Resolution improvement of noise sources in high-speed trains by inverse problem optimization

The delay-and-sum beamforming method for noise sources identification is useful for identifying noise sources in a high-speed train. However, the disadvantage of the method is the poor resolution at low-frequency ranges, so the low-frequency noise sources of the high-speed train were not identified. Therefore, the resolution improvement research was conducted for identifying low-frequency noise sources in the high-speed train. At first, relations between actual noise sources and beamforming outputs were derived by using the point spread functions. By the deconvolution approach of the relations, obtaining actual noise sources were transferred to an inverse problem. Then, in order to solve the ill-conditioned inverse problem, an optimization problem with non-negativity constraints was applied and an iteration method was conducted to solve the problem. After algorithm development of resolution improvement, an application on a high-speed train was conducted. The improved noise map obtained from this research provided detailed noise sources of the high-speed train in low-frequency ranges.