Research On Intelligent Robust Optimization Method Considering Engineering Hybrid Uncertainties

In the practical engineering problems,there are often a lot of uncertainties for optimal design of the electrical equipment,such as geometry,material properties,environmental conditions,etc.These uncertainties make the structural parameters of electrical equipment and nominal values exist some tolerances,which reduces the operating performance of the equipment,and leads to poor robustness.In severe cases,it may cause significant economic losses or social problems.Therefore,it is of great significance to study the robustness optimal design of electrical equipment.Existing robust optimal design theories and methods only address a single probabilistic uncertain variable or an interval uncertain variable,but in the face of these two kinds of uncertain variables of engineering problems,become certain limitations,for the above problems,based on the research at home and abroad,the work completed in this thesis is briefly described as follows.Firstly,for complex engineering problems,a large number of finite element calculations are needed in performance analysis,and the optimal design based on surrogate model is needed to improve optimization efficiency,but there are still many sample points required,low robustness,and sometimes it is difficult to find the optimal solution when combines with optimization algorithms.In this thesis,an adaptive Kriging surrogate model is studied,the method combines the improved composite shape addition strategy with the optimization search process,so that the sample information in the design space can be fully utilized.The proposed method is compared with the maximum expected point addition criterion and the minimum predicted point addition criterion by using classical test examples.Secondly,for engineering problems with both probabilistic and interval uncertainties,a robust optimal design method of three-layer nested optimal model is studied.The first layer model mainly calculates the midpoint and radius of the interval function through the interval algorithm.The second layer model uses random sampling methods to calculate the mean and variance of the midpoint and radius of the interval function.The underlying model focuses on selecting a suitable global optimization algorithm to find the optimal solution in the design space.The test function is selected and compares with the probabilistic robust optimal design method and the non-probabilistic robust optimal design method.In order to further verify validate the performance of the robust optimization algorithm with hybrid uncertainties,the superconducting magnetic energy storage model and the vibration noise of a motor are selected for optimization analysis.Finally,in order to facilitate the use of engineers,the existing experimental design methods,agent models,intelligent optimization algorithms and the theory of the robustness analysis methods studied in this thesis are analyzed to design and develop a robustness optimal design software for electrical products,and the software is analyzed through practical engineering problems.