<p>This paper proposes the level-up dynamic area sampling (LEDAS) method to alleviate the computational burden in multimodal optimization. Finite element analysis (FEA) is essential for evaluating the electrical and magnetic properties of devices such as the interior permanent magnet synchronous motor (IPMSM) in motor design. However, in multimodal optimization problems, such as in the optimal design of electric motors, the time-consuming characteristic of FEA can greatly reduce the efficiency of the design process. By utilizing a kriging surrogate model, LEDAS employs an efficient sampling strategy to reduce computation time while identifying optimal solutions. The method combines kriging interpolation with maximin-latin hypercube sampling (Maximin-LHS) and space filling method to ensure uniform sample distribution and improve model accuracy, while also integrating an adaptive sampling approach to continuously update the surrogate model. The performance of LEDAS was validated using two test functions, and its superiority was confirmed through comparisons with conventional optimization algorithms. Additionally, rotor shape optimization was performed to minimize torque ripple and line-to-line back electromotive force (B-EMF) total harmonic distortion (THD) while maximizing electromagnetic performance. Finally, stress and demagnetization analyses were conducted to evaluate the structural and thermal stability of the motor.</p>