Proceedings of the XMO Industrial Seminar 2026: Excellence in Manufacturing and Operations
Keywords
Digital twin; Cyber-physical production system; Machining process virtualization; Real-time geometric virtualization; Signed distance field
Tracks
DIGITAL MANUFACTURING
DOI
10.5703/1288284318688
Abstract
As the transition toward autonomous smart factories and Cyber-Physical Production Systems (CPPS) accelerates, real-time digital twin implementation of manufacturing processes has emerged as a critical enabling technology. However, existing virtualization approaches present fundamental limitations: mesh-based Boolean methods are computationally infeasible for real-time processing due to their CPU-bound operations, while pure voxel-based methods suffer from staircase artifacts at geometric boundaries and exhibit exponentially increasing computational demands as resolution improves. To overcome these limitations, this paper proposes a GPU-accelerated real-time geometric virtualization method combining Signed Distance Field (SDF) with the Marching Cubes algorithm. The proposed method executes massively parallel SDF subtractive operations between the tool and workpiece geometries via GPU Compute Shaders, while Marching Cubes extracts smooth, artifact-free iso-surfaces in real time even at low voxel resolutions. Comparative experiments across three methods demonstrate that the proposed approach achieves an average of 104.6 FPS, CPU frame time of 10.2 ms, and GPU frame time of 8.8 ms, representing approximately 35x improvement over the Mesh Boolean method (3.0 FPS, CPU 333.3 ms) and approximately 1.8x improvement over the Pure Voxel method (57.1 FPS, CPU 17.6 ms). The proposed framework supports configurable tool geometry and machining parameters, enabling broad applicability across diverse subtractive equipment including CNC machining centers and milling-capable robotic arms within factory digital twin platform.
A generalized real-time geometric virtualization of material removal processes for a factory digital twin platform
As the transition toward autonomous smart factories and Cyber-Physical Production Systems (CPPS) accelerates, real-time digital twin implementation of manufacturing processes has emerged as a critical enabling technology. However, existing virtualization approaches present fundamental limitations: mesh-based Boolean methods are computationally infeasible for real-time processing due to their CPU-bound operations, while pure voxel-based methods suffer from staircase artifacts at geometric boundaries and exhibit exponentially increasing computational demands as resolution improves. To overcome these limitations, this paper proposes a GPU-accelerated real-time geometric virtualization method combining Signed Distance Field (SDF) with the Marching Cubes algorithm. The proposed method executes massively parallel SDF subtractive operations between the tool and workpiece geometries via GPU Compute Shaders, while Marching Cubes extracts smooth, artifact-free iso-surfaces in real time even at low voxel resolutions. Comparative experiments across three methods demonstrate that the proposed approach achieves an average of 104.6 FPS, CPU frame time of 10.2 ms, and GPU frame time of 8.8 ms, representing approximately 35x improvement over the Mesh Boolean method (3.0 FPS, CPU 333.3 ms) and approximately 1.8x improvement over the Pure Voxel method (57.1 FPS, CPU 17.6 ms). The proposed framework supports configurable tool geometry and machining parameters, enabling broad applicability across diverse subtractive equipment including CNC machining centers and milling-capable robotic arms within factory digital twin platform.