Optimum Design Of Switched Reluctance Motor Based On Kriging Model

The inverse problem of electromagnetic field is to solve the optimum design of electromagnetic device in order to satisfy the requested performance or parameters. The global optimization methods to solve the inverse problem of electromagnetic field include generally genetic algorithm, simulated annealing algorithm, particle swarm optimization algorithm, ant colony algorithm, and so on. However, these random optimization algorithms have some disadvantages such as large calculation size, long computation time and low rate of convergence. To achieve the optimal design of Switched Reluctance Motor (SRM), an efficient adaptive global optimization algorithm is proposed in this thesis. The proposed algorithm is different from random optimization method, in which the numerical analysis of electromagnetic fields is carried out in each iteration, and from response surface model with fixed parameterized polynomial. In this algorithm, the semi-parameterization construction technique is used to approximate the response model to find the optimal design, and high occupancy of computer memory and long CPU computation time is overcame. The detailed works are as follows:Firstly, based on comparison of current optimal technology of electromagnetic device, the ordinary Kriging model is developed and some regular types of variogram in Kriging model are studied. The spherical model parametric estimation method based on genetic algorithm is proposed, and the corresponding computer program is complied by Visual C++ and Matlab software. The validity of Kring algorithm is verified with analytical functions, and the computing accuracy of thin elastic plates model (TEPM) and the spherical model employed in Kriging model is discussed.Secondly, to improve efficiency of optimization algorithm, a global optimization strategy employing multiple iterations and gradual refinement is developed, in which the Kriging algorithm with spherical model is used. In the optimization strategy, the uniform sampling technique is used to obtain some sampling points in the design space, and Kriging with spherical model is used as interpolation function to approximate the objective function, then genetic algorithm is to find the minimum of this objective function. In each iteration, new sampling points are gradually inserted to approximate the objective function, so that the efficiency and simulation accuracy is improved. The proposed optimization method has been verified using analytic function tests. The results show that the proposed algorithm has the property of fast global searching.Finally, an optimal stator pole face and pole shoe shape attached to the lateral face of the rotor pole are designed for reducing torque ripple of 3-phase 6/4 switched reluctance motor (SRM), and the proposed algorithm is used for optimal design. The calculation results show that the torque ripple is greatly reduced, and testify the high computation efficiency of the proposed algorithm.