关键词: 声爆, 室内烦恼度, 声品质, 数据驱动方法, SHAP

Abstract: Aiming at the impact of supersonic aircraft sonic boom on indoor occupants, this paper establishes a data-drivenprediction model of indoor annoyance induced by supersonic aircraft sonic boom. Firstly, based on the sonicboom experimental data (including signal records and subjective evaluation results) of the National Aeronau-tics and Space Administration (NASA), an objective index matrix covering both physical and psychological di-mensions is extracted. Then, the multi -layer perceptron regression model is used to construct the predictionmodel, and the contribution of the objective indices is analyzed in combination with the Shapley additive expla-nations (SHAP) values. The evaluation results demonstrate that the proposed model can effectively capture thenonlinear relationship between sonic boom characteristics and subjective annoyance evaluation, which providestheoretical support for aircraft noise reduction and optimization design. Quantitative results indicate that loudnessand overpressure act as major positive factors, while rise time and clarity index are negative factors. The data-driven method enables accurate and rapid prediction and evaluation of indoor annoyance induced by sonic boom,thereby laying a solid data foundation for the future development of supersonic aircraft.

Key words: sonic boom, indoor annoyance, sound quality, data-driven method, Shapley additive explanations (SHAP)