The grinding process of hardened mold steel can result in variations in grinding forces due to changes in machining depth, leading to unstable machining conditions. Therefore, utilizing single-point grinding with large-grain diamond, based on the analysis of the surface morphology characteristics of the machining surface, the dynamic properties and stability of the grinding process are investigated. The aim is to explore the mechanism of the effect of process parameters on machining efficiency and surface quality, with the goal of achieving high-efficiency and high-quality grinding. Firstly, the single-point diamond grinding system is dynamically modeled, and then the grinding vibration signals are measured using an accelerometer for modal analysis of the working system. The natural frequency and damping ratio of the machining system are solved. Next, based on the characterization data of surface waviness and roughness, a digital clustering analysis is performed to correlate the feed depth and wheel speed with the stable machining state. This is matched with the stability lobes diagram of grinding to fit the stiffness of the machining system and the coefficient of grinding force. Thus, a real-time controllable grinding process stability is established by controlling the feed depth and wheel speed. Finally, the grinding experiments on mold steel are conducted to validate and analyze the machining efficiency and quality. The results demonstrate that the modal analysis of the grinding process, along with the clustering matching of the surface morphology characteristics, can effectively map the machining process parameters within the stable domain of the grinding process. Within the stable domain of grinding, using a higher material removal rate can reduce the average surface waviness from 1.203 μm to 0.635 μm, and the average surface roughness from 0.267 μm to 0.143 μm. Moreover, under the same material removal amount, the average surface roughness of stable domain grinding can be reduced by 74% compared to unstable domain grinding. Therefore, by adjusting the feed depth and wheel speed in real-time based on the characterized stable domain of grinding during the machining process, it is possible to simultaneously improve the machining quality and efficiency.