Pareto-driven gradient-boosted Bayesian optimization for multi-objective inverse design of Aerogel-cement-EPS panels in small-sample regimes

This study addresses the challenges of escaping local optima in traditional engineering design methods and data scarcity during research and development. It proposes a novel multi-objective optimization approach integrating data augmentation, data expansion, gradient-boosted Bayesian optimization, a...

Full description

Saved in:
Bibliographic Details
Main Authors: Lu Lu, Anming Bai, Haodong Wang, Wenjia Xi, Shan Yun, Shangbing Gao, Mingming Wang, Xinguo Sun
Format: Article
Language:English
Published: Elsevier 2025-12-01
Series:Case Studies in Construction Materials
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2214509525009441
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849405219808477184
author Lu Lu
Anming Bai
Haodong Wang
Wenjia Xi
Shan Yun
Shangbing Gao
Mingming Wang
Xinguo Sun
author_facet Lu Lu
Anming Bai
Haodong Wang
Wenjia Xi
Shan Yun
Shangbing Gao
Mingming Wang
Xinguo Sun
author_sort Lu Lu
collection DOAJ
description This study addresses the challenges of escaping local optima in traditional engineering design methods and data scarcity during research and development. It proposes a novel multi-objective optimization approach integrating data augmentation, data expansion, gradient-boosted Bayesian optimization, and Pareto front analysis. By constructing a Pareto front solution set, this method successfully resolves the multi-objective optimization and inverse design challenges in the formulation design of aerogel-cement-expanded polystyrene building insulation panels under small-sample conditions. The integration of data augmentation and data expansion techniques significantly enhances the scale and diversity of the dataset, effectively mitigating the data scarcity issue. The gradient-boosted Bayesian optimization model rapidly establishes complex mappings between material composition and performance, enabling more efficient global optimum search. It elevates the prediction accuracy of small-sample learning to 0.99. Furthermore, the introduction of a Pareto front exploration strategy within the Bayesian optimizer balances the exploitation of known information and the exploration of unknown design space. This strategy overcomes the limitation of traditional Bayesian optimization, which tends to converge towards a single-objective optimum. Experimental results demonstrate that, while ensuring other performance metrics comply with standard requirements, the thermal conductivity was successfully reduced to 0.0374 W/(m·K), representing an approximately 12 % reduction compared to similar materials. At a thermal conductivity of 0.0397 W/(m·K), the material maintains optimal states in compressive strength, volumetric water absorption, and total thermal resistance. Simultaneously, the aerogel dosage is only 6.8 kg/m³ , demonstrating the best balance between performance and cost. This achievement provides a novel design strategy for large-scale industrial production.
format Article
id doaj-art-bf1272bd23594b77857c420604c93fe1
institution Kabale University
issn 2214-5095
language English
publishDate 2025-12-01
publisher Elsevier
record_format Article
series Case Studies in Construction Materials
spelling doaj-art-bf1272bd23594b77857c420604c93fe12025-08-20T03:36:44ZengElsevierCase Studies in Construction Materials2214-50952025-12-0123e0514610.1016/j.cscm.2025.e05146Pareto-driven gradient-boosted Bayesian optimization for multi-objective inverse design of Aerogel-cement-EPS panels in small-sample regimesLu Lu0Anming Bai1Haodong Wang2Wenjia Xi3Shan Yun4Shangbing Gao5Mingming Wang6Xinguo Sun7Jiangsu Provincial Engineering Laboratory for Advanced Materials of Salt Chemical Industry, Huaiyin Institute of Technology, Huaian, Jiangsu 223003, China; Jiangsu Smart Factory Engineering Research Center, Huaiyin Institute of Technology, Huaian 223003, China; Corresponding author at: Jiangsu Provincial Engineering Laboratory for Advanced Materials of Salt Chemical Industry, Huaiyin Institute of Technology, Huaian, Jiangsu 223003, China.Faculty of Computer and Software Engineering, Huaiyin Institute of Technology, Huaian 223201, ChinaJiangsu Provincial Engineering Laboratory for Advanced Materials of Salt Chemical Industry, Huaiyin Institute of Technology, Huaian, Jiangsu 223003, China; Jiangsu Smart Factory Engineering Research Center, Huaiyin Institute of Technology, Huaian 223003, ChinaJiangsu Smart Factory Engineering Research Center, Huaiyin Institute of Technology, Huaian 223003, ChinaJiangsu Provincial Engineering Laboratory for Advanced Materials of Salt Chemical Industry, Huaiyin Institute of Technology, Huaian, Jiangsu 223003, ChinaFaculty of Computer and Software Engineering, Huaiyin Institute of Technology, Huaian 223201, ChinaCollege of Water Resources and Environmental Engineering, Zhejiang University of Water Resources and Electric Power, Hangzhou 310018, ChinaJiangsu Smart Factory Engineering Research Center, Huaiyin Institute of Technology, Huaian 223003, ChinaThis study addresses the challenges of escaping local optima in traditional engineering design methods and data scarcity during research and development. It proposes a novel multi-objective optimization approach integrating data augmentation, data expansion, gradient-boosted Bayesian optimization, and Pareto front analysis. By constructing a Pareto front solution set, this method successfully resolves the multi-objective optimization and inverse design challenges in the formulation design of aerogel-cement-expanded polystyrene building insulation panels under small-sample conditions. The integration of data augmentation and data expansion techniques significantly enhances the scale and diversity of the dataset, effectively mitigating the data scarcity issue. The gradient-boosted Bayesian optimization model rapidly establishes complex mappings between material composition and performance, enabling more efficient global optimum search. It elevates the prediction accuracy of small-sample learning to 0.99. Furthermore, the introduction of a Pareto front exploration strategy within the Bayesian optimizer balances the exploitation of known information and the exploration of unknown design space. This strategy overcomes the limitation of traditional Bayesian optimization, which tends to converge towards a single-objective optimum. Experimental results demonstrate that, while ensuring other performance metrics comply with standard requirements, the thermal conductivity was successfully reduced to 0.0374 W/(m·K), representing an approximately 12 % reduction compared to similar materials. At a thermal conductivity of 0.0397 W/(m·K), the material maintains optimal states in compressive strength, volumetric water absorption, and total thermal resistance. Simultaneously, the aerogel dosage is only 6.8 kg/m³ , demonstrating the best balance between performance and cost. This achievement provides a novel design strategy for large-scale industrial production.http://www.sciencedirect.com/science/article/pii/S2214509525009441Aerogel-cement-EPS compositesBayesian optimizationSmall-sample learningMulti-objective optimizationInverse design
spellingShingle Lu Lu
Anming Bai
Haodong Wang
Wenjia Xi
Shan Yun
Shangbing Gao
Mingming Wang
Xinguo Sun
Pareto-driven gradient-boosted Bayesian optimization for multi-objective inverse design of Aerogel-cement-EPS panels in small-sample regimes
Case Studies in Construction Materials
Aerogel-cement-EPS composites
Bayesian optimization
Small-sample learning
Multi-objective optimization
Inverse design
title Pareto-driven gradient-boosted Bayesian optimization for multi-objective inverse design of Aerogel-cement-EPS panels in small-sample regimes
title_full Pareto-driven gradient-boosted Bayesian optimization for multi-objective inverse design of Aerogel-cement-EPS panels in small-sample regimes
title_fullStr Pareto-driven gradient-boosted Bayesian optimization for multi-objective inverse design of Aerogel-cement-EPS panels in small-sample regimes
title_full_unstemmed Pareto-driven gradient-boosted Bayesian optimization for multi-objective inverse design of Aerogel-cement-EPS panels in small-sample regimes
title_short Pareto-driven gradient-boosted Bayesian optimization for multi-objective inverse design of Aerogel-cement-EPS panels in small-sample regimes
title_sort pareto driven gradient boosted bayesian optimization for multi objective inverse design of aerogel cement eps panels in small sample regimes
topic Aerogel-cement-EPS composites
Bayesian optimization
Small-sample learning
Multi-objective optimization
Inverse design
url http://www.sciencedirect.com/science/article/pii/S2214509525009441
work_keys_str_mv AT lulu paretodrivengradientboostedbayesianoptimizationformultiobjectiveinversedesignofaerogelcementepspanelsinsmallsampleregimes
AT anmingbai paretodrivengradientboostedbayesianoptimizationformultiobjectiveinversedesignofaerogelcementepspanelsinsmallsampleregimes
AT haodongwang paretodrivengradientboostedbayesianoptimizationformultiobjectiveinversedesignofaerogelcementepspanelsinsmallsampleregimes
AT wenjiaxi paretodrivengradientboostedbayesianoptimizationformultiobjectiveinversedesignofaerogelcementepspanelsinsmallsampleregimes
AT shanyun paretodrivengradientboostedbayesianoptimizationformultiobjectiveinversedesignofaerogelcementepspanelsinsmallsampleregimes
AT shangbinggao paretodrivengradientboostedbayesianoptimizationformultiobjectiveinversedesignofaerogelcementepspanelsinsmallsampleregimes
AT mingmingwang paretodrivengradientboostedbayesianoptimizationformultiobjectiveinversedesignofaerogelcementepspanelsinsmallsampleregimes
AT xinguosun paretodrivengradientboostedbayesianoptimizationformultiobjectiveinversedesignofaerogelcementepspanelsinsmallsampleregimes