (2004). Expert Systems: Principles and Programming. Course Technology, ISBN 0534384471

Jackson, P. (1998). Introduction to Expert Systems. Addison Wesley; 3ed., ISBN 0201876868

Subbotin, C. (2008). Knowledge processing in expert systems and making decisions. Scientific book. Ukraine, ZNTU.

Tausend, K. & Foht, D. (1990). Expert system development and program realization for computers. Financial and Statistics

Waterman, D. (1989). A Guide to Expert Systems. Addison-Wesley, ISBN 9780201083132