確率付定性モデルのパラメータ表現

抄録

Qualitative reasoning is one of the powerful methods for the fault diagnosis of the complicated systems. The stochastic qualitative reasoning method, one of qualitative reasoning methods, has proposed. However, it is hard to construct an appropriate stochastic qualitative model for a target system because of a lot of stochastic variables in this one.
Toward the construction of a stochastic qualitative model automatically, parameter expression for them are proposed in this paper. All propagation rules and functions are represented with some parameters to construct a model easily. In this method, the variable elements of the model can be reduced extremely.
From application to some air conditioning system models, the effectiveness for the model generation is confirmed. As a result of this, the agreement rate, which which indicates how much the inference by the model agrees with the series of the measurement data, is smoothly changes according to the parameters, and the maximum value is as high as the one for the conventional modeling.