An Interpretable Multimodal Monitoring System for Mixing Process of Printcrete: Integrating Computer Vision, Torque, and Temperature

Authors:
  • RSRuyi Sheng
    Ruyi Sheng
  • YYYong Yuan
    Yong Yuan
  • JZJinyuan Zhang
    Jinyuan Zhang
  • YLYiqi Liu
    Yiqi Liu
  • XYXupeng Yao
    Xupeng Yao
  • YTYaxin Tao
    Yaxin Tao
Preprint
Publisher website
Google Scholar
Abstract
The mixing quality of 3D printable concrete, hereafter referred to as printcrete, is governed by the concurrent attainment of spatial homogeneity and rheological build-up. However, existing monitoring protocols suffer from a fundamental sensing disconnect. Visual inspection is inherently insensitive to bulk stiffening, whereas torque monitoring reflects only volume-averaged resistance and often overlooks localized compositional heterogeneities. To bridge this gap, this study develops an interpretable multimodal monitoring framework specifically designed for the automated mixing of printcrete by integrating computer vision, dynamic torque sensing, and thermal profiling. The proposed system establishes a physics-informed decision logic where the visual texture index serves as a high-sensitivity constraint for particle dispersion entropy, and dynamic torque characterizes the evolution of bulk yield stress. Crucially, thermal profiling acts as a definitive chemical arbiter, utilizing the order-of-magnitude difference in thermal derivatives to autonomously differentiate the distinct thermodynamic signatures across dry-mixing, lubricated liquid dispersion, and accelerator-induced rapid hydration. Experimental validation confirms that this cross-modal fusion effectively resolves the blind spots inherent to single-modal method, identifying the optimal timings for water dosing, admixture injection, and material discharge. By transitioning from static recipe adherence to dynamic performance control, the system ensures the precise rheological consistency required for defect-free extrusion, establishing a foundational architecture for closed-loop, unmanned, high-quality production of printcrete.

This publication has 31 references indexed in Scilit: