Algorithm to Adjust Fuzzy Inference System of Mamdani Type
Authors
Mikhail Sergeevich Golosovskiy
The State Scientific Research Testing Institute of Military Medicine of the Ministry of Defense of the Russian Federation, Saint-Petersburg
Aleksey Valer'evich Bogomolov
State Scientific Center of the Russian Federation – A.I. Burnazyan Federal Medical Biophysical Center, the Russian Federal Medical-Biological Agency, Moscow
Dmitriy Sergeevich Terebov
National Center for Defense Management of the Russian Federation, Moscow
Evgeniy Valerievich Evtushenko
The Black Sea Higher Naval Orders of the Red Star School named after P.S. Nakhimov, Sevastopol'
An algorithm to adjust the Mamdani-type systems is given; it uses the principle of a proportional-integral controller with a limited integral component to adjust fuzzy inference rules. To reduce the adjustment time in comparison with proportional controller and to reduce the overshoot amount in comparison with the proportional-integral controller with the same values of the coefficients of the integral and proportional components, the integral component limitation is used. The advantage of the developed algorithm is in the possibility of performing a local adjustment without a complete set of data for the domain of definition of input variables and the corresponding response values of the system. As a priority area for further research adaption of the application of the algorithms to the membership functions of other types (non-Gaussian) is considered. The efficiency of the algorithm is confirmed by the results of its comparison with algorithms of fuzzy inference system adjustment based on fuzzy neural networks and fuzzy clustering in solving identical problems.
