Fuel Pellets Automatic Visual Inspection

·Home
·Table of Contents
·Materials Characterization and testing

Fuel Pellets Automatic Visual Inspection

F.Aguirre A.Domingo - ENUSA (Empresa Nactional del Uranio S. A.) 'Spain
Contact

Nuclear Plants operate with UO₂pellets confined inside hermetic tubes. Fuel Pellets are ceramic cylindrical bodies of around 10 mm. length and Ć 9 mm.

Fuel Pellets

These Fuel Pellets have very tight requirements for surface integrity, so they must be visually inspected on a 100% sampling rate. This inspection is performed after the final grinding operation at the production line by visual inspectors. Production is intensive, and 300 pellets/minute are typical values obtained.

Main defects considered are cracks and chips in both lateral and end surfaces.

For avoiding tiredness problems in visual inspection, reducing radiological exposure on the inspectors and reducing subjectivity in the inspection, ENUSA has development an automatic Visual Inspection System.

This device is formed mainly by a table on which a two arms Cartesian robot is placed. One arm has a vision system and the other one has a grip. There is a loading area at one end of the table , then an inspection area, a rejection area and finally an unload area (Figure 1). The vision system is controlled by a PC, and the whole system by a robot and PLC integration.

Pellets are introduced in the loading area on trays containing around 1400 specimens. Afterwards they are manually pushed into the inspection area where Fuel Pellets are turned on rollers for the whole surface to be scanned. A digital linear CCD camera placed over one of the arms scan each pellet surface and at the same time a grip laid on the other arm retires previously inspected pellets from the rejection area. When the process is finished, pellets are placed again on another tray and they are ready to go on the process.

Fig 1:

Fig 2:

The vision system is formed by an halogen light source which is transmitted through an optical fibre to a cylindrical magnifying glass, oriented in the axial direction of the rollers. It is possible to have a light line over the generatrice of around ten pellets of a fuel pellet row. This lightened generatrice is scanned by a CCD linear camera with a fixed focus lens. As the pellets are turning on the rollers, it is possible with a proper time set-up to get a photograph of every fuel pellet surface. The camera, the lenses and the magnifying glass are attached to the robot arm, while the light source is fixed to the machine body. It is very important to have a perfect alignment of the light beam, the axis of the rollers and the focus of the camera, so the camera is mounted on a system with pins and screws which provides adjustment in two axis and displacement in the perpendicular direction to the inspection surface by means of a micrometric screw.(Figure 2)

Before the beginning of the inspection the system checks itself by comparing the signal of knowing dimensions an position notches against previously recorded values. In next Figure can be watched actual signal (blue) against previous (red) (Figure 3).

Fig 3:

Fig 4: Text refering this figure

After the checking, pellets can be inspected. Typical image processing is described below. Seven sample pellets are introduced in the system with the defects shown in (Figure 4).

The system scans a very narrow line (0.015 mm.) on the generatice of all the pellets. In Figure 5 scanning line is plotted in blue. Separation among pellets can be watched, and also the signal variations of the defects.

Fig 5:

Fig 6:

Because of the turning of the pellets a "photograph" of the pellets can be obtained, as shown in Figure 6. This "photograph" is formed by pixels 0.03 mm. long.

Fig 7:

Each pellet bright histogram is studied. A reference threshold is afterwards stablished, and in this way some areas are identified with "white blobs" and others with "black blobs" as shown in Figure 7. Inspection results depend on the good identification at this step of the defects.

Proper parameters (around 15) have been introduced to the systems for joining blobs that are very closed, for joining black and white blobs hat are related and for not considering minor defects noise).

Every blob is characterised in its dimension as, major and minor axis, area, gravity center location, area to inscribed area ratio, axis ratio, etc.

As defects are very good known, it is possible to classify them and perform the comparation against allowable defects dimension in visual standards. Limits can easily be changed when required. End chips defects is also possible to identify them as whenever there is a end chip defect, there is also a side indication.

Typical values obtained are 1% False Reject, 0.5% False Acceptance and a speed close to 300 pellets/minute for a product with 5% defects.

As conclussions:

An Automatic Visual Inspection System has been developed for the inspection of big quantities of small cilyndrical parts. The system is able to work with the same speed than a trained inspector.
Because the image analyses, real defects dimensions can be measured and compared against visual standards limits, without the subjectivity of an inspector.
Behaviour of the system is always the same, without the subjectivity among inspectors or the effect of tiredness.
System resolution is even better than a human eye.
Change in part dimensions or in the dimensions of the allowable defects are very easy to introduce in the system.