Robust Optimal Techniques For Inverse Problems In Electromagnetics

In the traditional optimal design of inverse problems in electromagnetics, the final target is to search the global optimum of the objective function. However, in practical engineering problems, it is inevitably to suffer from uncertain disturbances, such as ambient noise, manufacturing tolerances. Therefore, under the circumstance of uncertainties, the research for robust optimal design theories and techniques has much more practical engineering importance. It has become a newly growing significant research area of the computational electromagnetics worldwide.This paper studied the robust optimal design theory and methods. After a theoretical analysis of two common methods about robust optimal design, the random probability statistical method and the worst situation analysis method, some robust expected fitness functions were proposed for the analysis and computation for inverse problems in electromagnetics. And some parameter for measuring the robustness performance was designed, for example, the expectation of the objective function in case of disturbance and the deviation factor of the objective function performance. Some random optimization algorithms, which have been used for global optimization, were examined. Based on the characteristics of robust optimizations, the tabu search algorithm was improved. And some new methodologies to decrease the number of function evaluations for robustness performance were proposed. Some different robustness performance evaluation functions were examined, and the estimation for the key parameters in the robustness performance evaluation functions was analyzed.The algorithm codes were programmed under MATLAB 7.1 platform and were successfully applied to the robust analysis and computation of a typical mathematical function. In addition, the SMES (Superconducting Magnetic Energy Storage) coil of the TEAM problem 22 was solved by using ANSYS software. The robust design of the SMES coil configuration was finally realized using ANSYS combined with the powerful mathematical tool MATLAB 7.1 as a mixed programming method. The simulation results confirmed that, compared with the global optimal solution, the robust optimal solution has a much smaller deviation factor of the objective function performance, thus has a much better performance when suffering from disturbance.