US 12,296,869 B2
Rail corrugation recognition method and apparatus based on support vector machine, device, and medium
Zefeng Wen, Chengdu (CN); Xiaolong Liu, Chengdu (CN); Xinbiao Xiao, Chengdu (CN); and Shulin Liang, Chengdu (CN)
Assigned to Southwest Jiaotong University, Chengdu (CN)
Filed by Southwest Jiaotong University, Chengdu (CN)
Filed on Nov. 4, 2022, as Appl. No. 18/052,794.
Claims priority of application No. 202210464990.9 (CN), filed on Apr. 29, 2022.
Prior Publication US 2023/0347948 A1, Nov. 2, 2023
Prior Publication US 2024/0182089 A9, Jun. 6, 2024
Int. Cl. B61L 23/04 (2006.01); B61K 9/10 (2006.01); G01N 29/265 (2006.01); G06F 18/214 (2023.01); G06N 20/10 (2019.01)
CPC B61L 23/045 (2013.01) [G06F 18/214 (2023.01); G06N 20/10 (2019.01); B61K 9/10 (2013.01); G01N 29/265 (2013.01); G01N 2291/044 (2013.01); G01N 2291/2623 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A rail corrugation recognition method based on a support vector machine, wherein the method comprises:
obtaining wheel-rail noise signals in different time periods, and obtaining wheel-rail noise time domain information according to the wheel-rail noise signals;
dividing the wheel-rail noise time domain information into segmented wheel-rail noise time domain information corresponding to each of the different time periods, wherein each piece of segmented wheel-rail noise time domain information corresponds to an equal length of a moving path of an urban rail transit vehicle;
preprocessing each piece of segmented wheel-rail noise time domain information, and extracting a time domain statistical characteristic quantity and frequency domain eigenmode energy of each piece of segmented wheel-rail noise time domain information;
obtaining a multi-dimensional wheel-rail noise characteristic vector according to the time domain statistical characteristic quantity and the frequency domain eigenmode energy;
constructing a rail corrugation state recognition model based on a support vector machine according to the multi-dimensional wheel-rail noise characteristic vector, and training the rail corrugation state recognition model based on a support vector machine; and
recognizing to-be-recognized wheel-rail noise data by using the rail corrugation state recognition model based on a support vector machine, to obtain a rail corrugation state;
wherein the preprocessing each piece of segmented wheel-rail noise time domain information, and extracting a time domain statistical characteristic quantity and frequency domain eigenmode energy of each piece of segmented wheel-rail noise time domain information specifically comprises:
preprocessing each piece of segmented wheel-rail noise time domain information, removing abnormal data, and extracting the time domain statistical characteristic quantity of each piece of segmented wheel-rail noise time domain information; and
performing variational mode decomposition on each piece of segmented wheel-rail noise time domain information by using a variational mode decomposition method, and extracting each decomposition eigenmode coefficient to convert into frequency domain eigenmode energy of different frequency bands.