A method of evaluation product reliability in consideration of abnormal data removal is proposed in accordance with the abnormal data rejection criteria that the abnormal data caused by the test error and the sample dispersion can be removed. Firstly, the paper obtains a product life data based on extrapolating product pseudo life data by accelerating degradation test data, and the optimal distribution of the life time is determined by the distribution goodness of fit test of the lifetime data. And then, the paper estimates the parameters of the model by utilizing the least square estimation method combined with the normally used acceleration model, converts the pseudo life under the high force into a life value under normal stress level, and removes the unsocial life data according to the abnormal data elimination criteria. At last, the life distribution of the product is obtained by distribution fitting of the life data, and the reliability and the reliable life of the product is evaluated. The applicability and the effectiveness of the proposed method are verified by using a certain type of accelerometer as an example.