To address the difficulty in calculating modulation-type sound quality parameters—roughness and fluctuation strength—during the acoustic simulation of railway passenger vehicles where only frequency-domain responses are available, this paper proposes a frequency-domain prediction method based on segmented loudness statistical features and BP neural network correction. A genetic algorithm is introduced to optimize the weighting function of the traditional Zwicker sharpness model for better adaptability to railway vehicle scenarios. The prediction model is integrated with statistical energy analysis simulation, and a constrained damping metamaterial floor embedded with acoustic black holes is designed to evaluate sound quality improvement. Results show that the sharpness model optimized by the genetic algorithm agrees well with measured data across all Bark critical bands, with deviations within 5%. The BP neural network reduces the predictionerrors of total roughness and fluctuation strength from 140% and 170% to within 2%, respectively. After applying the metamaterial floor, the overall sound pressure level decreases from 55.9 dBA to 54.7 dBA, and loudness decreases from 12.15 sone to 11.39 sone, a reduction of 6.3%, Sharpness, roughness and fluctuation strength increased slightly instead, resulting in a limited overall improvement in sound quality without introducing obvious auditory deterioration. The proposed frequency-domain prediction method provides an effective engineering tool for sound quality assessment in the acoustic design phase of railway vehicles.

Dr. Li Qi pursued his undergraduate, doctoral studies at the Institute of Acoustics, Tongji University from 1999 to 2009. During his PhD studies, he was awarded the EU Marie-Curie Fellowship and conducted research as a visiting scholar at the MWL Laboratory of the KTH Royal Institute of Technology in Sweden. He founded Yiduo Information Technology (Shanghai) Co., Ltd. in 2015 and established Yiduo Smart Acoustic Valley (Nantong) Technology Co., Ltd. in 2022. He has been selected for Nantong Zilang Talent Program, Nantong Jianghai Talent Program, as well as Jiangsu Provincial "333 High-Level Talent Project".

Yiduo Company is mainly engaged in vibration and noise control, AI technical services as well as software and hardware development. It provides professional services for rail transit, energy and electric power, automotive, aerospace and other industries, and has been honored with the Second Prize of Shanghai Science and Technology Progress Award.