Synthetic Defect Geometries of Cast Metal Objects Modeled via 2d Voronoi Tessellations

TL;DR

The paper tackles the challenge of generating大量 synthetic, labeled data for visual surface defect detection in cast metal objects, where real defect examples are scarce. It introduces a rule-based, parametric pipeline that uses 2D Voronoi tessellations to create defect footprints, then lifts and triangulates them into 3D meshes for digital twins. The main contributions are unified pipelines for elongated defects and scabs, parameter-driven variability, and pixel-perfect annotation, enabling scalable data generation and edge-case coverage while remaining transferable to other manufacturing contexts. The approach supports realistic imaging with Monte Carlo rendering and can be adapted to other NDT modalities, offering a practical path to robust defect detection systems.

Abstract

In industry, defect detection is crucial for quality control. Non-destructive testing (NDT) methods are preferred as they do not influence the functionality of the object while inspecting. Automated data evaluation for automated defect detection is a growing field of research. In particular, machine learning approaches show promising results. To provide training data in sufficient amount and quality, synthetic data can be used. Rule-based approaches enable synthetic data generation in a controllable environment. Therefore, a digital twin of the inspected object including synthetic defects is needed. We present parametric methods to model 3d mesh objects of various defect types that can then be added to the object geometry to obtain synthetic defective objects. The models are motivated by common defects in metal casting but can be transferred to other machining procedures that produce similar defect shapes. Synthetic data resembling the real inspection data can then be created by using a physically based Monte Carlo simulation of the respective testing method. Using our defect models, a variable and arbitrarily large synthetic data set can be generated with the possibility to include rarely occurring defects in sufficient quantity. Pixel-perfect annotation can be created in parallel. As an example, we will use visual surface inspection, but the procedure can be applied in combination with simulations for any other NDT method.

Figures (14)

• Figure 1: Visualization of the synthetic data generation pipeline using rendering in a virtual environment for visual surface inspection.
• Figure 2: Examples of common defects in cast metal objects.
• Figure 3: Overview of modeling pipeline including the different steps, adaptation to and relations between different defect types.
• Figure 4: Synthetic images of different defect types in a cast metal cube obtained by rendering from a synthetic industrial image acquisition system created in Blender blender. A pinhole camera and two plane lights are used for illumination. Rendering is performed using 24 samples per pixel and 16 light bounces. The surface texture is chosen to resemble the appearance of a cast metal object. The texture model was integrated by Lovro Bosnar from Fraunhofer ITWM and is based on the cellular texture introduced by Worley castingWorley.
• Figure 5: Illustrations of different definitions for a dilated path.