Military, agricultural, and urban planning have all made extensive use of SAR (Synthetic Aperture Radar) in the realm of remote sensing. SAR images are more capable of identifying the details of the targets than optical images and can be taken in any condition. In the research of target detection and recognition, a large number of SAR image data under different imaging conditions is needed for learning and training to improve recognition rates. However, due to the challenges associated with SAR imaging, the lack of data causes many jobs relying on data-driven deep learning algorithms to perform less than satisfactorily. Obtaining real measured SAR images is difficult, as both the quantity and the conditions for coverage do not meet the requirements. Therefore, establishing a physical model of beam transmission for digital simulation has become an important means of data acquisition. This paper presents a comprehensive review of SAR image simulation methods and introduces the essence of possible development directions of SAR image simulation for machine learning.