Ariel Atenco, Jordy Rosbergen, Abhishek Chudal, Victor Reinders (2016)
Weighpack in The Netherlands was founded in 1974 with the aim to design, manufacture and service packaging lines. These packaging lines are suitable for high density, high weight, technical products, it marked the start of a company, which over the years would become the world’s most renowned specialist in complete hardware packaging lines now exporting its Hardware Systems to more than 45 countries around the world.
Weighpack is specialist in development and production of inspection and weighing machines as well as complete packaging lines for technical products. For many years Weighpack International is providing heavy duty, high quality packaging lines, therefore Weighpack machines guarantee reliability and safety. Weighpack also construct flexible custom made machines.
Weighpack does inspection of different products like anchors, automotive parts, bolts, nails etc. The motto of the company is “to think ahead” and therefore they want to improve the current inspection system of bolts. This improvement will be to add a camera to the system that checks the bolt for the right dimensions and damages to the bolt. Apart from the improvement they also want to make this inspection as fast as possible. However, this requires deep research before implementing the system previously mentioned. This research will consist of the type of camera that can check the desired criteria and a new process to inspect the product as fast as possible. After this research a proof of concept will be delivered to test the validity of the research.
The new process that is designed can be seen on Figure 1 and Figure 2. Because there is only one camera, which is used for inspection, it’s not known what the orientation and position of the bolt is at the end of the process. To have a complete inspection of the bolt it has to turn 120 degrees two times. To tackle this problem a guide rail is made so that the bolt will be on the same position and orientation at the end. So the robot knows the exact position of the bolt and will turn the bolt or will put the bolts in the approved bin. The guide rail system can be seen on Figure 1 and Figure 2. At the end of the guiding rail a funnel is placed to hold the bolt in the end position that can be seen on Figure
Figure 1: The new inspection process
Figure 2: The new inspection process
Figure 3: The funnel to hold the bolt at the end of the process
The automated process follows certain steps to reliably classify a bolt as approved or disapproved. The steps for this process are as followed:
- The bolt arrives at the conveyer between the guiderails
- The bolt will pass the light sensor that activates the camera
- The camera will scan the bolt and make a 3D image
- This 3D image is checked on the parameters
- The camera will communicate with the robot if the side of the bolt that is shown has defects
- If it has defects the robot will place the bolt in the disapproved bin
- If it didn’t see any defects the robot will turn the bolt 120 degrees
- The bolt will then be moved back before the light sensor of the camera
- Then the camera will be checked again at the 120 degree rotated side
- The conveyer moves the 240 degree turned bolt back before the camera
- The camera will then do a final check
- If still no damage have been found the bolt is classified as approved.
- The robot will pick up the bolt and put it in the approved output
For the new system it is required to use a Kawasaki robot FS03N robotic arm and a SICK IVC-3D camera. A new end-of-arm tool is designed to pick and place the bolts, this gripper has been made in a 3D-printer. The gripper can be seen on Figure 5. The IVC-3D camera makes highly detailed pictures off the surface area, an example picture can be found in Figure 4.
Figure 4: 3D profiles combined in an image.
Figure 5: The gripper.
After the 3D image have been made the camera is programmed to check on dimensions and damages on the bolt. The dimensions of the bolts are measured by finding the edges in the picture that the camera has taken. To calculate the diameter and length of the bolt, logically the left and right respectively bottom and top edged are detected. In Figure 6 and Figure 7 the detection of the left and right edges are shown.
Figure 6: Finding the right edge.
Figure 7: Finding the left edge
To check for the damage the blob feature is used, Figure 8. The program step called detailed extraction is used, Figure 9. In this step the 3D profile is checked for the points that are lower than the surrounding pixels, Figure 9. After this step the blob finder feature is used, Figure 8. In the blob finder the parameters for the blob that is consider to be damaged are given. If the blob is found with the given parameters then it’s known that the bolt is damaged. Figure 8 shows how the blob finder is setup. With this method the damage in the head and the flange can be detected quite reliably.
Figure 9: Detailed extraction.
Finding damage on the thread is really tricky. The camera have been able to detect damages in the thread but it’s not reliable all the time. Sometimes it sees damage where there is no damage. This problem is probably caused from the reflection of the laser within the thread and the method to find blobs (damage). In the blob finding method the program will look for the pixels that are lower than the profile in the surrounding. The thread does logically has a lower and a higher profile. Sometimes the blob finder will find a blob in that because of the height difference.
The camera is suitable to check on dimensions and damages on flat surfaces, but when the thread is checked for damages it is nog reliable because the laser scatters in between the thread. To get an explanation on this problem the SICK Company has been contacted, which then said the IVC-3D isn’t suitable for checking the thread. As a solution they recommended the IVC-3D30 camera, a newer and better version of the camera where have been made use of.