A. Macioł

An application of rule-based tool in attributive logic for business rules modeling

This paper presents results of the research being conducted in the Department of Applied Computer Science at Faculty of Management, AGH University of Science and Technology and concerning the use of database server for business rules modeling as knowledge based systems in information systems design and development. Our assumption was that most business rules would have complex structure and their interpretation would require new facts to be generated and analyzed, similarly as it happens in expert systems. Another assumption was that the rules would be stored in tables of relational database. Our experience gained so far indicates that the idea of the compilation of possibilities given by SQL language and interface engine based on if-then type rules together with the ability of new facts generation is right.

Knowledge based system for runtime controlling of multiscale model of ion-exchange solvent extraction

The hereby paper concerns the issue of solution of runtime controlling of multiscale model of ion-exchange solvent extraction. It is based on cooperation of a framework applying Agile Multiscale Modeling Methodology (AM3), and the REBIT Knowledge Based System. Ion-exchange solvent extraction has been shortly introduced. Design assumptions of AM3 and theoretical basis of REBIT have been described. Designed workflows and rules for simple laminar/ turbulent flow and extraction processes have been shown.

Comparison of Classical Multi-Criteria Decision Making Methods with Fuzzy Rule-Based Methods on the Example of Investment Projects Evaluation

In the process of investment decision making, next to financial indicators many other aspects of investment projects are increasingly often considered. This leads to the multi-criteria evaluation of a project. In the work one compared results of multi-criteria evaluation of the investment projects realized by using TOPSIS and AHP methods with results obtained at the use of rule-based methods, especially fuzzy reasoning techniques. To comparisons were used chosen investments in the metallurgical industry. The work finish conclusions defined on the basis carried out calculations.

Hybrid Data in the Multiobjective Evaluation of Investments

Abstract In the process of investment decision making, next to financial indicators many other aspects of investment projects are increasingly often considered. This leads to the multicriteria evaluation of a project. The last decades have shown that the number and complexity of dependencies both inside and outside a company makes it difficult to use the probability theory to represent all kinds of the uncertainty appearing in case of the evaluation of investment projects. Many authors have applied the alternative description of the uncertainty. First of all, fuzzy numbers may be mentioned as an example of the above. This leads to the hybrid description of the uncertainty in the process of the evaluation of the investment. This paper reports an investigation into the design and implementation of hybrid rule-based systems based on the Dempster–Shafer theory of evidence to investment project evaluation. As a result of the investigation, a new methodology will be proposed for multi-criterion project evaluation in the hybrid environment.

Relational Database as an Ontology Framework

In this paper we discuss application of relational data model and attributive logic as an alternative for knowledge modelling tools which are based on descriptive logic. Our experiences indicate that the most effective solution, capable of direct cooperation with majority of industrial information systems, which simultaneously provides decidability, is a combination of relational model with inference system, which utilizes attributive logic. Such solution, named Inference with Queries (IwQ) is realized in accordance with principles of Variable Set Attributive Logic (VSAL). The use of IwQ method for formulation of knowledge requires creation of data metamodel as well as increase in flexibility of inference engine by introduction of so called variable formula mechanism. Such framework is very useful, in the sense that: (i) direct access to information sources stored in relational databases is possible, (ii) well-known and reliable SQL language can be utilized, (iii) introduction of explicit knowledge structure and inference engine control is possible.

Integration of Data and Rules in Inference with Queries Method

The paper presents the effects obtained so far by the author in integration of data and rules in Expert Systems and designing complex information systems. The Inference with Queries (IwQ) method was used. This solution is based on using SQL mechanisms with data stored in a relational database as a knowledge base. Our solution joins the concept of frame-based knowledge representation, replaced partially with a relational database model, with inference possibilities given by procedural languages. Results of the tests show that the idea of joining the possibilities of rule systems and SQL is advisable. The prospects connected with using IwQ method for designing complex information system and their simulation seem to be highly interesting.

Ontology Dedicated to Knowledge-Driven Optimization for ICME Approach

Development of new materials, products and technologies with the ICME approach requires challenging computations, controlled by optimization algorithms. A computational time might be decreased with a “knowledge-driven optimization”—an optimization process is controlled not only by a numerical algorithm, but also by a Knowledge Based System. That requires development of a common language, able to cover communication between numerical models without sophisticated translators. There are several formalisms of knowledge representation, but the most common ones are based on First Order Logic (FOL) and Description Logic (DL). None of them meets all the requirements of knowledge management in ICME processes. We present an approach to development of an environment for knowledge management, combining DL and FOL. An exemplary multiscale problem is described, as well as an OWL2 based ontology and rules controlling an optimization process.

A Multi-agent Approach to Computational Optimization of Metal Forming Processes

In the paper we present a new original approach to a complex multi-phase process optimization problem that relies on dividing the optimization task into partial tasks related to the implementation of individual phases. To implement this concept, we propose a multi-agent approach. Its practical realization is shown on the example of manufacturing process for auto body parts made of the Advanced High Strength Steels (AHSS). Although it is a rather simple process consisting of only two phases, we assumed, however, that the results obtained will allow to extend further research to more complex problems. We present the operating principles of a multi-agent model, the flow of the messages between agents, and the architecture of the system. To ensure the proper speed of the whole system a simple and flexible multi-agent framework called Eve was used to develop a prototype of the system. Research performed, as well as preliminary tests have shown that a multi-agent approach can be successfully applied to reduce complexity of the whole optimization processes. Due to splitting a single, complex optimization process into several, partially independent optimization processes, delegating them to autonomous agents, and application of a knowledge-based reasoning system, significant advantage could be observed over to the solutions described so far in the literature.

Rule-based approach to human-like decision simulating in agent-based modeling and simulation

In the paper we present results of our research related to the application of the method of declarative description of knowledge possessed by people making decisions, characteristic for rule-based systems in ABMS. The purpose of this study is to examine whether, based on knowledge built on declarations of people, it is possible to create an adequate rule-based model of agents. It has been established that only about 60% of respondents take decisions in accordance with the previously declared preferences. Nevertheless, in our opinion, the rule-based approach to the agent behavior modeling can provide good results by introducing a random factor in decision-making rules.