Abstract：A method of dynamic knowledge representation and reasoning based on Fuzzy Petri Nets is proposed for problems of knowledge of Fuzzy characteristics, where frequent updating of knowledge is required in an expert system. Using the data obtained from the actual environment, the parameters of the knowledge model such as weights, threshold and reliability can be adjusted by training, so the knowledge can be updated dynamically. Some useful factors of knowledge representation based on production rules and Neural Networks are integrated into the knowledge model in a clear way, with well defined meanings for parameters, and with a learning and parallel reasoning ability. The result of simulation shows that the precision of the knowledge model is improved after training.
2007, Vol. 25 
