Derivation of determining variables for floor impact sound insulation performance in apartment buildings using machine learning

Ji-Hoon Park; Won-Hak Lee; Chan-Hoon Haan

doi:10.7776/ASK.2025.44.6.620

All Issue

2025 Vol.44, Issue 6 Preview Page Next Page

Research Article

Derivation of determining variables for floor impact sound insulation performance in apartment buildings using machine learning 머신러닝 기반 바닥충격음 차단성능 결정변수 도출

30 November 2025. pp. 620-629

PDF XML

Abstract

The impact sound insulation performance of apartment floors is affected by structural design, construction quality, and site conditions. This study quantitatively analyzed prediction performance according to input data composition for both design and construction stages, using machine learning regression models trained on building information and field measurement data from 415 apartment units. Four algorithms—linear regression, decision tree, support vector machine, and random forest—were trained and evaluated for explanatory power and accuracy using the coefficient of determination (R2). Variable importance analysis quantified the relative contribution of each factor. When field measurement data were excluded, all models showed lower accuracy; including measured data improved results, with the random forest model performing best. Because the single-number rating for impact sound is derived from frequency-specific measured levels, accuracy differed by input composition. Without measured data, major contributors were architectural factors such as resilient material shape, floor-finishing type, and thickness, whereas with measured data, frequency-band sound pressure levels dominated. In heavyweight impact sound prediction, some low-frequency bands (50 Hz – 250 Hz) from lightweight impact sound data were also included, indicating correlation between the two impact types. Based on these results, this study compared the predictive limitations in the design stage with the improved prediction performance in the construction stage, and verified the applicability of data-driven prediction methods.

Keywords

Floor impact sound

Machine learning

Random forest

Feature importance

Performance prediction

공동주택의 바닥충격음 차단성능은 구조적·설계적 요인뿐 아니라 시공 품질과 현장 조건 등 다양한 변수의 영향을 받는다. 본 연구는 국내 공동주택 415세대의 건축 정보와 현장 측정 데이터를 기반으로, 머신러닝 회귀모델을 적용하여 설계단계와 시공단계에서 활용 가능한 데이터 구성에 따른 예측 성능의 차이를 정량적으로 분석하였다. 예측모델은 선형회귀, 결정트리, 서포트벡터머신, 랜덤포레스트의 네 가지 기법으로 학습되었으며, 결정계수(R²)를 통해 모델의 설명력과 예측 정확도를 평가하고, 변수 중요도 분석을 통해 각 요인의 상대적 기여도를 산정하였다. 측정데이터가 제외된 경우 모든 모델의 예측 성능은 낮은 수준을 보였으나, 포함된 경우 전반적으로 향상되었으며 랜덤포레스트가 가장 높은 설명력을 나타냈다. 이는 단일수치평가량이 주파수별 측정데이터를 기반으로 산출되기 때문에, 입력데이터 구성의 차이에 따라 예측 성능이 달라진 결과로 해석된다. 변수 중요도 분석에서는 측정데이터가 없는 조건에서 완충재 형상, 마감재 종류 및 두께 등 건축적 요인의 기여도가 높았으며, 측정데이터가 포함된 조건에서는 주파수별 음압레벨이 지배적인 변수로 나타났다. 또한 중량충격음 예측 시 일부 경량충격음의 저주파수 대역(50 Hz ~ 250 Hz) 데이터가 낮은 수준이지만 변수로 포함되어, 두 충격음 간 상관성이 확인되었다. 이러한 결과를 통해 설계단계의 예측 한계와 시공단계에서의 예측 성능을 비교·검토하였으며, 데이터 기반 접근의 예측 방법을 확인하였다.

키워드

바닥충격음

머신러닝

랜덤 포레스트

변수중요도

성능예측

References

Statistics Korea, “2024 Population and housing census,” Statistics Korea, Sejong, Rep., 2024.

Korea Environment Corporation, “Noise complaints related to floor impact sound,” Korea Environment Corporation, Ansan, Rep., 2025.

Korea Environment Corporation, “Analysis report on floor impact noise complaints,” Korea Environment Corporation, Ansan, 2025.

J. M. Kim, S. Y. Park, and J. H. Lee, “Analysis of residential environmental factors affecting sensitivity to floor impact noise” (in Korean), J. Archit. Inst. Korea 38, 87-96 (2022).

S. H. Lee, “Effect of improved insulation performance on perception of indoor noise in apartment buildings” (in Korean), J. Build. Environ. Res. 22, 25-32 (2023).

The Board of Audit and Inspection of Korea “Operation status of apartment floor impact noise reduction system,” Audit. Rep., 2019.

Ministry of Land, Infrastructure and Transport, “Operational guidelines for the post-verification system of floor impact sound in apartment buildings,” MOLIT, Sejong, 2022.

Ministry of Land, Infrastructure and Transport, “Proposed amendment to the housing Act (Bill No. 12345),” MOLIT, Sejong, 2024.

K. W. Kim, G. C. Jeong, K. S. Yang, and J. Y. Sohn, “Correlation between dynamic stiffness of resilient materials and heavyweight impact sound reduction level,” Build. Environ. 44, 1589-1600 (2009).

10.1016/j.buildenv.2008.10.005

A. Schiavi, “Improvement of impact sound insulation: A constitutive model for floating floors,” Appl. Acoust, 129, 64-71 (2018).

10.1016/j.apacoust.2017.07.013

J. O. Yoen, H. M. Goo, and S. C. Lee, “The dynamic stiffness of resilient materials for floor impact sound according to temperature change” (in Korean), J. Acous. Soc. Kr. 37, 338-342 (2018).

W. H. Lee, Analysis of the influence factors on the floor impact sound of apartments through field experiments, (Master’s thesis, Chungbuk National University, Cheongju, Korea, 2020).

T. Cho, “Experimental and numerical analysis of floating floor resonance and its effect on impact sound transmission,” J. Sound Vib. 332, 6552-6561 (2013).

10.1016/j.jsv.2013.08.011

H. K. Shin, S. H. Park, and K. W. Kim, “Spatial distribution of heavy-weight floor impact sounds according to the household location in a box-frame concrete building,” App. Acoust. 203, 109214 (2023).

10.1016/j.apacoust.2023.109214

C. Y. Yun, C. G. Cho, and M. J. Kim, “Comparison of heavy-weight impact sound pressure level according to slab thickness and floor area” (in Korean), J. Korean Soc. Living Environ. Sys. 22, 768-776 (2015).

10.21086/ksles.2015.10.22.5.768

S. S. Hong, S. M. Lee, and C. H. Haan, “Floor structure for the reduction of floor impact noise of apartment with concrete slab less than 210 mm” (in Korean), J. Acous. Soc. Kr. 43, 724-736 (2024).

M. Bader Eddin, H. Kinns, and R. Wimmer, “Prediction of airborne and impact sound insulation in timber floors using artificial neural networks,” Acoustics 4, 13-25 (2022).

10.3390/acoustics4010013

N. R Draper and H. Smith, Applied Regression Analysis, 3rd ed (Wiley, New York, 1998), pp.1-706.

L. Breiman, J. H. Friedman, R. A. Olshen, and C. J. Stone, Classification and Regression Trees (Wadsworth, Belmont, CA, 1984), pp. 1-368.

C. Cortes, and V. Vapnik, “Support-vector networks,” Mach. Learn. 20, 273-297 (1995).

10.1007/BF00994018

L. Breiman, “Random forests,” Mach. Learn. 45, 5-32 (2001).

10.1023/A:1010933404324

KS F 2810-1:2015, Acoustics – Field Measurement of Impact Sound Insulation of Floors – Part 1: Method using Standard Light Impact Source, Korean Agency for Technology and Standards, 2015.

KS F 2810-2:2012, Acoustics– Field Measurement of Impact Sound Insulation of Buildings – Part 2: Method using Standard Heavy Impact Sources, Korean Agency for Technology and Standards, 2012.

KS F 2863-1:2017, Rating of Floor Impact Sound Insulation for Iimpact Source in Buildings and of Building Elements – Part 1: Floor Impact Sound Insulation Against Standard Light Impact Source, Korean Agency for Technology and Standards, 2017.

KS F 2863-2:2017, Rating of Floor Impact Sound Insulation for Impact Source in Buildings and of Building Elements – Part 2: FloorIimpact Sound Insulation Against Standard Heavy Impact Source, Korean Agency for Technology and Standards, 2017.

M. Kuhn and K. Johnson, Applied Predictive Modeling (Springer, New York, 2013), pp. 1-628.

10.1007/978-1-4614-6849-3_1

T. Chai and R. R. Draxler, “Root mean square error (RMSE) or mean absolute error (MAE)? – Arguments against avoiding RMSE in the literature,” Geosci. Model Dev. 7, 1247-1250 (2014).

10.5194/gmd-7-1247-2014

Information

Publisher :The Acoustical Society of Korea
Publisher(Ko) :한국음향학회
Journal Title :The Journal of the Acoustical Society of Korea
Journal Title(Ko) :한국음향학회지
Volume : 44
No :6
Pages :620-629
Received Date : 2025-06-05
Revised Date : 2025-08-11
Accepted Date : 2025-10-17
DOI :https://doi.org/10.7776/ASK.2025.44.6.620

The Journal of the Acoustical Society of KoreaISSN:1225-4428(Print) 2287-3775(Online)한국음향학회

All Issue