Road load spectrum preparation method based on density peak machine learning algorithm

The invention provides a road load spectrum preparation method based on a density peak machine learning algorithm. The method comprises the steps: testing the movement of a vehicle on an actual road,and obtaining an initial road load spectrum and speed information; performing statistical modeling on the load and the acceleration to obtain an acceleration signal-road load model; performing preprocessing such as low-pass filtering and resampling on the initial road load spectrum to obtain a preprocessed road load spectrum; based on a wavelet transform method, subjecting the preprocessed load spectrum to compression and noise reduction processing, and shortening the load spectrum time history while effective data is reserved; and performing clustering reconstruction on the compressed load spectrum based on a density peak clustering TNDP machine learning method to obtain an accurate road load spectrum. According to the method, real and accurate road load spectrums can be obtained on different test roads, the load spectrums can be collected only by installing an acceleration sensor in the test of the same type of vehicle model, the cost and the test complexity are greatly reduced, andfinally the compiled accurate load spectrum is obtained. The loaded conditions of automobile parts in the actual driving process can be rapidly reproduced for bench experiments, and high-precision load input is provided for automobile part fatigue reliability research.