Abstract: To solve the problem of deterioration of film thickness uniformity caused by nonlinear and multi-scale dynamic characteristics in the vacuum coating process, and to overcome the insufficient adaptability of traditional control methods in non steady state deposition conditions, an intelligent control method combining multi-scale convolutional neural network (MS-CNN) and fuzzy PID is proposed. Extracting dynamic characteristics of film thickness through MS-CNN, outputting target thickness benchmark, based on fuzzy PID controller, dynamically modulating key vacuum process parameters such as revolution rate and deposition time of sputtering process, transforming the uniformity control problem into real-time and accurate tracking of target thickness value, achieving automated control of film growth process in vacuum environment. The experimental results show that after the application of this method, the system stability time is shortened by 31%−39.5%, the peak uniformity deviation is reduced by 4%−7.4%, and the steady-state uniformity error is reduced by 66.7%−80%. Provided a data-driven and highly adaptive intelligent control solution for high-precision vacuum coating process.