A Bootstrap Data Expansion Method Based on SVR and Its Application In Small Sample Reliability Evaluation
A Bootstrap data expansion method based on support vector regression(SVR)is proposed to address the reliability evaluation problem of aeroengine with small sample characteristicshe.The SVR model is estab-lished and trained,and the input set is constructed using neighborhood sampling.The augmented samples are obtained by inputting the trained model.The simulation results show that the expanded samples obtained by this method are closer to the true distribution than traditional Bootstrap methods,and effectively expand the range of sample values.Taking the fatigue life test data of a small sample of a certain aircraft engine turbine disk as an example:① Non parametric method,the average fatigue life obtained by the two methods is very close,but the confidence interval obtained by the new method is larger,which is related to the expansion of the sample value range.②The parameter method,the results obtained from parameter estimation of the samples expanded by the new method are closer to the reference values,with a maximum relative deviation of-1.290 2％,while the maximum relative deviation of the traditional method reaches 29.477 6％.The average fatigue life calculated by both methods is closer to the reference value,but the confidence interval obtained by the new method is closer to the reference interval.Overall,the proposed method can effectively achieve sample expan-sion and has certain application value.
万方数据
维普
