Experimental and machine learning assessment of sea shell powder as a partial cement replacement

Résumé fourni par la source

The high level of carbon intensity in the cement industry is a spur for the use of low-impact mineral alternatives in structural concrete. This study assesses the use of Sea Shell Powder (SSP) as a partial cement replacement by 10%, 20%, 30% and investigates the mechanical performance, carbon footprint, and economic feasibility of sea-shell powder-modified concrete while evaluating the applicability of machine-learning (ML) models for predicting compressive strength and supporting sustainable mix design decisions. Experimental results reveal that 10–20% SSP replacement has structural-grade behaviour (28-day compressive strengths of 45.0 MPa and 41.9 MPa with 10 and 20% SSP replacement, respectively, compared to compressive strength of 51.2 MPa for control), while a substantial decrease (24.6 MPa) occurs for 30% SSP. Flexural and tensile strength showed a similar trend which further confirmed that the performance deterioration becomes critical after 20% replacement. Cradle to gate carbon analysis showed proportionate reductions in embodied carbon dioxide, with 10%, 18% and 26% CO 2 reductions for 10%, 20% and 30% SSP levels respectively. Cost analysis revealed a savings parallel of 6.7%, 13.3%, and 20.0% of per batch because of less cement consumption. ML models Random Forest and Gradient Boosting had a high accuracy in predicting and determining the dominant predictors of strength to be SSP content and curing age. An ML based optimization map showed the existence of an optimal window of sustainability performance at 10–15% SSP, where the strength, carbon and cost metrics converge favourably. Overall, the results support the fact that 10–20% SSP offers the best balance of mechanical reliability, carbon mitigation and cost efficiency. The integrated experimental-computational framework paves the way for establishing a decision ready pathway for the selection of SSP modified concretes to fulfil structural needs while greatly reducing environmental and economic burdens.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.