Integrating life cycle assessment (LCA) and machine learning for sustainable designs: a case study on protective layers made of mineral-bonded fiber-reinforced composites

Abstract Purpose In recent years, machine learning (ML) has become an important tool for predicting material properties and optimizing their mechanical performance without the need for large data sets. However, when design considerations only target structural characteristics, sustainability issues are often overlooked, leading to increased carbon dioxide emissions, energy consumption, and over-reliance on non-renewable materials. This study seeks to bridge this gap by optimizing the design of impact-resistant fiber-reinforced cement-based composites through a sustainability-driven, ML-based modeling approach. Methods In this context, we propose a three-stage integrated framework that combines experimental test databases, sustainability assessment, and ML modeling. The specific design scenario considered here is the use of these materials as protective layers for concrete structures under impact loading. A variety of combinations of fiber and/or textile-reinforced cementitious composites were considered, including single and multiple layers. A life cycle assessment (LCA) was performed in terms of global warming potential (GWP), as this is a sensitive and critical environmental parameter in the concrete industry. The experimental data, consisting of 193 tests on composites subjected to hard impacts, measured the dissipated energy and ballistic limits. Principal component analysis (PCA) was employed to identify patterns and correlations within the experimental data. Based on the database, an ML model was developed to predict energy dissipation, enabling optimization for untested configurations. A multi-objective optimization (MOO) strategy was applied to balance the energy dissipation and GWP constraints, enabling the identification of Pareto-optimal solutions that represent the best trade-offs between mechanical performance and environmental impact. Conclusions The ML model demonstrated high accuracy in predicting GWP and energy dissipation, closely matching experimental data for tested configurations and indicating strong generalization potential for unseen cases. For severe impact applications (above 4 kJ of impact energy), carbon textile grids outperformed other reinforcements, achieving a 20% performance increase with two layers of textile, while maintaining emissions at 5–6 kg CO $$_2$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mmultiscripts> <mml:mrow/> <mml:mn>2</mml:mn> <mml:mrow/> </mml:mmultiscripts> </mml:math> equivalent per plate. Glass emerged as the best alternative for strict GWP limits. Remarkably, hybrid composites, despite their higher cement content in comparison to conventional textile-reinforced concrete (TRC), are the most viable choice due to the superior toughness attained due to short fibers and the sustainability benefits of limestone calcined clay cement (LC 3 ) blended binders. Ultimately, the ML model and LCA framework developed in this study can be extended to other material systems and design scenarios.