Genetic Algorithms for Real Search Space and Their Use for Nonlinear Inverse Problems.

In this paper, a genetic algorithm for continuous search space is proposed and its use for nonlinear inverse problems is further described. The algorithm uses a population of individuals each represented by a real vector. The performance tests of the algorithm were conducted for the optimisation of some different functions with continuous variables, and the results of the tests were compared to the performance of the canonical GAs. The results show that the algorithms optimise the functions more efficiently than the canonical GAs in terms of the time and memory required for computation and the convergence rate. The algorithm was then applied to the parameter identification of a thermal conductivity problem. As a result, the algorithm was able to find a parameter set close to the exact solution even when the measured data were subject to noise.