|NDT.net - October 2002, Vol. 7 No.10|
Automatic Defect Recognition (ADR) systems are more commonly used in automotive industry and have to be less cost intensive. Five points for cost reduction are presented: Higher throughput, more flexibility, lower false reject rate, higher reliability and a shorter delivery time. A new mechanical concept with optimized path of motion gives a 30 percent higher throughput and less vibrations for better image quality. The modular design allows multiple parts in one machine and is producible in short time. The results of a feasibility study will be compared with insights of a test setup.
|Fig 1: Basic application of an ADR system.|
In the automotive industry is an increasing demand for lighter components. State of the art technology is to use aluminium casting; as castings may have voids or other type of defects all safety relevant parts have to be inspected one hundred percent. Along with the growing number of inspected parts the need for a cost effective Automatic Defect Recognition (ADR) system comes with it, where the basic task is to inspect the parts between an X-Ray tube and a detector which can be oversimplified to a configuration where the x-ray tube points to the detector. The specimen has to be placed in the beam that the detector can process the x-ray. For defect analysis the specimen has to be penetrated from different angles in any region of the object.
The following aspects have direct influence to the cost per inspected part in the production:
Part of the paper will examine the mechanical concept. Former techniques base on moving the specimen in the X-ray beam (robot solution) or manipulate the tube and detector around the fixed part. The disadvantage of the first technique is that the plastic gripper wears out and as a result of that the accuracy goes down. As a consequence of the second method big masses have to be set in motion for long travel ranges and short ramp down time produces vibrations that have to be absorbed or lead to false rejects.
The idea of the new concept is to fix the specimen on a cardanic frame with a revolving inner ring. The detector and the X-ray tube head are manipulated in gliding planes above and underneath the specimen carrier (Figure 2). This concept minimizes the movement particular rotation of big masses outside their center of gravity.
|Fig 2: Optimized mechanic concept.|
X-ray source and detector orbiting around the specimen. The translational motion can be done by the orbiting arrangement or the specimen.
The specimen stays in the same horizontal plane and can easy be transported through the machine e.g. with a chain- or belt conveyer. In addition the time for load and unload parts can be optimised. More then one part at a time can fit on a palette that keeps the non-productive time down.
The mass that has to be moved is high because there is typically a strong mechanical C-or U-arm that orbits the source and the detector around the specimen. The fact that the mass is for away from the centre of gravity makes it hard to control and stabilise.
X-ray source and detector are stationary and the specimen is manipulated. This is a typical example of a robot system.
Robots are standard products and wide-spreaded in automotive factories. They have a high degree of freedom to place the part in the x-ray beam.
The gripper possibly masks out areas that might be important for the analysis. Robots are designed for a big range of applications and are not optimised for this problem. The robot has to grip the part from a palette and then move it into the x-ray beam. The part moves a long path back and forth from the palette to the x-ray beam. Robots are expensive and it needs a high level of technical knowledge to teach them.
This is what the market offers currently as ADR systems. But there is another constellation that is capable for the demands of an ADR system. The new concept combines both option 1 and option 2. X-ray source and detector are moved by a X-Y manipulator to move translationaly. The specimen is suspended is a cardanic frame like a compass.
|Fig 3: cardanic manipulator.|
The x-ray tube and the detector can be moved separately with a standard X-Y manipulator. The mass that has to be move for different angels close to the centre of gravity. The three axis consist of round frames which always better then open C- or U-arms. A higher stiffness and less vibrations can be achieved. The masses in on the frame can be mounted opposite so that the drives have to works mainly against mass inertia.
All advantages of option no. 1 are valid for this concept too.
The parts have to be fixed on the palettes very tight to prevent them from falling out.
With the new concept a position change can be done in less then 0,8 seconds (later B). Currently the automatic systems perform in range of 1 to 2 seconds (later A). The difference of 0,2 can be convert into money by the following calculation.
A typical part with 20 views for the image processing can be processed in 10 seconds (0,5
seconds a view). With 4 seconds for load and unload one part the overall cycle time is
A: 19 movements times 1 second plus 4 seconds load/unload + processing
= 33 seconds per part
B: 19 movements times 0,8 second plus 4 seconds load/unload + processing
= 29,2 seconds per part
The clock ticks 86400 seconds a day hence the daily throughput is
A: 2618 parts per day
B: 2958 parts per day
With the new concept 340 parts more can be inspected. If the margin is just 1 USD per part the yearly earning 124.100 USD.
The new concept also handles more then part a palette so the calculation can be repeated. For three part on one palette the situation is:
B: 59 movements times 0,8 second plus 4 seconds load/unload = 81,2 seconds per 3 parts
3192 minus 2618 equals 574. A year at 1 USD a part the company can inspect parts worth
209.510 USD more.
Even if this is not a 100 percent realistic calculation for every situation the effect that 0,2 seconds faster does in money is quiet measurable.
In times where keywords like just-in-time and rapid prototyping getting more important the need for flexible machine is increasing. We understand flexibility as how easy and quick it is to teach a new program. The human interface is based on a computer monitor plus a light and handy operator panel. PC software guides the operator through the entire teach process.
One topic at the beginning of the new concept was the focus on a standardised solution to deliver machines to the market in a very short time. During research of the machine this issue was kept in mind for every single component. Furthermore the mechanical design is optimised for the efficient construction period. After the first delivered projects the target is to deliver units within six to eight weeks.
The machine consists of a core with well defined and easy to connect interface. If the machine has to be integrated into an existing production line tit can be done with a separate unit. This makes it possible to concentrate all manpower to the specific problem for the project. The result is a shorter time of manufacturing and opens the system to almost every situation.
ADR software demands a high level of precision in position accuracy. The software currently available use different filters in several regions which will be adapted to the structures of the test objects. One of the most critical adjustment in the software is to match the filters with the internal structures of the test object to ensure that defects are detected also at edges, holes etc..
Variances in positioning will shift the image on the detector and the filters that are set up to the edges/holes may produce false defects; also rotation variance will influence the grey values in the image which also may lead to a higher false reject rate or - which is much worse - to non-detection of real defects. To ensure that no real defect will be missed the software has to be adjusted to a more sensitive level and the regions for image processing have to be enlarged.
As lower the variance in positioning is as closer the regions for image processing can be drawn to the structure of the part. In addition the sensitivity of the ADR software can be reduced to ensure that no real defect will pass the system without getting a higher false reject rate.
The new concept enables a higher precision in the handling of the test objects; compared to traditional systems with C-arm manipulators or robot systems the accuracy in positioning is 20 % better. This reduces the false rejects and of cost of ownership a lot.
|Fig 4: Two drives for one axis.|
Often reliability is big issue in automotive industry. Downtime is often worth some 1000 US Dollars a minute. The new concept terminates this with backing up the key components within the machine. Only one example that the force that comes from servo motors are split-up in two units. If one fails the system runs with the second motor. The machine, of course, performs in a slower mode then but it can be kept alive until the part arrives. Consequently the cost for spare parts can be minimised.
|© NDT.net - email@example.com|||Top||