Abstract: To address the limitations of existing rolling force models in temper rolling, characterized by insufficient accuracy and oversimplification, a high-precision iterative rolling force prediction model is proposed. Initially, the rolling deformation zone is meticulously divided into three distinct segments: the entry elastic deformation zone, the central plastic defor-mation zone, and the exit elastic deformation zone. Building upon this refined division, a novel approach is introduced to construct three-segment quadratic equations for precisely fitting the work roll flattening curve, ensuring a smooth transition between these regions. Concurrently, based on the refined segmented deformation zones, rolling force distribution models within each region of the strip are established. The model achieves a synergistic solution of the rolling force model and the roll flattening model through an iterative coupling mechanism between the roll flattening curve and the rolling force distribution in the deformation zone. Validation using 6551 sets of production data from a specific temper rolling mill, covering various steel grades (e.g., DQ-IF, 1180DP, 590DP, 440DQ, 980DP, 780DP), after thorough denoising, demon-strates the model’s superior performance. The RMSE (root mean squared error) of the model is 435 kN, its R² (deter-mination coefficient R-squared) is 0.969, and the MAPE (mean absolute percentage error) is 4.6%. Compared to peer models and conventional models, the proposed model exhibits significant advantages in prediction accuracy and demonstrates excellent robustness across various steel grades. The developed temper rolling force prediction model provides a robust and reliable theoretical basis for precise control and process optimization in temper rolling.