Machine learning-based reactivity evaluation of solid wastes and development of a multi-component all-solid waste...

Area of Science:

• Materials Science
• Green Chemistry
• Civil Engineering

Background:

• Alkali-activated cementitious materials offer a low-carbon alternative to traditional concrete.
• Utilizing all-solid-waste in these materials reduces reliance on commercial activators, further lowering environmental impact.
• Current methods for assessing solid waste reactivity lack quantitative models based on microscopic features.

Purpose of the Study:

• To establish a quantitative relationship between the micro-characteristics of solid waste reactivity and the performance of cementitious materials.
• To develop a predictive model for material properties based on microscopic features.
• To support the development of high-performance all-solid-waste cementitious materials.

Main Methods:

• Characterization of 15 solid wastes using XRF, FTIR, XPS, and TG to determine reactivity parameters (silicate tetrahedron polymerization, ionic bond content, oxygen valence).
• Dimensional reduction of reactivity parameters into four principal factors using principal component analysis.
• Support vector regression (SVR) modeling to establish high-precision fitting between compressive strength and principal factors.

Main Results:

• Reactivity parameters quantitatively reflect the role of different solid wastes in the all-solid-waste system.
• Strength development mechanisms are influenced by phase-dependent hydration kinetics and time-dependent constituent contributions.
• The SVR model accurately predicts compressive strength based on micro-characteristics.

Conclusions:

• The proposed reactivity evaluation method provides quantitative support for performance regulation in all-solid-waste cementitious materials.
• This approach is effective for small-sample, nonlinear material property prediction.
• The findings contribute to the development of sustainable and eco-friendly construction materials.