Abstract: Data expansion is one of the important contents of small sample reliability research at present. Bootstrap method is widely used in engineering practice because it only needs actual observation data and does not need prior information. When the traditional bootstrap method is used to expand the samples, the value range of the new samples is only limited to the original small sample data range, which has become a disadvantage in the reliability assessment of the NC turret. In order to solve this problem, the traditional bootstrap method is improved to expand the sample range in two directions, and a random number that may be greater than 1 is randomly selected when calculating new data. In order to verify the improvement effect, based on a large sample of NC tool holder failure data, several groups of small sample data are obtained by sampling. The reliability assessment of small sample data is carried out by using the method before and after the improvement. The results show that the improved method is better than the traditional method in adaptability and accuracy. Therefore, the results of data expansion of the improved bootstrap method in reliability assessment can be closer to the ontology.