音樂廳的設計往往依賴經驗法則或是傳統公式，在計算機技術逐漸成熟的現在，雖然聲學模擬已經漸漸成為主流，然而模擬的技術門檻高，且在設計流程中，聲學模擬往往在設計定案後才執行，難以成為影響音樂廳設計的工具，更像是以經驗法則設計後進行驗證的工具。
本研究結合參數化建模、聲學模擬、機器學習、最佳化演算法，提出機器學習應用於音樂廳設計初期的最佳化演算法。首先，驗證了聲學模擬外掛軟體Pachyderm的適用範圍，以Odeon作為比較對象，當Pachyderm的材料散射係數被設為0.1時，500 Hz RT30、1000 Hz C80、SPL(A)模擬值的RMSE、MAPE均在2 JND以下，並採用此散射係數，以鞋盒型與環繞型音樂廳的參數化模型進行聲學模擬生成數據，結合基於等價吸音面積分布的特徵值，訓練機器學習預測模型，準確度評估以RMSE、MAPE小於2 JND為準確，鞋盒型與環繞型參數化模型在500 Hz RT預測上有87.37%與94.10%的準確度，而1000 Hz C80與SPL(A)D均有95%以上的準確度。除此之外，兩者的數據集可以混合並同時預測兩者的室內聲學品質，雖然準確度因彼此的干擾而稍微下降，其各項預測值的RMSE、MAPE仍均在2 JND以內。最後，將機器學習預測模型應用於最佳化演算法取代聲學模擬，大幅縮短多目的最佳化的運算時間，為了避免機器學習對於完全未知的模型無法預測的情況，提出以對總探索空間隨機抽樣的方式生成訓練集以訓練針對預測對象的機器學習預測模型，採用隨機抽樣與1/10的抽樣比例，並討論了對於聲學目標與複合目標的多目的最佳化，帕雷托前緣在目標函數的表現都優於初始模型，且針對不同目的有不同的最佳解。
與經驗法則和傳統公式相比，機器學習應用於音樂廳設計初期的最佳化演算法更能在設計初期以更快速準確的方式討論設計方案，並在設計初期導入室內聲學品質。

The design of concert halls often relies on thumb rules or traditional equations. While acoustic simulation has become more mainstream with advanced computer technology, its technical complexity makes it challenging to use in early design stages.
This study proposes a machine learning based optimization for early concert hall design. The applicability of the Pachyderm Acoustic simulation plugin was validated against Odeon. With a scattering coefficient of 0.1 in Pachyderm, the RMSE and MAPE for 500 Hz RT30, 1000 Hz C80, and SPL(A) were within 2 JND, leading to its adoption for data generation from parametric models of shoebox and surrounded halls. Machine learning models trained on eigenvalue based on equivalent absorption area distribution achieved prediction accuracies of 87.37% and 94.10% for 500 Hz RT in shoebox and surrounded hall models, respectively, and over 95% for 1000 Hz C80 and SPL(A)D. Combined datasets for both hall types could predict room acoustics with RMSE and MAPE still within 2 JND, despite slight accuracy drops due to mutual interference.
Machine learning models replaced acoustic simulations in optimization algorithms, cutting computation time for multi-objective optimization. To tackle the inability of machine learning models to predict entirely unknown models, a training set was generated via random sampling of the exploration space. Using a 10% sampling ratio, multi-objective optimizations for acoustic and hybrid goals were explored, showing that the Pareto front's performance in objective functions surpassed initial models.

1.Abarghooie, R., Zomorodian, Z. S., Tahsildoost, M., & Shaghaghian, Z. (2021). A Machine-learning Framework for Acoustic Design Assessment in Early Design Stages. arXiv preprint arXiv:2109.06459.https://doi.org/10.48550/arXiv.2109.06459
2.Arau-Puchades, H. (1988). An improved reverberation formula. Acta Acustica united with Acustica, 65(4), 163-180.
3.Aretz, M., & Orlowski, R. (2009). Sound strength and reverberation time in small concert halls. Applied Acoustics, 70(8), 1099-1110.https://doi.org/10.1016/j.apacoust.2009.02.001
4.Ahearn, M. J., Schaeffler, M. J., Celmer, R. D., & Vigeant, M. C. (2009). The just noticeable difference of the clarity index of music, C80. J. Acoust. Soc. Am, 126(4), 2288-2288.https://doi.org/10.1121/1.3249382
5.Badino, E., Manca, R., Shtrepi, L., Calleri, C., & Astolfi, A. (2019). Effect of façade shape and acoustic cladding on reduction of leisure noise levels in a street canyon. Building and Environment, 157, 242-256.https://doi.org/10.1016/j.buildenv.2019.04.039
6.Beranek, L. L. (2016). Concert hall acoustics: Recent findings. The Journal of the Acoustical Society of America, 139(4), 1548-1556.https://doi.org/10.1121/1.4944787
7.Beranek, L. L. (1992). Concert hall acoustics—1992. The Journal of the Acoustical Society of America, 92(1), 1-39.https://doi.org/10.1121/1.404283
8.Beranek, L. L. (2014). Concert hall design: new findings. Proceedings of the Institute of Acoustics, 36(Pt 3).
9.Bianco, M. J., Gerstoft, P., Traer, J., Ozanich, E., Roch, M. A., Gannot, S., & Deledalle, C. A. (2019). Machine learning in acoustics: Theory and applications. The Journal of the Acoustical Society of America, 146(5), 3590-3628.https://doi.org/10.1121/1.5133944
10.Broyles, J. M., & Rusk, Z. T. (2023, May). Predicting the reverberation time of concert halls by use of a random forest regression model. In Proceedings of Meetings on Acoustics (Vol. 51, No. 1). AIP Publishing.https://doi.org/10.1121/2.0001751
11.Cox, T. J., Avis, M. R., & Xiao, L. (2006). Maximum length sequence and Bessel diffusers using active technologies. Journal of sound and vibration, 289(4-5), 807-829.https://doi.org/10.1016/j.jsv.2005.02.019
12.Cox, T. J., & Lam, Y. W. (1994). The performance of realisable quadratic residue diffusers (QRDs). Applied Acoustics, 41(3), 237-246.https://doi.org/10.1016/0003-682X(94)90074-4
13.DISSIMILARITY, I. B. Binaural dissimilarity and optimum ceilings for concert.https://doi.org/10.1121/1.382601
14.Eyring, C. F. (1930). Reverberation time in “dead” rooms. The Journal of the Acoustical Society of America, 1(2A), 217-241.https://doi.org/10.1121/1.1915175
15.Forsyth, M. (1988). Buildings for Music. The American Music Teacher, 37(6), 12.
16.Falcon Perez, R. (2018). Machine-learning-based estimation of room acoustic parameters.
17.Galiana, M., Llinares, C., & Page, Á. (2012). Subjective evaluation of music hall acoustics: Response of expert and non-expert users. Building and Environment, 58, 1-13.https://doi.org/10.1016/j.buildenv.2012.06.008
18.Gao, J., Zhao, Y., & Pan, L. (2024). The Effects of Sound Absorption of Stage House on the Acoustics of Auditorium in an Opera House. Buildings, 14(3), 718.
19.K.B.Ginn.M.Sc. (1978).Architectural Acoustics . Denmark
20.Haan, C., & Fricke, F. R. (1997). An evaluation of the importance of surface diffusivity in concert halls. Applied Acoustics, 51(1), 53-69.https://doi.org/10.1016/S0003-682X(96)00060-6
21.Jeon, J. Y., Kim, Y. S., Lim, H., & Cabrera, D. (2015). Preferred positions for solo, duet, and quartet performers on stage in concert halls: In situ experiment with acoustic measurements. Building and Environment, 93, 267-277.https://doi.org/10.1016/j.buildenv.2015.07.010
22.Jeon, J. Y., Lee, S. C., & Vorländer, M. (2004). Development of scattering surfaces for concert halls. Applied acoustics, 65(4), 341-355.https://doi.org/10.1016/j.apacoust.2003.11.001
23.Jo, H. I., & Jeon, J. Y. (2022). Optimizing sound diffusion in a concert hall using scale-model measurements and simulations. Journal of Building Engineering, 50, 104240.https://doi.org/10.1016/j.jobe.2022.104240
24.Jurkiewicz, Y., Wulfrank, T., & Kahle, E. (2013, June). How far should the geometry of a concert hall be optimized?. In International Symposium on Room Acoustics.
25.Jeon, J. Y., Seo, R., & Jo, H. I. (2020). Effect of stage volume ratio on audience acoustics in concert halls. Sustainability, 12(4), 1370.https://doi.org/10.3390/su12041370
26.Johnson, J. H., Turner, C. W., Zwislocki, J. J., & Margolis, R. H. (1993). Just noticeable differences for intensity and their relation to loudness. The Journal of the Acoustical Society of America, 93(2), 983-991.https://doi.org/10.1121/1.405404
27.Kawase, S. (2013). Factors influencing audience seat selection in a concert hall: A comparison between music majors and nonmusic majors. Journal of Environmental Psychology, 36, 305-315.https://doi.org/10.1016/j.jenvp.2013.08.002
28.Lokki, T., & Pätynen, J. (2019, May). Architectural features that make music bloom in concert halls. In Acoustics (Vol. 1, No. 2, pp. 439-449). MDPI.https://doi.org/10.3390/acoustics1020025
29.Long, M. (2005). Architectural acoustics. Elsevier.
30.Marshall, A. H. (1967). A note on the importance of room cross-section in concert halls. Journal of Sound and Vibration, 5(1), 100-112.https://doi.org/10.1016/0022-460X(67)90181-2
31.Nannariello, J., & Fricke, F. (1999). The prediction of reverberation time using neural network analysis. Applied Acoustics, 58(3), 305-325.https://doi.org/10.1016/S0003-682X(98)00081-4
32.Oldland, K., Teufl, H., & Mahdavi, A. (2022). Computational and empirical assessment of the acoustic performance of repurposed university spaces. In E3S Web of Conferences (Vol. 362, p. 05001). EDP Sciences.https://doi.org/10.1051/e3sconf/202236205001
33.Park, H., & Jeon, J. Y. (2016). Acoustical investigation of ambience sound for auditory environments in music performance spaces. Building and Environment, 107, 10-18.https://doi.org/10.1016/j.buildenv.2016.06.019
34.Peters, B., Nguyen, J., & Omar, R. (2021). Parametric acoustics: design techniques that integrate modelling and simulation. In Proceedings of Euronoise Congress.https://doi.org/10.1177/00375497980700020
35.Palma, M., Sarotto, M., Echenagucia, T. M., Sassone, M., & Astolfi, A. (2014). Sound-strength driven parametric design of an acoustic shell in a free field environment. Building Acoustics, 21(1), 31-41.https://doi.org/10.1260/1351-010X.21.1.3
36.Pan, L., & Zhao, Y. (2022). An experimental investigation into the scattering properties of maximum length sequence diffusers and quadratic residue diffusers. Applied Acoustics, 185, 108347.https://doi.org/10.1016/j.apacoust.2021.108347
37.Pilch, A. (2021). Optimized diffusers for shoe-box shaped performance halls. Applied Acoustics, 178, 108019.https://doi.org/10.1016/j.apacoust.2021.108019
38.Perez, R. F., Götz, G., & Pulkki, V. (2019). Machine-learning-based estimation of reverberation time using room geometry for room effect rendering. In Proceedings of the 23rd International Congress on Acoustics: integrating 4th EAA Euroregio (Vol. 9, p. 13).
39.Peters, B., Nguyen, J., & Omar, R. (2021). Parametric acoustics: design techniques that integrate modelling and simulation. In Proceedings of Euronoise Congress.
40.Rindel, J. H. (1995). Computer simulation techniques for acoustical design of rooms. Acoustics Australia, 23, 81-86.https://doi.org/10.1109/AMS.2017.18
41.Robinson, P. W., Siltanen, S., Lokki, T., & Savioja, L. (2014). Concert hall geometry optimization with parametric modeling tools and wave-based acoustic simulations. Building Acoustics, 21(1), 55-63.https://doi.org/10.1260/1351-010X.21.1.55
42.Rindel, J. H. (2000). The use of computer modeling in room acoustics. Journal of vibroengineering, 3(4), 219-224.
43.Seo, J. G., & Jeon, J. Y. (2019). Impact of early reflection on real-time estimation of direction of arrival in different building environments. Building and Environment, 147, 434-443.https://doi.org/10.1016/j.buildenv.2018.10.029
44.Summers, J. E. (2008). Auralization: Fundamentals of Acoustics, Modelling, Simulation, Algorithms, and Acoustic Virtual Reality.https://doi.org/10.1121/1.2908264
45.Sato, S. I., Otori, K., Takizawa, A., Sakai, H., Ando, Y., & Kawamura, H. (2002). Applying genetic algorithms to the optimum design of a concert hall. Journal of Sound and Vibration, 258(3), 517-526.https://doi.org/10.1006/jsvi.2002.5273
46.Skålevik, M. (2010, August). Reverberation Time–The mother of all room acoustic parameters. In CD Proceedings of 20th International Congress on Acoustic, ICA (Vol. 10).
47.Tsay, Y. S., & Yeh, C. Y. (2020). A machine learning based prediction model for the sound absorption coefficient of micro-expanded metal mesh (Memm). Applied Sciences, 10(21), 7612.https://doi.org/10.3390/app10217612
48.Vlaun, N. J. V. (2015). Sound working environments: Optimizing the acoustic properties of open plan workspaces using parametric models.http://resolver.tudelft.nl/uuid:9ea83381-3a19-4486-ae16-f25a68811b6f
49.Vlaun, N., van Waart, A., Tenpierik, M., & Turrin, M. (2016). A Sound Working Environment: Optimizing the Acoustic Properties of Open Plan Workspaces Using Parametric Models. In SimAUD EU 2016: 7th annual Symposium on Simulation for Architecture and Urban Design (pp. 239-246). simAUD.
50.Wanniarachchi, D. M. Acoustic consideration and musical influences in the architecture of concert halls.
51.Yu, Y. C., & Tsay, Y. S. (2023, November). Optimizing the shapes of concert halls through parametric design. In INTER-NOISE and NOISE-CON Congress and Conference Proceedings (Vol. 268, No. 4, pp. 4683-4688). Institute of Noise Control Engineering.https://doi.org/10.3397/IN_2023_0665
52.Yeh, C. Y., & Tsay, Y. S. (2021). Using machine learning to predict indoor acoustic indicators of multi-functional activity centers. Applied Sciences, 11(12), 5641.https://doi.org/10.3390/app11125641