Permanent magnet linear generator for the application as range extender in full electric vehicles

In the thesis a numerical optimization method is utilized to find the most appropriate generator design for the application of the free piston linear generator. The developed analytical model is employed here to study the Pareto front of the fractional slot linear machine for the maximization of the efficiency and the minimization of the active mass subjected to the end-winding temperature with 6 different pole and slot numbers, which are expected to show the high force constant by the given design boundary. One of them is chosen for more detailed study, including the minimization of the end effect force. This work proposes the Bayesian optimization approach to deal with a constrained bi-objective optimization for the machine design problem. The algorithm used employs Kriging to build surrogate models in relations of high fidelity with the actual function. The methodology to normalize an objective function is introduced to maintain the diversity of the solutions, and three different acquisition criteria are implemented to approximate the Pareto front of a constrained bi-objective problem in an effort to find the optimal solutions within few number of sampling. The developed algorithm integrates the single objective evolutionary algorithm. However, it is still computationally more efficient when compared to employing them directly without the surrogate models, as their objective functions are relatively not expansive to compute.