A Framework for Adaptive Width Control of Dense Contour-Parallel Toolpaths in Fused Deposition Modeling

1. Introduction

Fused Deposition Modeling (FDM) is a dominant additive manufacturing technique prized for its versatility and low cost. A critical step in FDM process planning is generating toolpaths to fill the 2D cross-section of each layer. Contour-parallel toolpaths, created by inward offsetting the layer boundary, are preferred for accuracy. However, a fundamental flaw arises when using a uniform bead width (typically the nozzle diameter): if the shape's internal width is not an exact multiple of this bead width, it results in overfill (material overlap causing pressure buildup and bulging) or underfill (gaps leading to reduced stiffness or failed features). This issue is particularly detrimental for parts with thin walls or fine details, common in applications like microstructures, topology-optimized components, and functional prototypes.

This paper presents a comprehensive framework to solve this by generating adaptive-width contour-parallel toolpaths. The core innovation is a method to decide the number of beads and their individual widths to densely fill any polygon without over-/underfill, while critically constraining the width variation to be manufacturable by standard FDM hardware.

2. Methodology & Framework

2.1 Problem Definition & Uniform Offset Limitations

Given a simple polygon representing a layer and a nominal bead width $w_n$, the uniform offset method generates paths at distances $w_n, 2w_n, 3w_n,...$ from the boundary. The filling fails when the remaining unfilled region's width $d_r$ is not equal to $w_n$. If $d_r < w_n$, it causes overfill; if $d_r > w_n$ and cannot fit another bead, it causes underfill. This is illustrated in Figure 1a of the paper, showing clear gaps and overlaps in the center of a rectangular shape.

2.2 Adaptive Width Framework Overview

The proposed framework is agnostic to the specific scheme, structured around a core width decision function. For a shape with a certain fillable diameter $D$, this function determines the number of beads $n$ and their respective widths $\{w_1, w_2, ..., w_n\}$ such that $\sum_{i=1}^{n} w_i = D$, and each $w_i$ is within the printer's feasible range $[w_{min}, w_{max}]$. The framework can integrate different optimization objectives (e.g., minimize width variance, maximize minimum width).

2.3 Novel Scheme: Minimizing Extreme Width Variation

The authors' primary contribution is a novel scheme that prioritizes reducing extreme bead widths (those very close to $w_{min}$ or $w_{max}$) while limiting the number of toolpaths that need to deviate from the nominal width. The logic is that a few moderately adjusted widths are preferable to many severely adjusted ones or one extremely thin/thick bead, as the latter are harder to print reliably. This scheme strategically alters a minimal subset of beads from a baseline uniform-offset plan.

3. Technical Implementation

3.1 Mathematical Formulation & Width Decision Function

The core problem is formulated as an optimization. Let $D$ be the total width to fill. Find integer $n$ and widths $w_i$ that solve:

$$\text{Minimize } f(\{w_i\}) \quad \text{subject to:}$$ $$\sum_{i=1}^{n} w_i = D, \quad w_{min} \le w_i \le w_{max} \quad \forall i$$ where $f$ is an objective function. The novel scheme uses $f$ designed to penalize widths near the boundaries $w_{min}$ and $w_{max}$ more heavily than deviations in the middle of the range, formalized as a piecewise cost function.

3.2 Medial Axis Transform (MAT) Application

For complex polygons, the fillable "width" $D$ is not constant; it varies along the medial axis (the skeleton of the shape). The framework utilizes the Medial Axis Transform (MAT) to decompose the polygon into segments. Along each segment of the MAT, the local width is treated as $D$ for the adaptive width calculation, ensuring the toolpaths conform to the shape's varying geometry. This is crucial for handling branches and non-convex features.

3.3 Back Pressure Compensation Technique

Adaptive width requires real-time control of extrusion flow. The authors develop a back pressure compensation technique for off-the-shelf FDM systems. By modeling the extruder as a fluid dynamic system, they relate commanded flow rate $Q_{cmd}$ to nozzle pressure and, consequently, to final bead width $w$. An inverse model is used to adjust $Q_{cmd}$ for a desired $w$, compensating for hysteresis and pressure buildup effects that cause inaccuracies in non-standard widths.

4. Experimental Validation & Results

4.1 Statistical Analysis on 3D Model Dataset

The framework was tested on a dataset of representative 3D models containing thin walls, small holes, and complex contours. Key metrics analyzed included: Percentage of filled area without over-/underfill, Maximum and minimum bead width generated, and Width variation (max/min ratio).

Results: The novel scheme achieved near-100% fill density (eliminating gaps/overlaps) across all models. Crucially, it reduced the occurrence of beads at the extreme limits ($w_{min}$, $w_{max}$) by over 70% compared to a naive adaptive width method that simply divides $D$ by $n$. The width variation ratio was consistently maintained below a factor of 2.5, within a more manufacturable range.

4.2 Physical Validation & Print Quality Assessment

Physical prints were made using a modified open-source FDM printer implementing the back pressure compensation. Test artifacts included tensile bars with thin gauge sections and models with intricate lattice structures.

Findings: Parts printed with adaptive toolpaths showed:
1. Superior visual quality: No visible bulging in center regions, smooth top surfaces.
2. Improved mechanical properties: Tensile tests on thin sections showed a 15-25% increase in ultimate tensile strength and stiffness compared to parts with uniform toolpaths, directly attributable to the elimination of underfill voids.
3. Reliable feature reproduction: Small holes and narrow bridges were printed completely, whereas uniform toolpaths often failed to close gaps or produced weak, stringy features.

Chart/Figure Description: A key figure (implied as Fig. 5 or similar in the paper) likely presents a bar chart comparing the "fill efficiency" (100% - % area of gaps/overlaps) between Uniform Offset, a Basic Adaptive method, and the proposed Novel Scheme. The Novel Scheme's bar would reach ~99-100%, significantly higher than the others, especially for a category of "Thin Features (< 5mm width)".

5. Analysis Framework & Case Example

Case: Printing a Topology-Optimized Bracket
A common outcome of topology optimization is a organic, thin-walled structure. A uniform 0.4mm toolpath fails in the varying-width members.
Framework Application:
1. Input: Layer polygon of a bracket arm, MAT computed. Local width $D$ varies from 1.1mm to 2.3mm.
2. Width Decision: For $D=1.1mm$, $n=3$ beads. Naive division: $w_i = [0.367, 0.367, 0.367]mm$. One bead is at $w_{min}=0.3mm$, risk of flutter.
3. Novel Scheme: Optimizes for $f$. Solution: $w_i = [0.35, 0.40, 0.35]mm$. All widths are farther from extremes, total $D=1.1mm$ maintained.
4. Output & Print: Toolpaths are generated at offsets calculated using these adaptive widths. Back pressure compensation adjusts flow for each segment. The resulting print has dense, void-free infill in the thin arm, translating to higher load-bearing capacity.

6. Future Applications & Research Directions

• Multi-Material & Functional Grading: Adaptive width control can be coupled with variable material composition. Imagine a toolpath where width and material (e.g., stiff vs. flexible filament) change synchronously along the MAT to create spatially tailored mechanical properties, pushing towards "process-property co-design" as explored in projects like the MIT Center for Bits and Atoms' hyperform work.
• Integration with Slicing Software: The next step is embedding this framework into mainstream slicers (e.g., Ultimaker Cura, PrusaSlicer) as an advanced infill mode, making it accessible to engineers and hobbyists.
• Machine Learning for Width Prediction: A neural network could be trained on simulation data to instantly predict the optimal $\{n, w_i\}$ for any local geometry $D$, bypassing iterative optimization and speeding up slicing for complex parts.
• Beyond FDM: The core principle applies to other AM processes with a deposition toolpath, such as Direct Ink Writing (DIW) for bioprinting or Wire Arc AM (WAAM) for metals, where controlling deposited track geometry is equally critical.

7. References

1. Ding, D., et al. "A tool-path generation strategy for wire and arc additive manufacturing." The International Journal of Advanced Manufacturing Technology (2014).
2. Wang, W., et al. "Manufacturing of advanced topology-optimized structures via additive manufacturing." Science (2021) - Related work on AM for complex structures.
3. Gibson, I., Rosen, D., & Stucker, B. "Additive Manufacturing Technologies: 3D Printing, Rapid Prototyping, and Direct Digital Manufacturing." Springer (2015) - Standard reference for FDM fundamentals.
4. "Medial Axis Transform." In: CGAL User and Reference Manual. CGAL Editorial Board (2023). - Computational geometry basis for MAT.
5. MIT Center for Bits and Atoms. "Hyperform: Computational Design for Digital Fabrication." [Online Project Description]. - Relevant research on co-design.

8. Original Analysis & Expert Commentary