摘要： Physics-based thermal simulation of laser powder bed fusion can greatly help understanding the process and reducing time and cost associated with experimental approaches. Numerical methods including the finite element method have been widely used to predict parts thermal-structural history during the build process. However, the simulation has scalability issues because both layer thickness and laser spot diameter are too tiny compared to the part-scale. Therefore, it is only feasible to run these simulations with massive computational resources or if the boundary conditions are simplified, which inevitably reduces the prediction accuracy. This paper proposes a tetrahedral element-based scan-wise adaptive remeshing framework for thermal simulation of the selective laser melting (SLM) process. Results show that a significant reduction of the computational time and resources can be achieved due to the reduction of the number of nodes. The proposed framework enables part-scale thermal modeling of the SLM process with detailed thermal history without sacrificing the accuracy. The effectiveness of the proposed framework is demonstrated through a large size cantilever problem.