Abstract: In view of the existing low accuracy of dynamic prediction in mining subsidence and in order to improve the prediction accuracy and make the dynamic prediction results play an important role in guiding the production of the mining area and the land use of the goaf,the theoretical model and the corresponding programming algorithm of dynamic prediction is developed for the strike main section,both of which are based on the the probability integral method and optimized segmented Knothe time function. The parameters of the time function of the model have clear meaning and are easy to be obtained without any other monitoring data.Specifically,first,the calculation method of the time function value of each dynamic mining unit is studied and determined at the predicted time; second,the calculation model of the surface dynamic movement and deformation is derived based on the time function value of each unit under different mining speeds ; third,the calculation steps and programming algorithm of the time function and the surface movement are given according to the model.Using the model and algorithm in this paper,the prediction program is compiled and applied to the prediction practice,the prediction practice shows that when the given prediction time is long enough,the prediction of dynamic subsidence is consistent with that of static prediction,and at the inflection point,the corresponding tilt value is maximized and it is completely consistent with the measured results. That is to say,the model can also be used for static prediction after the surface movement is stable. Due to the difference of algorithm,it will take more time than before,but it realizes the calculation assumption of dynamic and static prediction integration. In addition,through the dynamic prediction of mining in No.29401 working face of Guandi mine,and sampling comparison,statistical prediction results and measured data,the relative accuracy of the dynamic prediction of the points on the strike main section is within 6%,which proves the reliability of the model and algorithm.