A COMPUTATIONAL METHOD FOR INVERSE PROBLEMS BY USING AUTONOMOUS DECENTRALIZED APPROACH : Proposal of autonomous decentralized finite element method and its application

Abstract

In this paper, we discuss the applicability of life-like approach to inverse problems. Our approach is based on the concept of autonomous decentralization that is the one of characteristics in life or biological system. The approach is developed using the idea of the cell automaton. The finite element method with the relaxation technique is adopted as a solution procedure in this approach. The parameter of the inverse analysis is converted into local rules. The search algorithm for the solutions is a procedure from which the subsystem originally processes information regardless of the entire system, and the information is transmitted only to the adjacent subsystem. When the computation in the subsystem is processed, their rules are used. The inverse problems of three different following types are presented as computational example with our algorithm. 1) Minimum weight problem in 3-dimensional truss structure. 2) Self-organizing problem in 2-dimensional continuous structure. 3) Identification problem for thermal conductivity in 2-dimensional continuous body. The effectiveness and validity of our idea is shown from their numerical results. This computational method is classified into the emergent computation.