<p>This study proposes an efficient optimization framework for permanent magnet tubular linear machines (PMTLMs) under spring-loaded conditions. The objective is to achieve high thrust output, maintain force linearity, and ensure stable plunger return. Two-dimensional (2D) and three-dimensional (3D) finite element analysis (FEA) showed strong agreement with experimental results. Therefore, the optimization is primarily conducted using the computationally efficient 2D model. To further reduce computational costs, the 2D FEA is integrated with kriging-based surrogate modeling. The Latin hypercube sampling (LHS) method is employed to generate distributed design points, and correlation analysis identifies key design parameters for sequential two-stage optimization. The optimized design achieves a peak thrust exceeding 462·7N at the maximum stroke while satisfying the spring constant range of 297.72–350·25N/mm. Restoring force margin analysis confirmed that the spring consistently surpassed the no-load thrust across all displacements, securing stable plunger recovery. These results demonstrate that the proposed framework enables accurate, robust, and computationally efficient optimization of PMTLMs, offering a practical solution for high-performance direct-drive linear actuators requiring both strong thrust and reliable spring-loaded operation.</p>