Scientific and rational prediction of post-construction settlement is a key link of high-speed railway construction. Based on the field observation data of subgrade settlement of Wuhan-Guangzhou high-speed railway

aiming at measured settlement data being characteristic of small in magnitude

but large relative fluctuation

the suitability of exponential curve method for predicting settlements of subgrade under ballastless track has been studied synthetically. it was found that exponential curve method can't be directly used for predicting subgrade settlements in high-speed railway. The basic idea of three-point method is introduced into exponential curve model

a three-point modified exponential curve method for predicting subgrade settlements has been proposed. Combining the measured settlement data of subgrade in Wuhan-Guangzhou high-speed railway

the characteristics of three-point modified exponential curve model have been analyzed. The analysis shows that selecting three points as forecast sample on settlement-time curve of subgrade can commendably evade the influence brought by data fluctuation; and the settlement regularity after inflection point on settlement-time curve of subgrade more tally with exponential curve

therefore

the samples must be selected after inflection point on settlement-time curve of subgrade; and three samples should reflect the settlement development tendency as far as possible. The prediction results of three point modified exponential curve method are stable with high correlation coefficient. The new prediction method has engineering value.