Enhanced high-level Petri nets with multiple colors for knowledge verification/validation of rule-based expert...

Area of Science:

• Computer Science
• Artificial Intelligence
• Formal Methods

Background:

• Petri net models are used to analyze rule-based expert systems (RBSs).
• Traditional Petri nets have limitations in knowledge verification and validation (KVV) for RBSs, particularly with variables and negative information.
• Existing methods struggle to accurately model complex RBS properties like conservation of facts, refraction, and the closed-world assumption.

Purpose of the Study:

• To propose an enhanced high-level Petri net model for improved KVV of RBSs.
• To accurately model rule inference, including variables, negative information, and key RBS properties.
• To formally investigate the detection of various knowledge flaws within RBSs.

Main Methods:

• Development of an enhanced high-level Petri net model incorporating a coloring scheme.
• Modeling rule inference, conservation of facts, refraction, and the closed-world assumption.
• Formulating KVV tasks as reachability problems and solving equations with multiple colors.

Main Results:

• The proposed model accurately represents variables and negative information in RBSs.
• Key RBS properties (conservation of facts, refraction, closed-world assumption) are effectively modeled.
• The method enables formal detection of redundancy, conflicts, cycles, unnecessary conditions, dead ends, and unreachable goals in RBSs.

Conclusions:

• The enhanced high-level Petri net model offers a more robust approach to KVV for RBSs compared to traditional methods.
• The coloring scheme facilitates the formal detection of logical errors and knowledge flaws.
• This work provides a foundation for more rigorous analysis and validation of complex rule-based expert systems.