Abstract: The determination of the load sample size, as a critical step of compiling load spectrum, directly affects the accuracy of fatigue life estimation. In order to provide an effective method to determine the load sample size for engineering vehicles, the wheel loader was taken as an example and the load characteristics under the loading cycle condition were analyzed. Under the assumption that both the mean of time-history load and the prior distribution of the mean following the normal distribution, Bayesian method and quadratic loss function with weighting were adopted to calculate the sample size. When the posterior loss expectation reached the minimum value, the estimated values of posterior parameters were acquired, which included the prior information, the sample information and the general information. Finally, based on t distribution theory, the load sample size was obtained, which could meet a given confidence level and the error limit.