Fuzzy inference is usually performed with the compositional rule of inference (CRI). To be applied, two operators, an implication function I and a modus ponens generating function M (MPGF) must be previously chosen from a wide range of operators, which is usually done on an empirical or numerical basis. A pair (M,I) has been called a rule model. This paper proposes a new methodology to analyse the semantical behaviour of fuzzy rule models. Two main types of models are analysed, possibility and necessity models, and two main types of possibility distributions are proposed, possible and necessary possibility distributions, that are used to interpret the conclusions obtained from each type of model. The three main semantics proposed by professors Dubois and Prade ( and ) are again identified and enriched taking into account the influence that the MPGF has in the conclusion. This allows to simplify the CRI into a compatibility modification inference (CM, see ) based on two new implication functions with two compatibility indexes, simplifying both the calculus and the interpretation of the CRI, but preserving its main semantical behaviour. The new inference rule is applied to a reasoning strategy based on two complementary reasoning modes leading to a two step inference process. A practical example showing the applicability of the proposed methodology can be found in .
VI Congresso Ibero-Americano de Inteligencia Artificial (IBERAMIA'98). I Concurso Ibero-Americano de Teses e Dissertações em Inteligência Artificial (CTDIA'98), Lisboa, Lisboa (Portugal). 06 octubre 1998
Fecha de publicación: octubre 1998.
J. Villar, M.A. Sanz-Bobi, Aplicación de la teoría de conjuntos borrosos al diagnóstico de procesos industriales, VI Congresso Ibero-Americano de Inteligencia Artificial (IBERAMIA'98). I Concurso Ibero-Americano de Teses e Dissertações em Inteligência Artificial (CTDIA'98). Lisboa, Portugal, 06-09 Octubre 1998