cs.CVMar 13, 2026

BenDFM: A taxonomy and synthetic CAD dataset for manufacturability assessment in sheet metal bending

Authors: Matteo Ballegeer, Dries F. Benoit

Organizations: Ghent University CVAMO Core Lab Research Group Data Analytics

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Predicting the manufacturability of CAD designs early, in terms of both feasibility and required effort, is a key goal of Design for Manufacturing (DFM). Despite advances in deep learning for CAD and its widespread use in manufacturing process selection, learning-based approaches for predicting manufacturability within a specific process remain limited. Two key challenges limit progress: inconsistency across prior work in how manufacturability is defined and consequently in the associated learning targets, and a scarcity of suitable datasets. Existing labels vary significantly: they may reflect intrinsic design constraints or depend on specific manufacturing capabilities (such as available tools), and they range from discrete feasibility checks to continuous complexity measures. Furthermore, industrial datasets typically contain only manufacturable parts, offering little signal for infeasible cases, while existing synthetic datasets focus on simple geometries and subtractive processes. To address these gaps, we propose a taxonomy of manufacturability metrics along the axes of configuration dependence and measurement type, allowing clearer scoping of generalizability and learning objectives. Next, we introduce BenDFM, the first synthetic dataset for manufacturability assessment in sheet metal bending. BenDFM contains 20,000 parts, both manufacturable and unmanufacturable, generated with process-aware bending simulations, providing both folded and unfolded geometries and multiple manufacturability labels across the taxonomy, enabling systematic study of previously unexplored learning-based DFM challenges. We benchmark two state-of-the-art 3D learning architectures on BenDFM, showing that graph-based representations that capture relationships between part surfaces achieve better accuracy, and that predicting metrics that depend on specific manufacturing setups remains more challenging.

Figures & tables

StudyProcess TypeBase DesignDFM Stage
 SubtractiveFormativePrimitiveComplexSelectionIPMA
Zhao and Melkote (2024)✓ ✓ ✓ 
Hussong et al. (2025)✓ ✓ ✓ 
Zheng et al. (2025)✓ ✓ ✓ 
Zhang et al. (2018)✓ ✓ ✓ 
Cao et al. (2020)✓ ✓ ✓ 
Table 1: Overview of introduced synthetic DFM datasets.
Figure 1: Proposed taxonomy for manufacturability metrics. Each quadrant represents a distinct class of manufacturability evaluation relevant for data-driven DFM.
Figure 2: Core steps of the bend creation process.
Figure 3: Bend reliefs, flange types, and symmetry bias enhance the realism and variability of generated designs.
Figure 4: Punch and die geometries are built around the bend center using bend parameters and simple trigonometry.
Figure 5: (a) As the bend progresses, a greater portion of the bend allowance is incorporated into the bend itself (red), while the remainder is distributed as flat flanges (green) on either side. (b) Punch and die tools are dynamically positioned relative to the evolving part geometry throughout the bend trajectory, shown here for 0∘0^{\circ} , 45∘45^{\circ} , and 90∘90^{\circ} angles.
Figure 6: Example of a bent part and its corresponding unfolded flat pattern, accounting for material deformation during bending using bend allowances.
Figure 7: (a) Placing the punch on the left or middle of the bend (depicted in red) results in a collision, offsetting it to the right (depicted in green) avoids this. (b) Adding future bends in their flat state (depicted in orange), reveals a die collision with later flanges during the first bend (in purple).
Figure 8: Example of a part whose unfolded flat pattern exhibits self-intersection, making it impossible to cut from a single sheet.
Figure 9: Illustration of part reorientation metrics: the travel distance between consecutive bends (yellow) and the rotation angle required to align the part for the next bending operation (purple).
 Tool collisionUnfolding overlap
ModelAUCAcc (%)F1 (%)AUCAcc (%)F1 (%)
UV-Net0.840 (0.003)76.07 (0.43)75.41 (0.06)0.896 (0.009)81.80 (1.37)81.31 (1.93)
PointNext0.827 (0.003)73.83 (0.10)71.74 (1.27)0.844 (0.008)76.13 (1.21)76.55 (0.53)
Baseline0.50050.0050.000.50050.0050.00
Table 2: Feasibility classification results. Tool collision detection is evaluated on the BenDFM subset; unfolding overlap detection is evaluated on the BenDFM-U subset. Reported values are means (standard deviations) over 5 runs with different random seeds.
 Part flipsUnfolded bounding area
MethodMAERMSEMAPE (%)MAERMSEMAPE (%)
UV-Net0.54 (0.01)0.86 (0.02)35.52 (0.90)14.60 (0.86)20.68 (1.17)5.90 (0.41)
PointNext0.59 (0.01)0.90 (0.01)39.33 (1.70)20.24 (0.93)28.82 (1.37)8.28 (0.49)
Baseline0.9841.18067.6789.81111.3546.01
Table 3: Complexity regression results. Both tasks are evaluated on the BenDFM subset. Reported values are means (standard deviations) over 5 runs with different random seeds.
More in this topic

Synthetic Benchmark Generation

Synthetic benchmark generation creates controlled, verifiable evaluation tasks and datasets to test models across diverse settings.

See all 287 papers

Explore similar work

CardsList
May 12, 2026

CAD-feature enhanced machine learning for manufacturing effort estimation on sheet metal bending parts

Graph-based machine learning has emerged as a promising approach for manufacturability analysis by learning directly from CAD models represented as Boundary Representations (B-reps), exploiting both surface geometry and topological connectivity. However, purely geometric representations often lack the process-specific semantics required for accurate manufacturability prediction: many manufacturing factors, such as surface roles or bend intent, are not explicitly encoded in shape alone and are difficult for data-driven models to infer reliably. We propose a hybrid approach that addresses this challenge by enriching B-rep attributed adjacency graphs with manufacturing features recognized through a rule-based module. Applied to sheet metal bending, recognized features, such as bend characteristics, flange lengths, and surface roles are integrated as node attributes, concentrating the learning signal on process-relevant geometric patterns. Experiments on both a large-scale synthetic manufacturability benchmark and a real-world industrial dataset with measured bending times, one of the first such validations on genuine production data, demonstrate that combining domain knowledge with graph-based learning improves prediction accuracy across both tasks. The results demonstrate that hybrid modeling offers a feasible and effective path toward deployable tools for manufacturability assessment and effort estimation in industrial CAD environments.

Matteo Ballegeer, Toon Van Camp, Willem Jaspers +6

ManufacturingGraph Representation Learning
AskPDF
Sep 3, 2026

LLM-Aided Design for Manufacturing: A Multi-Agent System for Intent-Preserving Redesign of CAD for Improved Manufacturability

We introduce autonomous, intent-preserving Design for Manufacturing (DFM) redesign of CAD parts: given an engineer's CAD model, the method returns a variant that is easier to manufacture without losing its design intent. Generating such a redesign in a single shot is unreliable, since CAD fidelity degrades as parts grow complex; we instead produce it as a sequence of individually verified design transitions. Our DFM-Redesign pipeline realizes this with two coupled agent subsystems driven by a pretrained multimodal LLM: a DFM Reviewer that inspects the current design and proposes one intent-preserving manufacturability improvement at a time, and a CAD Modifier that executes each proposal as an edit to the part's CadQuery program. The CAD Modifier closes a verification loop, compiling every candidate edit and visually checking it against the intended change from multi-view renderings, then re-generating or re-instructing until the edit is accepted or abandoned. Iterating review and verified modification compounds edits into parts more complex than one-shot generators reliably produce, preserves the original intent at each step, and requires no fine-tuning. On a 46-part benchmark scored by chamfer distance to reference geometries, the CAD Modifier reproduces target parts more accurately on average than chain-of-thought single agents given the same tools, and ablations isolate the contributions of the visual review loop and of captioning the design state before each edit. A centrifugal pump casing built from 32 chained transitions illustrates the complexity reachable by compounding verified edits. This is a preliminary report: evaluation of the full review-and-redesign loop, including manufacturability gain and an operational measure of intent preservation, is ongoing.

Kojo Welbeck, Xiangyu Shi, Zahra Sadeghi +2

Department of Mechanical Engineering, Northwestern University, Evanston, USA · Department of Electrical and Computer Engineering, Northwestern University, Evanston, USA

Multi-Agent LLM SystemsParametric CAD ModelingLLM-Based Program Synthesis
AskPDF
Aug 10, 2026

CADEngBench: It Looks Like CAD, but Does It Work? Evaluating Parametric Design, Assembly Reasoning, and Physics Simulation

A CAD model is not engineering-grade merely because it looks correct. It must satisfy design requirements, respond predictably to parameter changes, support controlled edits, match a reference structural response under a declared analysis, and connect to other parts through valid joints. We present CADEngBench, a two-track benchmark for these capabilities. CADEngBench-P evaluates 300 parametric parts, each used for one zero-to-CAD task and one functional-editing task (600 tasks in total), through boundary-representation (B-Rep) validity, engineering and DFM checks, parameter-family perturbations, functional editing, and matched linear-static FEA in CalculiX. CADEngBench-A evaluates 150 body pairs through ranked joint retrieval, exact face-and-edge grounding, joint-frame prediction, and kinematic verification. Across eight multimodal, code-capable models, editing supplied CAD is substantially easier than generating it, while complex edits and matched FEA remain difficult. Assembly predictions often locate the relevant region but fail to recover the recorded joint or mating entities. These results show that CAD evaluation must test engineering behavior rather than appearance alone.

Harmanjot Singh, Abhra Dubey, Jorge Alejandro Amador Herrera

Mohamed Bin Zayed University of Artificial Intelligence · Abu Dhabi, UAE

Benchmark ConstructionParametric CAD ModelingBenchmark ValidityText-to-CAD Generation
AskPDF