<p>This study presents an optimal design approach leveraging machine learning to mitigate thrust ripple, which is a primary contributor to vibration and noise in permanent magnet linear synchronous motors (PMLSMs). Thrust ripple originates from detent force, which can be alleviated through the introduction of auxiliary teeth. To maintain output capability while suppressing thrust ripple, optimizing the thrust output is crucial. In this work, the geometric parameters of the auxiliary teeth are defined as design variables, and the objectives are set to simultaneously minimize thrust ripple and maximize thrust. A machine learning–assisted optimization framework is developed, comprising a meta-model and a niching genetic algorithm to efficiently search for optimal solutions in a multi-modal, multi-objective design space. The optimization methodology is validated using two test functions. Based on this methodology, a range of candidate models across diverse regions of the design space was obtained. Subsequently, to prevent performance degradation due to manufacturing uncertainties, a robustness evaluation was conducted to determine the final model. As a result, the average thrust increased by 2.84%, and the thrust ripple was reduced by 8.73%_p. Finally, an irreversible demagnetization analysis is conducted to validate the feasibility of the final design.</p>