To solve the problem that the influence coefficient (c) of the dynamic prediction time of mining subsidence

the subsidence value (Wv) and the time point (??) at the maximum ground subsidence speed are set as fixed values

these parameters were theoretically and experimentally analyzed using segmented Knothe function. Based on the theoretical knowledge of the probability integral method

the functions of these three parameters were deduced by considering time as the independent variable. Then

the segmented Knothe function was optimized

which is showing better generalization performance and prediction accuracy. The experimental results reveal that the optimized Knothe function eliminates the defect that c

? and Wv of the conventional one were set as fixed values

and compensates for the shortage in similar geological conditions were used to generate a fitted empirical formula. After comparing and analyzing the prediction results from the optimized and the conventional models

it was concluded that the predicted results in the optimization model remain within the time range of the measured maximum subsidence velocity. The maximum mean square error and relative mean error of the optimized model are 0.116 m and 3.8%

respectively

which are improved by 65.1% and 51.9% compared with the conventional model. It proves the superiority of this model.