Rising industrial carbon dioxide emissions necessitate utilization technologies. Carbon dioxide solidification captures carbon dioxide by reacting with alkaline compounds in concrete, improving its properties. This study integrates a life cycle assessment (LCA) model to evaluate carbon reduction potential with Machine Learning (ML) models to predict complex production dynamics. It investigates solidification mechanisms. Results show co-solidification and external solidification achieve reductions of 676.45 and 704.9 kilograms per tonne, respectively, with notable environmental benefits. A comparison of three predictive models, namely Feedforward Neural Networks (FNN), Polynomial Regression, and Support Vector Regression, confirms that FNN is the optimal choice. It exhibits a lower mean absolute error (791) and a higher coefficient of determination (0.91). SHAP analysis revealed that ‘Coal consumption’ and ‘Electricity consumption’ were the primary drivers of the FNN prediction, confirming the model’s reliance on essential energy inputs, while the ‘date’ feature exerted minimum influence. Projections indicate China’s 2024 concrete production emissions could be 4.23 billion tonnes via synergistic curing, versus 4.37 billion tonnes with conventional external curing. Case and visual analyses further validate carbon dioxide curing’s advantages in improving concrete performance and cutting energy use.

Machine learning model prediction of CO2 production and resource utilization of commercial concrete: curing utilization mechanism, carbon emission assessment and visualisation analysis
    Abstract
    1 Introduction
    2 Theoretical background and methodology
        2.1 Mechanism of CO₂ curing in concrete
        2.2 Key advantages and influencing factors
        2.3 Life cycle evaluation with CO2-cured concrete
        2.4 Machine learning models
    3 Different models in carbon dioxide emission projections and analysis of key indicators
        3.1 Comparison of the performance parameters of each model
        3.2 Comparison of the stability of the models
        3.3 Analysis of the effect of fitting the predicted values to the true values of each model
        3.4 Comparative analysis of residual plots across models
        3.5 FNN model interpretability and key input variable contribution analysis
        3.6 Concrete production and carbon emissions forecast 2024
    4 Process and current status of CO2 co-conditioning of concrete
        4.1 Conservation processes and research status
            4.1.1 Supercritical CO2 cured glass fiber cementitious materials
            4.1.2 CO2 cured ordinary cement concrete
            4.1.3 Carbon dioxide conservation of solid-waste carbon-negative Building materials
    4.2 Problems and control measures
        4.2.1 Problems
        4.2.2 Control measures
    5 Conclusion
    References

• Cong-jun Qi(Chongqing Ruixuan New Building Materials Co., LTD, Chongqing 408000, China)
• An-ping Zuo(School of Materials Science and Engineering, Chongqing Jiaotong University, Chongqing 400074, China)
• Chao-qiang Wang(School of Materials Science and Engineering, Chongqing Jiaotong University, Chongqing 400074, China) Corresponding author
• Yan-yan Liu(School of Materials Science and Engineering, Chongqing Jiaotong University, Chongqing 400074, China)