Abstract This investigation analyzed the propagation of the uncertainty of the cycle number to fatigue failure, taking into account the strain life method, considering as random variables some physical parameters of the material and load vector in a hole-in-center; plate composed of an aluminum alloy 2024-T351 under fatigue, with variable amplitude loading through multidimensional Hermite polynomials. The application of series of multidimensional Hermite polynomials allowed the prediction of the randomness of the output vector. This research demonstrated that a 2nd degree multidimensional Hermite polynomial adequately estimates the propagation of uncertainty with 45 points per random variable, while the use of Monte Carlo simulation requires 1E + 06 points per random variable. This generates significant computational savings and time. The results showed that variations in the material parameters are not important compared to the variations in the load parameters, which generate an increase in the probability of failure in aluminum alloy parts subjected to fatigue.