Load Spectra Extrapolation by Bandwidth-Optimized Kernel Density Estimation Based on DBSCAN Algorithm

Load spectra extrapolation is the basis of fatigue analysis and life prediction in engineering. This paper extrapolates the loads based on the kernel density estimation method, and a new bandwidth calculation method of kernel density estimation is proposed by introducing the DBSCAN algorithm. This bandwidth calculation method has high accuracy and efficiency. Torque load time history in the brake stabilization phase was first collected through the data acquisition system. Then, the mean and amplitude distributions of loads were obtained by means of the rain flow counting method. The DBSCAN algorithm was used to cluster the load data and divide them into different clusters. Data in the same cluster share the same bandwidth, and the rule of thumb is used for calculation. Experiments were conducted with fixed bandwidth, adaptive bandwidth, and the bandwidth calculation method proposed in this article, respectively. The statistical load amplitude extremes were extrapolated with 100 folds by three different methods, respectively, and 100 braking trials were conducted as a comparison. The results showed that the method of this article has much higher accuracy than the method based on the rule of thumb, meanwhile, it has similar accuracy and lower algorithm complexity compared with the adaptive method. The research content in this paper has a certain guiding significance for the extrapolation of load spectrum in practical engineering.