The rainflow amplitude distribution function model is the core component for estimating random vibration fatigue life using the frequency domain method, with double-rectangular spectrum Gaussian random vibration being a common random vibration process,to improve the accuracy of the Rayleigh model's description of the rainflow amplitude distribution for double-rectangular spectrum Gaussian random processes, a sampling statistical analysis evaluation was conducted. For the double-rectangular random vibration spectrum, the power spectral density function was employed to characterize its statistical properties. Time-domain signals were generated using trigonometric series methods, and rain-flow cycle counts were performed on these signals to obtain statistical results for the rain-flow amplitude distribution. The influence of irregularity factors on model prediction outcomes was investigated. A correction factor was proposed based on the normalized moment of the rain-flow amplitude distribution. Prediction results from different models in the high-amplitude region were compared with the statistical outcomes of the rain-flow amplitude distribution. Results demonstrate the new model's simplified form, Convenient calculation ，with an average prediction deviation of 0.0078—positioned between the Dirlik and Tovo-Benasciutti models—representing a significant improvement over the original Rayleigh model.