PurposeDue to the limitation of experimental conditions and budget, fatigue data of mechanical components are often scarce in practical engineering, which leads to low reliability of fatigue data and reduces the accuracy of fatigue life prediction. Therefore, this study aims to expand the available fatigue data and verify its reliability, enabling the achievement of life prediction analysis at different stress levels.Design/methodology/approachFirst, the principle of fatigue life probability percentiles consistency and the perturbation optimization technique is used to realize the equivalent conversion of small samples fatigue life test data at different stress levels. Meanwhile, checking failure model by fitting the goodness of fit test and proposing a Monte Carlo method based on the data distribution characteristics and a numerical simulation strategy of directional sampling is used to extend equivalent data. Furthermore, the relationship between effective stress and characteristic life is analyzed using a combination of the Weibull distribution and the Stromeyer equation. An iterative sequence is established to obtain predicted life.FindingsThe TC4–DT titanium alloy is selected to assess the accuracy and reliability of the proposed method and the results show that predicted life obtained with the proposed method is within the double dispersion band, indicating high accuracy.Originality/valueThe purpose of this study is to provide a reference for the expansion of small sample fatigue test data, verification of data reliability and prediction of fatigue life data. In addition, the proposed method provides a theoretical basis for engineering applications.