Fatigue life assessment method of bogie frame with time-domain extrapolation for dynamic stress based on extreme value theory

Abstract

Structural fatigue reliability assessment is one of the research directions for railway vehicles' mechanical problems. Obtaining the representative stress spectrum is the basis of fatigue life assessment, and its adequacy and integrity will directly affect the accuracy of the assessment results. However, because of the limited conditions, many dynamic stress tests are completed in the short-term, which makes the fatigue evaluation results based on short-term test data challenging to reflect the structure's anti-fatigue status under the long-term service. Therefore, it is necessary to extrapolate the stress spectrum from small samples to large samples while retaining real data probability distribution characteristics. In this paper, a fatigue life assessment method with time-domain extrapolation for dynamic stress based on extreme value theory is proposed for the first time. Take a test point near the weld of the rotary arm positioning seat of a railway vehicle's bogie frame as a case study. Firstly, three time-series signals of dynamic stress of the test point in the initial, middle, and final stages of a whole wheel reprofile period are selected. The pre-processing and equal-weight linear superposition are carried out to them. Secondly, the tail probability distribution of the time-series signal after linear superposition is estimated by applying the POT Pareto distribution model, and its confidence level is 99%. Finally, the dynamic stress spectrum under the long-term service is obtained by several-multiple extrapolation of the superimposed time-series signals with the fitting probability distribution function, and the fatigue life of the test point based on the extrapolated dynamic stress spectrum is assessed. Compared with the commonly-used linear extrapolation method, the cumulative fatigue damage with the time-domain extrapolation method based on extreme value theory increased by 0.168%, and the safe operation mileage decreased by 2670 km. Therefore, indicate that using the fatigue analysis results under short-term load (i.e., linear extrapolation) to replace the fatigue analysis results under long-term load will cause the apparent error, which is indeed worthy of attention. And the fatigue life assessment method based on time-domain extrapolation is safer. Provide the method reference and theoretical support for fatigue life assessment under the long-term service.