Abstract: According to measured wheel wear data， the nonstationary Gamma process was used to establish the degradation model of locomotive wheel rim， and the repair time of 95% reliability was predicted by using the two methods a and b combined with wheel rim wear threshold. Method a uses the Bootstrap method to randomly generate a set of empirical distributions of pseudo-life， with Weibull distribution fitting forecasting repair time of 473900km. Method b uses the secondary fourth-order moment method based on the maximum entropy to predict the repair time of 488900km. The results showed that the lifetime of method a is more conservative than that of the empirical method， and the failure distribution of method b is more consistent with that of the experience distribution， and the repair time is 450000km.

Key words: nonstationary Gamma process, wear, degradation model, secondary fourth-order moment method, Weibull distribution