An Improved Light Beam Search Method in Multiobjective Inverse Problem Optimizations

An improved light beam search method is proposed to find a series of preferred subsets of the entire Pareto optimal solutions of a multiobjective inverse problem. The main improvements include: 1) the utopia plane (line) is proposed to be divided into subdomains, and an evaluation mechanism is introduced to identify the topology relationship between each individual and the preference zone to improve the distribution of the searched Pareto solutions and 2) a multiexternal archive methodology is designed in order to use different archives in the mating pool evolutionary process for different searching stages to guarantee the diversity of the population to prevent premature convergences. Mathematical test functions and a benchmark inverse problem, TEAM Workshop Problem 22, are solved to testify the effectiveness and the efficiency of the proposed method.