Abstract: At the present condition that most of China's plate mills have no gauge-meter, a high-precision self-learning model is therefore developed for the rolling force in plate rolling process without gauge-meter. With the natural logarithm applied to the division of layers of thickness, the division of deformation resistance parameters is carried out according to different layers of thickness to solve the self-learning of rolling force. Then, the method of exponential smoothing is introduced into the model to deal with the deformation resistance parameters of each and every layer of thickness. Based on the high precision spring model, a new approach is proposed that the workpiece thickness after last rolling pass or at exit is set as the target thickness so as to regress the deformation resistance parameters which have been rolled via all scheduled passes. The model has been applied to the process control system of a 3000 mm plate mill in China with favorable result obtained.