Traditional train load spectrum statistics are all based on actual measurements,using Weibull distribution to fit test results and calculate load at different levels. But the required testing mileage is long and the entire process is timeconsuming. It is also difficult to fit Weibull parameters. More importantly,the final results are not necessarily reflect the real characteristic of a sample due to the assuming of load distribution beforehand. In order to overcome the disadvantages of traditional load spectrum statistical methods,in this paper,we use kernel density estimation techniques in statistical work of side frame load spectrum for SRM80 full ballast cleaning machine’s T100 C new bogie,which is based on the use of rain-flow counting method in time-load history statistics. Results show that the kernel density estimation method can restore the original load spectrum,remedy the problems of shortage of small sample data and save manpower and resources for train load spectrum statistical work.