Abstract: To address the problems of complex operating environments, difficulty in direct monitoring caused by the enclosed structure of the screen surface, and low control accuracy of existing control systems in cross screens, a digital twin-based intelligent monitoring system was developed. The virtual model of the cross screen was divided into four hierarchical components: the feeding device, screen surface, base, and discharge box assembly. Unity3D’s built-in rendering pipeline was adopted for rendering and physical simulation. A rotational speed regulation logic for the screen shafts was established based on a PLC and frequency converter, and an S7 protocol was employed to construct the communication architecture. The K-Nearest Neighbors algorithm was applied to interpolate missing data under abnormal communication conditions. Furthermore, a MEA-BP prediction model integrating a Back Propagation (BP) neural network with a Mind Evolutionary Algorithm (MEA) was established to predict screening efficiency and power consumption. The results show that when the data missing rate was 15%, the KNN algorithm with K = 5 achieved the optimal interpolation performance, with a Mean Squared Error (MSE) of 0.3267 and a Mean Absolute Percentage Error (MAPE) of 32.15%. During both no-load and loaded tests, the variations in operational state data conformed to actual operating patterns, verifying the effectiveness of the monitoring system. Compared with the BP neural network prediction model, the MEA-BP model reduced the MAPE values for screening efficiency and power consumption from 9.15% and 19.46% to 6.97% and 7.87%, respectively, improving prediction accuracy by 2.18 and 11.59 percentage points. The digital twin-driven intelligent monitoring system for cross screens enables real-time monitoring of equipment operating conditions and accurate prediction of screening performance, providing theoretical foundations and technical support for intelligent regulation and optimization of cross screening equipment.