Abstract: At present, the laser welding heat source model is mainly loaded in the form of energy density distributed with geometric regions, and its accuracy is closely related to the geometric region size of energy distribution. However, it lacks a mapping relationship with laser process parameters and physical mechanisms, making it difficult to use for constructing digital twin technology for laser process equipment In response to this issue, this article explores a semi physical and semi intelligent heat source model based on laser welding keyhole heat transfer and energy attenuation Firstly, based on the heat transfer physical mechanism formed by laser welding keyhole, the nonlinear geometric region of laser energy distribution is determined through algorithms; Then, based on the absorption rate attenuation phenomenon along the depth direction of the material with the laser, a knowledge-based nonlinear attenuation curve parameter is defined to express the attenuation law of laser energy By combining the above two methods, a semi physical and semi intelligent heat source model can be established Furthermore, the rationality and accuracy of the heat source model were verified through experiments and simulations of 316L stainless steel laser welding On this basis, the mapping relationship between heat source model parameters and laser power, defocus, and welding speed was explored, laying the foundation for the digital twin implementation of laser welding technology.