Neurocomputing for Fuzzy Finite Element Analysis of Structures Based on Fuzzy Coefficient Programming

There exist problems in fuzzy finite element methods because technique of solving fuzzy equations is not perfect. For example, computation amount is too big and both sides of the equality are not exactly equal when solutions are substituted into the original equation. The concept of monosource fuzzy number is developed to simplify the calculation process of fuzzy equations. However the source of fuzziness in practical engineering is difficult to be judged and the source of fuzzy coefficient is non-unique. Indeed, no efficient method is available to solve fuzzy finite element equations. In this paper, fuzzy coefficient programming is combined with the essence of elasticity. In other words, the force equilibrium of elastic object is the process of minimizing energy of a quadratic equation. A new fuzzy finite element solution and a new neural network algorithm of fuzzy finite element are developed. The method was proved to be efficient and feasible through circuit simulation.