Most of recently developed knowledge-based systems perform very well in cases where encoded knowledge plays a role to solve given problems. However, since they do not have models inside, they fail in solving simpler problems than the problems currenlt...
Most of recently developed knowledge-based systems perform very well in cases where encoded knowledge plays a role to solve given problems. However, since they do not have models inside, they fail in solving simpler problems than the problems currenlty solved by the system. There are many cases where currently known equation-based methods are good as models for system, but there are also many cases where quantitatively correct equations which defines a system model cannot be found. The FCM(fuzzy cognitive map) method has been developed for the cases so that models can be defined in fuzzy terms. It turns out that this method does not support time-derivative constraints. In order to solve the problems, we have developed MFCM which allows us to define fuzzy qualitative constraints to build a model for a system and infers the system behavior qualitatively. A set of constraints describing relationship between two or more system variables, includes only M+, M-, ADD, MULTI, and d/dt at the moment, and it is enough to define a simple system. A simulation algorithm has been developed to show derivable system behavior from the model. System variables, their initial values, and a set of constraints are inputs to the algorithm, and simulated system behavior is output. We show the implementation result with a frictionless spring system, and discuss possible problems of this method to be solved in the future and possible applications.