Research On Swarm-based Stochastic Optimization Algorithms And Its Application In Electromagnetic Inverse Problems

The history for electromagnetic inverse problem studies exceeds more than one century.However,only till late 20th century are there substantial efforts to develop a general framework in solving electromagnetic inverse problems.Indeed,pratical electromagnetic inverse problems are all solved by using numerical methods.The booming advancement of electronic computer and technologies allignes the accelerated the application of various new numerical methods into this research field.Since the general methdology for solving inverse problems is to convet a inverse problem into a direct one,and to solve it iteratively using a optimizer,optimization algorithm constitues one major issue in inverse problem studies.Population-based algorithms have become a trend over the last two decades and been accepted as an indispensible tool in the arsonal of optimizations.In this dissertation,we systemically analyse the development of fast global stochastic search algorithms and based on which develop two engineering-oriented improved variants of the algorithms.The focus of this work can be summarised as follows:Fisrtly,we introduce two classes of population-based fast global stochastic search algorithms,particle swarm optimization(PSO)and quantum-behaved particle swarm optimization(QPSO),from the perspective of their origin,on-going development and relations and differences.Next,we describe our improved variant of QPSOs.With the goal of balancing exploration and exploitation searches,our improvements consist of new adaptive dynamic parameter control stategies,dunamic neighborhood topoly,additional stop criteria and a partially regenerated swarm scheme.Finally,the effect and effectiveness of our improvments are validated by solving a benchmark design optimization problem:TEAM Workshop problem 22.