The challenge
Each element is manufactured based on a digital work drawing in IFC format. Manual quality control does not provide sufficient guarantee to detect production deviations timely, completely, and reproducibly.
The main challenges are threefold.
- First, it must be determined whether the physical elements match the digital working drawing with sufficient measurement accuracy.
- Secondly, there is a need for automatic detection of assembly errors and missing parts — such as pipes, insulation material, and cutouts.
- Thirdly, there is a lack of systematic recording of production steps per element for traceability and quality assurance.
The large work surface, the variety of wood materials, and the dynamic production environment—with machine vibrations, an overhead crane, wood dust, and changing lighting—make a direct application of standard vision technology not straightforward.
The proposed solution
Vision Partners has designed a central Smart Assembly Vision architecture, in which multiple workstations are connected via Ethernet to one central image processing module. The scalable setup is designed for a maximum of 16 workstations, distributed across multiple halls if desired.
The system consists of:
- Central image processing unit Control cabinet with industrial Windows/Intel vision processor, controlled via an HMI interface on a high-resolution monitor.
- Multi-camera setup (high resolution): Multiple industrial cameras per workstation are software-merged into one combined image of over 60 megapixels, for accurate comparison with the digital work drawing.
- IFC to DXF Conversion Customer master system data is automatically converted into a 2D inspection model, suitable for comparison with camera images.
- Best-fit and floating tolerance algorithms Take into account permitted variations in wood material and minor shifts or rotations within predefined tolerances to minimize false rejections.
- AI inspection Neural networks for anomaly detection and object detection check for the presence of pipes, insulation material, cutouts, and other assembly parts.
- Geometric controls Automatic check for squareness and other relevant geometric parameters.
- QR data code reader per workstation Provides automatic element identification and linking with the correct working drawing and product data from the parent system.
The next step: Proof of Concept
Before the full Smart Assembly Vision solution is deployed, Vision Partners will conduct a Proof of Concept. The goal is to validate the technical and commercial feasibility of the architecture under realistic production conditions. If the PoC is successful, the system can be scaled up to multiple workstations in different production halls.
Want to learn more about machine vision in manufacturing? Get in touch
Read more about this case: Optimize construction process on the Sensor Partners website.
