ShapeGen3DCP: A Deep Learning Framework for Layer Shape Prediction in 3D Concrete Printing

TL;DR

ShapeGen3DCP develops a fast, physics-informed predictor for 3DCP filament cross-sections by mapping material properties $(\rho, \tau_0, \mu)$ and process parameters $(\phi_n, h_n, v_p, u_f)$ to Fourier descriptor coefficients that define the contour. The training dataset is synthetically generated with a validated PFEM model, and inputs are transformed into dimensionless groups $\tau_0^*$ and $v^*$ to improve generalization. The model achieves strong agreement with both numerical PFEM results and experimental data, with typical feature errors in the $1$–$10\%$ range, enabling pre-calibration, toolpath optimization, and potential digital-twin integration. An open-source web implementation provides practical accessibility for designers and engineers seeking rapid, reliable predictions of layer geometry in 3DCP.

Abstract

This work introduces ShapeGen3DCP, a deep learning framework for fast and accurate prediction of filament cross-sectional geometry in 3D Concrete Printing (3DCP). The method is based on a neural network architecture that takes as input both material properties in the fluid state (density, yield stress, plastic viscosity) and process parameters (nozzle diameter, nozzle height, printing and flow velocities) to directly predict extruded layer shapes. To enhance generalization, some inputs are reformulated into dimensionless parameters that capture underlying physical principles. Predicted geometries are compactly represented using Fourier descriptors, which enforce smooth, closed, and symmetric profiles while reducing the prediction task to a small set of coefficients. The training dataset was synthetically generated using a well-established Particle Finite Element (PFEM) model of 3DCP, overcoming the scarcity of experimental data. Validation against diverse numerical and experimental cases shows strong agreement, confirming the framework's accuracy and reliability. This opens the way to practical uses ranging from pre-calibration of print settings, minimizing or even eliminating trial-and-error adjustments, to toolpath optimization for more advanced designs. Looking ahead, coupling the framework with simulations and sensor feedback could enable closed-loop digital twins for 3DCP, driving real-time process optimization, defect detection, and adaptive control of printing parameters.

Figures (12)

• Figure 1: Schematic representation of ShapeGen3DCP predictive framework. Given input material properties and process parameters, the model predicts the cross-section of the deposited material for one or two printed layers.
• Figure 2: Numerical computation of the 3D filament geometry using the PFEM model, followed by extraction of the 2D midspan cross-section.
• Figure 3: Fourier descriptors approximations of filament cross-sections varying the number of coefficients. As the number of harmonics increases, the reconstructed contours progressively approach the true shape.
• Figure 4: ShapeGen3DCP architecture, consisting of three blocks: an encoder that maps input parameters into a latent space, a processor with $r$ residual layers, and a decoder projecting back to the output space. The predicted Fourier coefficients are then used to reconstruct the closed curve representing the filament cross-sectional profile.
• Figure 5: Training loss (blue) and validation error (orange) during training of the two-layer model. Both metrics are shown in logarithmic scale. The decline and convergence of the curves indicate stable learning and good generalization.