As an important enabling way to upgrade shop-floor digitally and intelligently, the digital twin shop-floor has been paid much attention to the manufacturing. The digital twin models of manufacturing resources were gradually moving from theoretical research to application. However, as a key part of shop-floor, human’s high autonomy and uncertainty led to the difficulty in digital twin modeling for human behavior. Therefore, we collected and proposed the theoretical system for cross-scale human behavior modeling in digital twin shop-floor, powered by the data fusion of macro-behavior and micro-behavior virtual models. Through the closed-loop interaction with physical behavior, can realize monitoring, forecasting and real-time mapping of production activities. Based on this modeling approach, the cross-scale behavior control system was developed and applied in detail in aerospace products assembly shop-floor, and got a good performance.