Abstract:

In order to overcome the limitation of dynamic stress spectrum estimation using parameter estimation method, a kind of non-parameter estimation method which is kernel density estimation method is employed. In order to validate the validity of the method, five hundred random points are selected from standard normal distribution function to estimate by using of kernel density method. The result shows that the estimated probability density curve and the real probability curve are well fitted. At the same time, it also shows that the length of group gap has no effect on estimation result. Then the dynamic stress data from a high-speed train is selected to do the kernel density estimation and deduction. The result shows that the trend between damage curve of kernel density estimation and actual test spectrum are the same and each grade’s damage of the former one is a little bigger than the latter one. This means the estimation result of kernel density method is conservative which leads to lower fatigue life of dynamic stress spectrum estimation by using of kernel density method. The conservative estimation is very important to ensure the security of the structure in the application.

Key words: distribution estimation, fatigue life, kernel density, dynamic stress spectrum