Impeller vane and surface quality inspection

The fuel pumps of modern cars use a high quality impeller. For a long lifespan and smoothless operation the surface of the impeller is lapped. During this lap process small fractions or damages can occur. One of the vanes can be damaged which results in leakage of the fuel and in time to malfunction of the pump mechanism. Pits and scratches will also cause a malfunction of the fuelpump. Because both sides of the impeller have to have a certain flatness quality both sides will be lapped, which means both sides need an inspection.

Vision inspection setup

The vision inspection system inspects the front side as well as the backside of the impeller. Therefore two measurement positions on the inspection machine are used. At every inspection position a digital camera with optics and a circular led illumination is used. After the impeller is positioned in front of the camera the vision system acquires an image and starts processing.


Thooth-error-subimage For this inspection the graphical programming language LabVIEW is used. The vision software runs on a Industrial Personal Computer. Due to position tolerances of the pick and place unit the position of the impeller in the image is not accurate enough. To Solve this problem first of all the outside of the impeller is detected and a circle fit displays the accurate position of the impeller within the images. The second step is a surface inspection of the circular area outside the vanes.

The vision system locates the first vane within a certain area followed by a position detection of every vane. For every vane the width and the distance between the vanes are measured followed by a surface inspection.

The final step is to detect the fuel holes in the inner part of the impeller. These areas are excluded from the surface inspection.

The front side as well as the backside do have a different pattern. Therefore each side has his own mask.



Unfortunately the backside also has a few marks which the software doesn't inspect. The surface inspection outside of the holes is performed by using a dynamic object finding algorithm.

When a failure exceeds a certain predefined adjustable limit the impeller is classified as a failure. If a failure is detected, a sub image of the failure (or failures) is visualised in the left top corner of the image.

In the enlarged sub images the error is visualised. An error is reported to the machine control system, and the impeller is rejected.

The error images are saved on disk for off line analysis and improvement.