Scholarlink

<p>The acoustic performance of concrete panels was evaluated through a hybrid framework. In this study, we have integrated COMSOL Multiphysics<sup>®</sup> finite element simulations (FEM) and machine learning (ML) models. Four mixes were tested with varying proportions of OPC, barite, and hematite. This targeting improved sound transmission loss (STL) while maintaining structural integrity. Experimental results confirmed that Mix 3 (OPC 75% + 13% barite + 12% hematite) achieved the best balance, with low density (2315.4&#xa0;kg/m³), high stiffness (38.22 GPa), enhanced damping (0.0362), and peak compressive strength (38.1&#xa0;MPa). COMSOL-based STL predictions demonstrated a sharp drop in transmitted sound from Mix 1 (50–70 dB) to Mix 3 (10–30 dB), aligning with ASTM E90 and ISO 10140-2 standards. Deep neural networks (DNN) provided the most accurate surrogate for STL prediction (MSE = 0.009, R² = 0.99, runtime = 26&#xa0;s). While gradient boosting machines (GBM) and extreme learning machines (ELM) showed higher errors. The findings highlight Mix 3 as an optimised acoustic solution, reducing transmitted power by orders of magnitude (up to 1,000,000× in hospitals) and meeting WHO/EPA quiet-environment targets. Overall, the work demonstrates both engineering impact, designing quieter, healthier built environments and modelling reliability, leveraging DNN for efficient STL-driven material design.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid computational and AI/ML framework for predicting sound transmission loss in concrete structures using COMSOL multiphysics

  • 期刊论文

  • 发表日期：

    2025年11月12日

查看全文
  • Sudarshan Sampatrao Bobade,
  • Shailendra Pandurang Banne,
  • Dhanashree Ashish Latthe,
  • Umesh Vishnu Jadhav,
  • Girish Baburav Mahajan

摘要

The acoustic performance of concrete panels was evaluated through a hybrid framework. In this study, we have integrated COMSOL Multiphysics® finite element simulations (FEM) and machine learning (ML) models. Four mixes were tested with varying proportions of OPC, barite, and hematite. This targeting improved sound transmission loss (STL) while maintaining structural integrity. Experimental results confirmed that Mix 3 (OPC 75% + 13% barite + 12% hematite) achieved the best balance, with low density (2315.4 kg/m³), high stiffness (38.22 GPa), enhanced damping (0.0362), and peak compressive strength (38.1 MPa). COMSOL-based STL predictions demonstrated a sharp drop in transmitted sound from Mix 1 (50–70 dB) to Mix 3 (10–30 dB), aligning with ASTM E90 and ISO 10140-2 standards. Deep neural networks (DNN) provided the most accurate surrogate for STL prediction (MSE = 0.009, R² = 0.99, runtime = 26 s). While gradient boosting machines (GBM) and extreme learning machines (ELM) showed higher errors. The findings highlight Mix 3 as an optimised acoustic solution, reducing transmitted power by orders of magnitude (up to 1,000,000× in hospitals) and meeting WHO/EPA quiet-environment targets. Overall, the work demonstrates both engineering impact, designing quieter, healthier built environments and modelling reliability, leveraging DNN for efficient STL-driven material design.