Source-linked AI summary
Influence of Extruded Filament Shape on Buildability in 3D Concrete Printing: A Geometry-Informed Deep Learning-FEM Approach
Giacomo Rizzieri, Saif-Ur-Rehman, Jörg F. Unger, Annika Robens-Radermacher
TL;DR
3DCP buildability assessment commonly simplifies deposited filaments as rectangles, although filament morphology can affect structural stability and prediction accuracy. This paper couples ShapeGen3DCP with layer-activation FEM to model realistic geometries from material and process parameters. The study finds that geometry matters especially for free-flow deposition, while elliptical and volume-conserving rectangular representations provide practical modelling choices.
Problem
Most FEM-based buildability models simplify printed layers as rectangles, while the influence of realistic filament geometry on buildability remains insufficiently characterized.
Method
The framework couples ShapeGen3DCP-based filament-shape prediction with layer-activation FEM and mesh generation for geometry-aware buildability assessment.
Results
Filament representation strongly affects buildability predictions for free-flow deposition but less so for layer pressing; the ellipse approximation agrees best overall with reference results.
Takeaways & Limitations
An elliptical representation balances geometric accuracy and modelling effort, while rectangular models can remain useful when their dimensions are selected appropriately for efficient or conservative simulations.
Takeaways & Limitations
The simulations do not provide a representative comparison of computational efficiency because mesh resolution was held constant and selected for curved representations.
Abstract
from arXiv · showhide
The geometric morphology of deposited filaments can significantly influence the structural performance and stability of 3D concrete-printed (3DCP) structures. However, most finite element (FEM)-based approaches for buildability assessment represent printed layers as simplified rectangles, potentially limiting predictive accuracy. This study proposes a geometry-informed modelling framework that integrates the deep-learning-based filament shape prediction tool ShapeGen3DCP with a layer-activation FEM approach to investigate the effect of realistic filament geometries on buildability. The framework generates geometry-aware numerical models directly from material and process parameters, eliminating the need for experimental filament characterization or computationally intensive fluid-flow simulations. Validation against experimental data and a parametric study of rectilinear walls demonstrate that extrusion parameters and the resulting filament geometry can significantly influence buildability predictions. Realistic filament representations are particularly important for free-flow deposition, whereas layer-pressing strategies are less sensitive to geometric simplifications. Among the investigated representations, an elliptical approximation provides an effective balance between geometric fidelity and modelling simplicity. When rectangular representations are preferred to enable regular computational meshes for faster simulations, defining their dimensions based on volume conservation improves prediction reliability compared with calibrating them using either the maximum filament width or the interlayer contact width. Overall, the proposed methodology demonstrates the importance of incorporating filament geometry into 3DCP simulations and provides practical guidance for selecting efficient and accurate geometric representations for buildability assessment.
1 Introduction
3DCP buildability models often simplify deposited filaments as rectangles, despite evidence that filament morphology affects structural behavior and that its influence on buildability remains insufficiently quantified. The paper therefore motivates a geometry-informed FEM framework that predicts filament shapes from material and process parameters without relying on extensive experiments or costly fluid simulations.
- Motivation: Filament morphology affects interlayer bonding, hardened-material performance, durability, and potentially the buildability of printed structures.Buildability is defined as the maximum number of layers that can be stacked before structural collapse.
- Limitations of existing models: Most solid-based FEM models represent extruded filaments as idealized rectangles, which may substantially oversimplify circular-nozzle deposition.Rectangular dimensions are commonly derived from nozzle size or measured filament width.
- Limitations of existing models: Experimental filament characterization is time-consuming under complex conditions, while high-fidelity fluid models are generally too computationally demanding for simulating complete printing processes.These constraints limit the direct use of detailed shape information in predictive buildability frameworks.
- Related approaches: Existing data-driven approaches predict filament dimensions or shapes from printing and material information, including regression, support-vector, and deep-learning methods.ShapeGen3DCP predicts layer shapes using rheological properties and key process parameters for circular-nozzle printing.
- Proposed approach: The paper proposes coupling ShapeGen3DCP with layer-activation FEM to generate geometry-aware buildability models from material properties and printing parameters.The framework is designed to examine geometry effects, process and material dependencies, and the geometric fidelity needed for accurate yet efficient predictions.
- Proposed approach: The workflow combines predicted filament geometries, toolpath-based mesh generation, and elastic or elasto-plastic FEM buildability assessment, with validation against experiments and a multi-condition parametric study.The study considers six printing strategies, five geometric discretizations, and three material models; layer-pressing regimes are reported as less sensitive to geometric simplification.
2 Workflow description
The workflow links machine-learning filament-shape prediction, geometry-aware mesh generation, and layer-activation FEM to assess buildability. It derives filament representations from material and process inputs, then simulates sequential deposition with evolving fresh-concrete properties and failure tracking.
- The automated workflow sequentially predicts filament shapes, generates a layered computational mesh, and evaluates buildability with progressively activated finite elements.ShapeGen3DCP supplies geometries, Gmsh constructs the domain, and the FEM framework simulates deposition and failure.
- ShapeGen3DCP estimates single- and double-layer cross-sections from fluid-state material properties and process parameters without computationally expensive multiphysics simulations.Inputs include density, yield stress, plastic viscosity, nozzle diameter and height, printing velocity, and flow velocity.
- The mesh phase converts predicted profiles into mechanically meaningful rectangular, elliptical, or point-based cross-sections and stacks them layer by layer in three dimensions.Profiles may be analytically generated or derived from processed ShapeGen3DCP points before extrusion and tetrahedral meshing.
- Buildability is assessed with quasi-static elastoplastic FEM, progressive layer activation, and a pseudo-density field that scales stiffness and body-force contributions.The formulation captures deformation accumulation and failure as the structure evolves during fabrication.
- Material parameters vary linearly with simulation time to represent the early-age gain in stiffness and strength during printing.The parameters are calibrated from compression tests, and the linear approximation is adopted for the minute-scale dormant phase.
- Each layer’s load is ramped over its activation time, approximating continuous deposition while layers are activated simultaneously in the simulation.The activation time is tl = ll/vp, where ll is layer length and vp is printing velocity.
- Different filament representations can change total structural volume and applied load; only the points and rect. adapted representations preserve volume.The study discusses density adjustment as a way to equalize total weight across representations.
3 Results
The study first validates the proposed workflow against experiments, then examines how filament shape and discretization affect rectilinear-wall buildability across material behaviours.
- The framework is evaluated against published experiments to assess buildability-prediction performance.
- A parametric study investigates filament shape and geometrical discretization for rectilinear walls.
- The analysis covers purely elastic and various elasto-plastic material regimes.
- Buildability is additionally assessed for cylindrical structures.
3.1 Validation of geometry-informed approach
The geometry-informed framework is assessed against experiments using calibrated material models, ShapeGen3DCP-predicted filament geometries, and layer-activation FEM simulations. Simulations agree well with experiments overall, while buildability predictions vary with filament representation and density treatment.
- Validation setup: The framework was assessed against Tripathi et al.’s experiments because their printing parameters, material properties, compositions, and wall-buildability data were available.The comparison used simple wall geometries and multiple material compositions.
- Validation setup: Material parameters for LM 30 and LSM 30 were calibrated by reproducing reported compression stress-strain curves at two ages with numerical simulations.The model used an elastoplastic material law with nonlinear hardening.
- Validation setup: For the reported process conditions, ShapeGen3DCP predicted LM 30 dimensions of 30.3 mm width, 22.13 mm contact length, and 11.7 mm height, versus 27.63 mm, 19.22 mm, and 12.21 mm for LSM 30.The predictions used a 20 mm nozzle, 15 mm nozzle height, 15 mm/s printing velocity, 17.92 mm/s flow velocity, and 2112.5 kg/m3 density.
- Validation results: The simulations showed good agreement with experimentally observed numbers of printable layers, supporting the framework’s ability to reproduce reported buildability values.Activated layers at failure were compared for simulations with and without density adjustment.
- Validation results: Filament representation affected predictions most for the stiffer LSM 30 material: ellipse and points agreed closely, while rect. width overestimated and rect. contact underestimated buildability.The rect. adapted representation produced an intermediate failure level.
- Validation results: Density adjustment had negligible influence for LM 30 and changed non-volume-preserving LSM 30 representations by fewer than three layers.It does not affect the volume-preserving points and rect. adapted representations; without adjustment, rect. width predicts lower buildability because of increased cross-sectional volume.
- Validation scope: The comparison is not a complete validation, but it demonstrates agreement with observed values and shows that filament cross-section assumptions matter for reliable buildability prediction.This validation supports using the framework for the subsequent comparison of filament representations.
3.2 Numerical investigation on the influence of filament shape on buildability: rectilinear walls
Rectilinear-wall simulations show that filament geometry affects buildability predictions, especially for rounded free-flow deposits, while ellipse and point representations closely agree.
- Filament geometries: ShapeGen3DCP predicted flatter, wider filaments for layer pressing and rounder, taller filaments for free-flow deposition.Layer pressing corresponds to high aspect ratios and roundness ratios near unity; free-flow corresponds to aspect ratios near unity and larger roundness ratios.
- Filament geometries: Cross-sectional area decreases as speed ratio increases within each fixed-nozzle-height case, consistent with mass conservation.
- Elastic buckling: Point-based and elliptical representations produce nearly identical buildability predictions, whereas contact rectangles underestimate and width-based rectangles overestimate buildable layers.The adapted rectangle also overestimates buildability, but less than the width-based rectangle.
- Elastic buckling: Representation effects are weaker for pronounced layer pressing and stronger for free-flow scenarios with more rounded filament geometries.
- Material stiffness: Lower Young’s modulus reduces buildability, while contact rectangles remain lower and width-based and adapted rectangles remain higher than the ellipse representation.The ellipse lies between the rectangle-based approaches, although its analytical comparison is unavailable for rectangular-only formulas.
- Elasto-plastic buckling: Plasticity reduces absolute buildability but does not change the fundamental influence of filament representation; LSM 30 permits more layers than LM 30.In ID2, LSM 30 fails within the 30th layer, while LM 30 fails within the 18th layer; the elastic case reaches 36 layers.
- Comparative effects: Relative deviations are similar across elastic and elasto-plastic materials: width-based and adapted rectangles overestimate by up to approximately 80%, while contact rectangles underestimate by up to about 30%.The results suggest that representation effects are governed primarily by geometry rather than material behavior within the density-adjusted setup.
- Geometric slenderness: Increasing layer width increases wall height at failure across material models and representations, while plastic yielding lowers buildability relative to purely elastic behavior.
3.3 Numerical investigation on the influence of filament shape on buildability: cylinders
The framework was extended to cylindrical structures with radii of 50 and 150 mm using rounded filaments and a softer elasto-plastic material. The same representation trends appeared, but cylinder radius changed the governing failure mechanism.
- Setup: Cylindrical simulations assessed radii of 50 and 150 mm using rounded ID3 filaments and the softer elasto-plastic LM 30 material.The analysis was qualitative and was not validated against experimental data.
- Representation choice: The ellipse representation was used as the realistic reference because it closely matched the point representation in rectilinear-wall simulations.
- Results: The cylindrical results reproduced the rectilinear-wall trends across filament representations.
- Results: Table 6 compares activated layers at failure for the two cylinder radii and the different filament representations.
- Failure mechanisms: The 50 mm-radius cylinder failed by global buckling, whereas the 150 mm-radius cylinder exhibited pronounced upper-region bulging.Different failure mechanisms influenced the relative over- and underestimation rates compared with the ellipse representation.
4 Discussion: results summary and simulation guidelines
Filament cross-sectional geometry substantially affects buildability and the reliability of numerical predictions, especially for rounded free-flow deposits. Elliptical representations offer a strong accuracy–simplicity compromise, while volume-conserving rectangular models improve reliability when regular meshes are preferred.
- Geometric effects: Filament cross-sectional shape strongly influences structural buildability through differences in width, contact area, bending stiffness, and buckling resistance.Layer pressing produces wider, flatter filaments with higher bending stiffness, whereas free-flow deposition produces rounded filaments with reduced stiffness and earlier instability.
- Geometric effects: Buildability predictions are particularly sensitive to geometric representation for free-flow deposition, while flatter layer-pressed filaments are less sensitive to rectangular approximations.Rounded free-flow geometries require more faithful representation than the flatter cross-sections produced by layer pressing.
- Error estimation: The roundness ratio w1/wc correlates strongly with approximation error and supports a design chart for estimating buildability-prediction errors in rectilinear walls.The chart relates expected prediction error to roundness ratio using linear regression trends fitted to numerical results.
- Practical guidance: The design chart offers practical guidance for selecting representations, but computational-efficiency comparisons remain inconclusive because all simulations used the same element size.Rectangle-based models may still reduce time and memory through coarser regular meshes, while further correction factors are needed for highly curved structures.
- Representation guidelines: Ellipse approximations provide results close to the reference geometry while requiring a simpler description, making them the best compromise among the investigated representations.A point-wise filament representation is generally unnecessary for accurate buildability predictions.
- Representation guidelines: Rectangular contact discretization gives a conservative lower bound, whereas rectangular width discretization overestimates buildability and requires caution in safety-critical assessments.The adapted rectangular discretization also tends to overestimate buildability, but its error is smaller and its volume conservation preserves total weight without density adjustment.
5 Conclusion
The study presents a geometry-informed workflow that couples ShapeGen3DCP with layer-activation FEM to improve buildability modelling without preliminary filament characterization or fluid-flow simulation. Results show that geometry sensitivity depends on printing strategy, while elliptical and volume-conserving rectangular representations provide practical modelling choices.
- Method: The workflow integrates ShapeGen3DCP with a layer-activation FEM model to automatically generate geometry-aware buildability simulations from material and process parameters.This avoids preliminary filament characterization and computationally expensive fluid-dynamics simulations.
- Findings: Rounder free-flow filaments are substantially more sensitive to geometric simplifications than flatter filaments produced by layer pressing.The influence of filament geometry also depends on the printed structure’s governing failure mechanism.
- Guidelines: Elliptical geometries provide the best compromise between prediction accuracy and modelling simplicity, while volume-conserving rectangular dimensions improve reliability when regular meshes are preferred.More complex realistic polylines yield only marginal accuracy improvements while adding substantial geometric complexity.
- Implications: Coupling machine-learning-based geometry prediction with existing FEM workflows provides a practical route to geometrically informed buildability models without excessive computational increase.The approach explicitly integrates extrusion process parameters and resulting filament geometry into structural simulations.
- Implications: The workflow and design chart help users select filament representations according to the required balance between prediction accuracy and computational cost.The chart estimates uncertainty from filament roundness ratio.