“There exist multiple programmatic methods and technical architectures for editing industrial mesh-based 3D models with a focus on preserving precision, geometry, and engineering features, including direct mesh manipulation, mesh-to-CAD reconstruction, voxel/SDF workflows, primitive recognition, and hybrid pipelines, each with distinct trade-offs in dimensional accuracy, feature retention, and industrial suitability.”

In an industrial context, recovering a continuous model is necessary to make modifications or to exchange data with a format including continuous representation of objects like STEP. The first step is to detect simple primitives like: plane, sphere, cone and cylinder from a 3D CAD mesh. This method of detection uses curvature features to recover each primitive type, with segmentation based on curvature features computed for each vertex, allowing extraction of sub-meshes corresponding to primitives. Parameters of these primitives are found with a fitting process, yielding angle errors lower than 0.01% for planes and between 0.03% and 1.99% for spheres, cones, and cylinders, which are deemed sufficient for industrial applications.

A robust surface mesh denoising method that can effectively remove mesh noise while faithfully preserving sharp features using surface fitting and projection techniques. Sharp features are preserved in the surface fitting algorithm by considering an anisotropic neighborhood of each vertex detected by the normal-weighted distance. The method achieves promising mesh denoising while retaining both sharp (high-curvature) and non-sharp (low curvature) features, with excellent performance in preserving geometry such as surface areas, volumes, and local curvatures.

Presents a feature-preserving mesh decimation technique that adapts the triangle mesh to local geometry in complex surface structures while removing oversampling from flat featureless regions. The method uses an anisotropic formulation that aligns edges and vertices with ridges and furrows of underlying geometry, maintaining delicate surface details even at high compression ratios (around 90%) while achieving sub-millimetre surface accuracy. The approach combines quadric error measures with optimal Delaunay triangulation for screen space applications.

Proposes a non-iterative feature-preserving filtering technique based on robust statistics and local first-order predictors applicable to arbitrary triangle soups and non-manifold meshes. Feature preservation is achieved through a robust influence weight function combined with a predictor for vertex positions based on tangent planes, which does not move vertices at sharp features separating smooth areas. The method provides a fast and stable one-step smoothing operator suitable for large, noisy meshes without requiring manifoldness of input data.

Open CASCADE Technology (OCCT) is the only open-source full-scale 3D geometry library. It is a backbone of many projects including FreeCAD (providing parametric 3D modelling), KiCad (free 3D CAD software for PCB layouts design), Gmsh (a fast meshing tool), and Salome (a platform for numerical simulation).

OCCT3D includes a set of C++ class libraries providing services for 3D surface and solid modeling, visualization, data exchange and rapid application development. It is a software development kit (SDK) intended for the development of applications dealing with 3D CAD data and is available under the GNU Lesser General Public License (LGPL).

We introduce the first framework that reformulates mesh editing as an image editing–mesh generation–seamless fusion pipeline integrating 2D and 3D generative models. We propose a Joint Geometry and Appearance Fusion framework consisting of Poisson Geometry Blending for seamless geometric integration and Poisson Texture Harmonization for natural texture blending. Compared with direct 3D editing, these 2D approaches are lightweight, controllable, and well-suited for performing high-quality object-centric image manipulation.

This patent describes a global structure-based 3D mesh repair method in four stages: hole detection, decomposition into base model and high-frequency details using improved bilateral filtering, parameterization of control regions to 2D plane, and repair of geometric detail images mapped back to the 3D mesh. It preserves geometric details and structure, reducing complexity and improving efficiency over prior methods, exemplifying direct mesh manipulation and hybrid pipelines for precision editing.

To improve robustness and accuracy in 3D mesh segmentation, this method uses energy optimization and distinguishability features, addressing over/undersSegmentation, jagged boundaries, and manual intervention issues. It defines energy functions for clustering, enhancing segmentation precision and feature retention in mesh models.

libigl is a simple C++ geometry processing library with functionality including construction of sparse discrete differential geometry operators and finite-elements matrices such as the cotangent Laplacian and diagonalized mass matrix, simple facet and edge-based topology data structures, and mesh-viewing utilities. It is a header-only library that operates on generic triangle meshes stored in vertex positions and triangle indices matrices.

CGAL, libigl, and PMP have an emphasis on the implementation of sophisticated mesh processing algorithms, lacking rendering and interactive capabilities. Easy3D is an open-source library for 3D modeling, geometry processing, and rendering, designed to integrate end-to-end 3D data processing capabilities that existing libraries like CGAL and libigl do not provide together.

Describes an adaptive remeshing scheme that preserves features in original meshes by maintaining a feature list throughout the pipeline. Feature edges are determined by computing the angle between normals of vertices adjacent to an edge, with this angle used to classify edges as features or non-features. The approach combines adaptive remeshing with explicit feature preservation mechanisms.

Demonstrates that feature-preserving mesh decimation can maintain delicate surface details effectively even at high compression ratios by aligning mesh edges and vertices with ridges and furrows of underlying geometry. The method includes three local mesh operations: edge collapse, edge flip, and vertex position update, achieving sub-millimetre surface accuracy while reducing processing time from hours to minutes for high-resolution normal map integration.

The mastery stack for faster reconstruction is consistent: clean and segment the mesh, recognize analytic geometry, build a curve skeleton that reflects design intent, surface with controlled continuity, then validate with deviation-driven iteration. Region/feature segmentation separates the mesh into logical surface families such as planes, cylinders, blends, and freeform zones so that each can be fit and surfaced using the most appropriate method. Feature Extraction: Best-Fit Primitives & Analytical Recognition (planes, cylinders, cones, spheres) turning triangle regions into analytic CAD features.

This paper compares six methods for morphing hexahedral and tetrahedral meshes, including the previously published FEMWARP and LBWARP methods as well as four new methods: smoothing alone, weighted residuals, simplex-linear, and simplex-natural neighbor. Element quality and performance results show that different methods are superior on different models. We recommend that designers consider both the FEMWARP and a linear simplex based method.

This article outlines best practices for 3D meshing in CAE workflows, including user-defined size controls, adaptive meshing, gradient, boundary layers, and iterative adaptation to ensure precision in geometry capture for simulation. It supports mesh-based inputs like BREP and ACIS, with physical and geometric sizing for consistent element sizes and curvature approximation, highlighting trade-offs in accuracy and mesh quality.

Meshtron uses machine learning from artist-created meshes for data-driven high-fidelity 3D mesh generation at scale, addressing scalability limits in prior ML methods and enabling precise geometry preservation in large-scale editing workflows.

QUICKSURFACE converts raw 3D scan data into precise, manufacturable CAD models, combining powerful reverse engineering tools with an intuitive workflow. It supports parametric and freeform modeling, allowing for both precise engineering parts and organic geometries, and exports clean, production-ready CAD models in formats like STEP and IGES for integration into downstream systems. The software is designed to handle complex industrial parts, rebuilding mechanical components and detailed assemblies with high geometric accuracy even from dense or imperfect scan meshes.

Reverse engineering is the transformation of the mesh data that is created during 3D laser scanning, into a precise and editable CAD file. Feature extraction techniques allow designers to identify and extract key geometric elements, such as holes and complex curves, from the mesh model and convert them into parametric CAD features. Mesh to CAD conversion plays a pivotal role in this process, allowing designers to obtain accurate CAD models from 3D scanned or digitized mesh data.

We describe a way to construct a plausible and intuitive low-level workflow that turns one of two given meshes into the second by building mesh correspondences. Analogous to text version control tools, we visualize the mesh changes in a two-way, three-way, or sequence diff, and demonstrate how to merge independent edits of a single original mesh, handling conflicts in a way that preserves the artists’ original intentions.

Proc3D is a system designed to generate editable 3D models and enable real-time modifications by introducing a procedural compact graph (PCG) representation. This graph encodes algorithmic rules and structures, exposing key parameters for intuitive manual adjustments and automated modifications via natural language prompts using Large Language Models (LLMs). Experimental results show Proc3D achieves over 400x speedup in editing efficiency compared to conventional methods, enabling precise, real-time parametric edits.

We present an algorithm to reconstruct the computer-aided design model from a deformed mesh. Based on the extraction of geometric primitives from a 3D mesh to detect geometric primitive types of the mesh and to compute the parameters that give the best fit. To define the topology of the object, intersections between primitives should be calculated.

The CGAL Mesh Generation Package provides tools for generating high-quality tetrahedral meshes for 3D domains, as well as 2D triangular and 1D meshes. It is designed for applications such as finite element methods (FEM) and numerical simulations, with advanced algorithms for mesh refinement and optimization. The package supports complex geometries, including implicit functions, CAD data, and polyhedral surfaces, and offers adaptive meshing based on user-defined criteria.

Focusing on the best performing meshes, tetra patch conforming and the method multizone hexa dominant, the hexahedral mesh showed values of the Jacobian ratio and of the aspect ratio closer to the optimum target values. In terms of the aspect ratio, the hexahedral mesh showed better values. Regarding the two meshes, multizone hexa/prism hexa core and multizone hexa/prism tetra prism, the improvement of the accuracy of the solutions by increasing the number of elements and nodes was observed, with distinct trade-offs in computational resources.

Cyborg3D MeshToCAD is a standalone product designed for reverse engineering complex scanned meshes and ZBrush sculpted meshes into CAD systems. It offers powerful tools to repair, modify, and smooth meshes, capture fine details, and convert Sub-D meshes to high-quality NURBS-based boundary representations (G2 along the edges) for export to standard CAD formats like STEP, IGES, and SAT. While it simplifies reverse engineering of freeform organic shapes, existing products often specialize in precise prismatic parts, indicating a trade-off in handling different geometry types.

Voxel meshing fills an enclosed volume with voxels of a pre-defined size, primarily used for topology optimization. This method requires the volume to be enclosed as geometry or shell elements without T-connections or free edges for successful creation. While useful for design exploration and generating design spaces for topology optimization, it does not give meaningful results in a stress analysis, highlighting a trade-off in its industrial suitability for certain analytical tasks.

Cg3Lib is a C++ geometry processing library developed by the CG3HCI Group of the University of Cagliari. It is composed of different modules for geometry processing tasks.

Remesh Surface: The mesh generation blocks in nTop do not produce mesh with an exact element size. This block can be used to create a mesh with explicitly chosen properties using triangle or quad elements. Remeshing comes in handy when you need to control the mesh properties and prepare them for simulation, with considerations for manufacturing tolerances.

This research on 3D mesh deformation algorithms supports industrial design, animation, and geometric analysis, providing methods for precise manipulation while preserving geometry and features, as essential material for CAD and engineering applications.

VoxelMeshSDF allows baking a mesh into a signed distance field (SDF) stored as an internal voxel buffer, which can then be used for voxel-based operations. However, it is noted that procedural shapes (like spheres or boxes) are generally preferred for better quality and performance, as VoxelMeshSDF is more expensive and currently has lower quality compared to its procedural equivalents. Additionally, not all meshes can be baked, with best results from manifold and closed shapes.

Mesh-to-CAD reconstruction methods convert triangle meshes into parametric CAD models by identifying geometric primitives (planes, cylinders, spheres) and fitting them to mesh data, enabling feature-based editing with dimensional control. Voxel-based and signed distance field (SDF) workflows provide alternative representations that trade mesh precision for volumetric editing flexibility, commonly used in additive manufacturing and topology optimization. Hybrid pipelines combine multiple representations—such as mesh decimation followed by primitive recognition or voxel-to-mesh conversion—to balance accuracy, editability, and computational efficiency for industrial applications.

How to convert a triangle-based model to a parametric CAD model. Workflow for converting mesh to parametric CAD.

Hyper3D.ai's mesh editor offers exceptional precision control through advanced vertex and polygon tools, enabling meticulous adjustments to individual components with high accuracy for professional 3D work. It emphasizes workflow efficiency by consolidating complex tasks into a unified environment and provides automated mesh repair to ensure models are print-ready and impeccably structured. The editor's versatility caters to diverse industry needs, from game optimization to physical manufacturing, focusing on geometric accuracy for 3D model specialists.

Learn why mesh quality is critical in FEA. Explore key metrics, challenges, and actionable tips to optimize mesh accuracy, efficiency. Different meshing strategies impact dimensional accuracy and suitability for engineering analysis.

MeshLab, a free, open-source application, provides an extensive array of tools for mesh manipulation, including cutting, trimming, and slicing options, indispensable for professionals in digital fabrication and 3D printing. It facilitates precise editing through intuitive 'Selection' and 'Delete' capabilities, allowing users to outline and remove specific portions of a model. For more accurate cuts, filters like 'Cut with a Plane' can be applied, enabling users to specify a plane for slicing the mesh and export the polished model.

Best way for STL to be modified within a CAD solution without losing surface quality. Depends on the STL; for heavy high-poly mesh, convert to make additional cuts and changes.

Traditional solutions use 'segmentation-reconstruction' pipelines for editable part-level meshes, assessing geometric integrity, texture quality, and IoU for components, representing primitive recognition and hybrid workflows for precise, editable industrial 3D models.