Fully Automatic X-ray Inspection of Aluminium Wheels
Dipl.-Ing. Georg Theis
YXLON International X-Ray GmbH, Germany
Corresponding Author Contact:
Email: yxlon@yxlon.com, Internet: www.yxlon.comPaper presented at the 8th ECNDT, Barcelona, June 2002
Abstract
This paper describes the requirements, the structure, software setup, inspection procedure
and economical advantages of fully automatic X-ray inspection systems for aluminium
wheels in an industrial environment.
Introduction
In the automotive industry aluminium wheels are safety relevant parts which have to be
inspected by X-ray.
Most machines use a visual inspection by an operator. In the future, most of the inspection-
systems will be designed as fully automatic inspection systems with an ADR (Automatic
Defect Recognition) – software (1).
Wheel customers now require these fully automatic inspection systems, because the defect
recognition is more reliable than visual inspection. When comparing a fully automatic
wheel inspection system with visual inspection, the automatic testing is usually much more
economical.
The following regard and examples are done with the YXLON MU231 X-ray system with
Automatic Inspector (AI) image processing. This solution is used more than any other
worldwide for X-ray wheel inspection.
Inspection System requirements for industrial use
The inspection systems are mostly integrated as inline-systems in a foundry. Most
requirements of an inspection system are based on detection of defects and on the inline -
integration in an industrial environment.
-
high availability (approx. > 95%)
- high throughput (>100 wheels per hour)
- detection of all relevant defects with a minimum of false rejects
- minimum of maintenance effort
- easy system-handling by the operator
- mixed operation mode (different wheels)
- fast and easy setup of the image processing software
- complete and fast service by manufacturer of the system
- space-saving system
Hardware - structure of an inspection system
Inspection Cycle
- Conveying a decollated wheel by a shuttle in the cabin/manipulator
Wheel-type recognition by the Vision system of the AI:
It selects the inspection program for this type of wheel in the AI and recognizes the
orientation of the wheel. The information will be send from the AI to IPC and PLC.
The X-ray parameters and the mechanical position for every wheel type is stored in the
IPC. This ensures that the system can be used in semiautomatic mode without AI
(visual inspection that requires no positioning by the operator, because the
manipulation of the wheel is done by PLC).
- Manipulation and image processing:
To get the necessairy inspection Steps of the wheel, the MU231 uses the tilt axis (X-ray
tube and Image Intensifier) and the wheel manipulator (rotation and linear movement
of the wheel). If the wheel is in the programmed position, the image intensifier sends
the X-ray image to the AI. The AI grabs this image and tells the PLC to position the
next view. Then the wheel will be positioned for this view and so on. This ensures that
the mechanical positioning doesn’t have to wait for the image processing, so cycle time
will always be as short as possible. The number of views required for the inspection
depends of the design, test specification and size of wheel. A typical 17‘‘-wheel
requires approx. 23 views.
- Assigning the result of inspection:
At the end of the inspection cycle, the result will be sent from the AI to the PLC while
the wheel is conveyed outside the cabin. With this information, the conveying system
of the manufacturer decollates the rejected wheels.
Setup of the image processing software
After a short training (approx. 5 days), a new operator will be able to setup a new
inspection program with the AI software quickly and reliably.
The setup requires the following Steps:
Step 1: Teaching the Vision Software for wheel-type recognition (see fig. 3).
Fig 3: Image of the Vision System with a marked valve hole.
|
Step 2: Sampling X-ray images of different views of the wheel with the AI to teach
variations.
Step 3: Drawing the Regions Of Interest (ROIs) in one X-ray image (see fig. 4 a). With
ROIs one has full control and can inspect every part of the image with different
settings.
Fig 4 a – c: X-ray images with ROIs, different views.
|
Step 4: Teach the system to place the ROIs on sampled images (see fig. 4 b, c)
Step 5: Training of the Neural Network in the AI (no action of the operator required)
Step 6: Entering the test specification (can be different for every ROI!)
Step 7: Testing the new inspection program in production
For the Steps 1,2 and 7, a complete X-ray system with image processing is required. Steps
3 to 6 can be continued at an offline system (PC with the image processing software; e.g.,
placed in an office). During the work on the offline system, the complete X-ray system can
be used for production.
Wheel - Inspection with the image processing software
The inline - inspection of the wheel has no influence on the cycle time; it is working
parallel to the mechanical positionning and image sampling. To get enough processing
power, the hardware is using a Host-PC with several Slaves-PCs (see fig. 2).
The main Steps of the inspection are:
Step 1: Substracting the incoming image (see fig. 5) from a filtered image to receive a
background (see fig. 6) image showing differences in material (like defects)
Fig 5: Incoming Image.
| 
Fig 6: Background Image.
|
Step 2: Binarization - Distinguish noise from defects and reports for each defect the
-
Type of defect (gashole, porosity, shrinkage cavity, pores)
- Size of defect
- Total amount of defects in a given zone
Fig 7: Image after Binarization.
|
The Binarization can use a different classification for each ROI. The classification
can be setup to fulfill standards of ASTM.
Step 3: Detection of large areas of missing material by comparing of the incoming image
and reference image.
Step 4: Before giving the final result to the PLC, the AI will verify the defects with special
algorithms.
With a system like AI its also possible to setup a mould recognition by X-ray.
Integration of the inspection system in the casting process
The wheel-manufacturers install the X-ray system close to the casting process, because
they want to sort out the rejects as soon as possible (to save costs in the following
processes like machine processing) and to recognize and to solve casting-problems.
Fig 8: Optimizing the casting process.
|
If the actual tested images and statistic informations are displayed near the mould, the
moulder can correct the quality of the wheels (see fig. 8).
Economical consideration
An (very simplified) economical comparison of visual and a fully automatic inspection
system with the following parameter
-
additional Costs for the ADR – Software: approx. 130 kEURO
- Rate of interest: 8%
- Maintenance: approx. 500 EURO per month
- Cost for labour for visual inspection: 20 EURO/h
- 3 shifts per day, 7 days per week
shows that there’s a return of investment in less than one year (see fig. 9). There are
additional financial advantages by optimizing the casting process with a system like AI.
Fig 9: Costs for automatic and visual inspection.
|
Prospect
Comparing the fully automatic X-ray inspection of aluminium wheels with the visual
inspection, the main advantages of the automatic inspection are:
- more informations to optimize the casting process by having statistical data and
showing X-ray images at the casting machine (mould)
- reliable detection of defects
- saving labour costs because no operator is required for inspection
In the future, the fully automatic inspection will replace the visual inspection. The number
of wheel customers (automotive industry) is increasing who demand automatic inspection
in foundries.
Reference:
- T. Kahrs, G. Theis: Vollautomatische Inline-Röntgenprüfung von Aluminiumrädern,
DGZfP May 2001, Berlin
Fig 1: MU231 X-ray Inspection System
Fig 2: MU231 X-ray Inspection System with Image Processing System