A double Gamma distribution model to determine the probability density function （PDF） of the time domain rainflow range corresponding to the broadband random stress power spectral density （PSD） is proposed， and a neural network method is used to implement the parameter prediction of the model. A series of stress PSDs are given， and the corresponding stress time his tories are generated using the time-domain randomization method. The number of rainflow-range is counted for the stress time his tories using the rainflow counting method， and the stress rainflow-range probability density values are calculated. Based on the cal culation results of each stress PSD mentioned above， the proposed stress rainflow-range probability density double Gamma distribu tion model is parametrically fitted to obtain a set of corresponding model parameters. The results of the double Gamma distribution model are compared with the Dirlik method and fatigue life prediction is carried out， and the results show that the proposed double Gamma distribution model is more accurate for determining the broadband random stress rainflow-range PDF.