The accurate prediction of wear depth is of great significance to the failure evaluation of gear transmission. At present, the traditional wear calculation method is still unable to accurately predict the wear depth of helical gears under variable working conditions. The present work proposes a novel wear depth prediction method of helical gear, which is based on the theory of thermodynamics and the degradation entropy generation (DEG) theorem. The method can be used in not only dry contact friction but also mixed elastohydrodynamic lubrication (EHL) condition. The degradation coefficient of helical gear material was determined by wear experiment. The advantages of using DEG theorem to calculate wear loss under variable operation conditions are demonstrated by wear experiment. Moreover, the influence of ignoring the update of tooth surface load on the calculation results in wear process is further studied. The results reveal that the wear depth prediction method of helical gear proposed in this work can obtain the consistent wear depth distribution with the traditional method. The results calculated by the traditional method are generally higher than the method in this paper, and ignoring the load update in the wear process will make the results larger. The prediction method of helical gear wear depth presented in this paper will be useful for accurately predicting the wear of helical gear.