Multimodal machine learning approach for exploring the 28-day compressive strength of nanomaterials-reinforced cement composites

Accurately predicting the 28-day compressive strength (CS) of carbon nanotubes-reinforced cement composites (CNTRCCs) and graphene oxide-reinforced cement composites (GORCCs) is crucial for accelerating their potential application in civil engineering. However, traditional experimental and theoretical modeling methods suffer from problems, including time-consuming, costly, and inefficient. Moreover, it is also challenging to consider the effects of multiple coupling factors. In this work, a multimodal machine learning (ML) approach is proposed as the first attempt to explore the complex relationships between the CS of hybrid system containing both CNTRCCs and GORCCs. The proposed multimodal ML shows great potential in estimating the nanomaterials-reinforced cement composites with a coefficient of determination (R2) of 0.96, surpassing the single-modal ML approaches. The results demonstrate the effectiveness of the developed model in accurately predicting the 28-day CS of hybrid system containing both CNTRCCs and GORCCs. Shapley additive explanations (SHAP) analysis illustrates that the optimal concentration of CNT is approximately 0.5 wt%, and preferred length of CNT and sheet size of GO are within a range of 20–30 μm and below 10 μm, respectively. Additionally, the enhancement effect of a single-layer GO is better than its multilayer counterparts.