The problem
Plastic defects that 2D vision misses
Inspection of moulded, extruded and thermoformed plastic parts is still largely manual or semi-automated, and usually happens after production. Sink marks, warpage, flash, surface cracks, colour variation and contamination are hard to catch with conventional 2D vision because of how plastics reflect light.
The WIZARD approach
- A 6-axis industrial robot scanning at constant distance with adaptive path correction
- High-resolution vision, 3D laser triangulation and optional infrared thermal sensing
- Multimodal fusion with CNN and Transformer models for sub-second inference
- 2D and 3D defect heatmaps showing location and severity
- Defect maps turned into path coordinates for the MAGICIAN rework robot to polish or deburr
- ROS2 and OPC UA interfaces into MES and ERP systems
Objectives
What WIZARD sets out to do
01
Integrate with MAGICIAN
Full technical integration with the MAGICIAN framework: ROS2 interoperability, OPC UA links to production systems, and secure, encrypted pipelines for multimodal sensor data.
02
Industrial-grade detection
Optimise the multimodal AI architecture to industrial performance, classifying at least six defect categories with confidence scores and severity estimates.
03
Human-centred interface
An explainable dashboard with interactive heatmaps and operator-in-the-loop validation, so people stay in control of quality decisions.
04
Validate in industry
Demonstrate the integrated system under real production conditions in a plastic-moulding environment.
05
Reach TRL 7
Move from TRL 5 to TRL 7 through integrated demonstration and validation, with a commercialisation roadmap.
Proposed targets
Measurable from the start
>99%
Detection accuracy
<2%
False-positive rate
0.1 mm
Smallest detectable defect
<1 s
Inference time
<200 ms
Multimodal fusion latency
≥50 mm/s
Inspection speed
72 h
Continuous operation
≥98%
System availability
Targets are as proposed to the host project and will be reported against as the work progresses.
Work plan
Five work packages over twelve months
- WP1Project set-up, requirements and MAGICIAN onboarding
- WP2Multimodal AI model development for plastic defect detection
- WP3Sensing hardware adaptation and system integration
- WP4TRL 7 industrial demonstration, human-robot interface and rework hand-off
- WP5Project management, dissemination and exploitation
The host project
MAGICIAN
MAGICIAN is a Horizon Europe project building a modular automation solution in which robots detect and rework production defects before the final production stages. It runs from October 2023 to September 2027.
Visit the MAGICIAN projectOpen Call 2
The second open call funds SMEs to take MAGICIAN’s sensing and rework robots into sectors beyond automotive. WIZARD extends the sensing robot with 3D and thermal modalities and links its output directly to the rework robot.
Funded by
the European Union
WIZARD receives funding from the MAGICIAN project, which is funded by the European Union’s Horizon Europe research and innovation programme under grant agreement No 101120731.
Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the granting authority. Neither the European Union nor the granting authority can be held responsible for them.