NEUROFUZZY MODELING OF NATURAL FREQUENCIES OF CYLINDRICAL SHELLS APPLIED TO EVOLUTIONARY BASED OPTIMAL DESIGN OF SR MOTORS

Abstract

A thorough analysis of cylindrical shells' dynamical behavior is essential in many different industrial design problems, and particularly in electric motor design. Shell vibration equations form a set of partial differential equations of order eight, where their closed form solution is only known for few special cases with a few known boundary conditions along with many not necessarily realistic assumptions. On the other hand, finite element based numerical solutions does not yield a lumped model that can be regarded as a general solution for natural frequencies of cylindrical shells. In this paper, a neurofuzzy model for natural frequencies of cylindrical shells is developed. At first, natural frequencies are calculated for a wide range of cylindrical shells' dimensions, using either closed form solution or finite element method. Gathered data is exploited for training of a Locally Linear Neurofuzzy Network, which yields a general model for calculation of natural frequencies of cylindrical shells. While the developed neurofuzzy model may be used in different design problems that deals with cylindrical shells, as a case study, the proposed model along with an evolutionary algorithm are utilized in the optimal design of a Switched Reluctance motor.