QMT Features: June 2009
Inspection opens doors
An in-line 3D vision system provides automatic classification of door facings based on moulding design and surface texture.


Masonite is a global building products company with its Corporate Headquarters in Tampa, Florida, USA.  Masonite leads the world in manufacturing quality moulded door facings, which they supply direct to door-manufacturing partners across the world. Its Irish facility, in Carrick on Shannon, Co. Leitrim, is the result of a 138 million dollar investment and is the second largest of the Masonite group. It is a state-of-the-art production facility, and ships door facings to 32 countries.

Door facings with different patterns are produced simultaneously, and are required to be automatically classified and sorted into the correct collection bays at the end of the production lines. Incorrectly classified facings must be manually re-sorted. Manual re-sorting is very undesirable as it is highly labour intensive, resulting in damaged facings and subsequent yield loss.

Masonite Ireland had a legacy door facing inspection vision system that had served its original purpose well, but was becoming antiquated and unreliable. As it contained a library of all possible door facings, it could not easily accommodate the addition of new facings, requiring specialised programming by engineering personnel. Moving forward, Masonite was keen to apply a vision system solution that would meet their requirements for the foreseeable future. It was also critical that the system would be highly accurate, easy to use, quick for operations personnel to train, and could accommodate the large number of moulded door designs used in production.

Bring in the specialist
Masonite commissioned Alpha Vision Design, a specialist industrial machine vision company, to work alongside Masonite Ireland engineering personnel to deliver an in-line automatic classification system. The system was required to identify the door facings based on the moulding design and surface texture. Two door facings are pressed side by side, with the different door pattern combinations produced from a large stock of single door facing dies. There are thousands of possible die combinations, so the sorting system was required to accommodate multiple die combinations and frequent die changes.

At the beginning of the production run, the operator would train the system for the first set of facings using a simple training wizard. Subsequent facings would then be automatically inspected and classified. The vision system would communicate with the production line’s control system, which would route the classified facing to the appropriate collection bay.

The project posed some interesting technical challenges. Due to the large surface area involved (one door length x two door widths), providing even illumination over the entire surface area was difficult to achieve.  Due to the physical attributes of the door facing (large area, not perfectly flat), conventional lighting was inadequate to provide good quality images for the vision system. Also, a robust system was required that could deal with rotation and some bending or warping of  the door facings, as the facings can sag where there is no direct support from the conveyor belt. As a result, the pattern profile of the door facing can appear distorted, which typically causes stretching or compression within the acquired image. Standard pattern matching tools based on cross correlation techniques would perform poorly under these circumstances.

The solution
Globally, the vast majority of machine vision applications are based on 2D inspection techniques. 3D inspection still remains relatively specialized although new markets are emerging as the technology matures and gains more widespread acceptance. 3D inspection using laser triangulation is probably the most common platform used today for deploying 3D applications (although many other 3D technologies exist). Lasers, used for illumination, offer cost effective solutions and are readily available from many suppliers and manufacturers. In addition, laser based 3D systems are particularly effective for scanning large non-reflective parts with well defined profiles and were therefore applicable for use in this vision application.

The system is designed to capture 3D data of the moulded door facing’s patterns, using diode laser line generators and a line-scan 3D image acquisition camera (supplied by other equipment manufacturers). The laser line generators produce a visible laser line across the width of the double door facing as it travels along the inspection conveyor. Using a shaft encoder, the 3D camera is synchronised with the conveyor speed to capture 3D data at a suitable resolution along the length of the door facing. The 3D camera is configured for a similar pixel resolution for the facing width.

The height sensitive image data is transferred to the PC’s framegrabber via a CameraLink interface to build and create a 2D height profile image (in bitmap format) of the double door facing being inspected. The image is then processed using powerful VisionPro software from Cognex. The advanced PatMax tool provides the ability to perform unique geometric pattern matching and to classify shapes, even at low contrast, as encountered in this application.

“Smart” 2D vision cameras and halogen lighting are used to provide the required surface texture information, which is integrated into the vision system using digital input signals.

Once Masonite Ireland engineering personnel had completed the specification, design, installation and commissioning of the 3D & 2D vision hardware, the system was ready to be programmed to extract and process the image information. A custom Graphical User Interface (GUI) was created by AVD according to Masonite’s requirements. The interface was designed for ease-of-use by the operations personnel. It was designed to be uncluttered, and contains a colour coding scheme consistent with other systems in place at the manufacturing facility.

As part of the ease-of-use ethos, the number of operational settings required to use the system were kept to a minimum. An over-complicated engineering tool that would be limited to programming / configuration by technical or engineering personnel only was avoided. The delivered system requires less than 4 minutes for production personnel to train for new door facings, and requires little more than the ability to draw boxes and position detection tools.

Robert Sadlier: Senior Controls Engineer, Masonite Ireland said, “The advanced functionality of the Cognex VisionPro software contributed significantly to the project’s success, allowing the system to exceed the original User Requirement Specification, providing a sorting accuracy in excess of 99.97% (i.e. less than 3 in 10,000 incorrectly classified). Manual re-sorting of product has virtually been eliminated, increasing yield, and allowing operations personnel to perform tasks that add value to the product.”

www.avd.ie.
http://www.cognex.com.
  
You can now view all QMT Magazine issues on your favourite tablet or smart phone.
Download the free Quality Manufacturing Today App from the Apple iTunes App Store or from QMT Magazine on Google Play.

Rob Tremain Photographer
www.4exposure.co.uk
slideShow
Click above to see full page display and links to QMT articles.
Untitled
Bowers logo
TCT Inspex 2016 logo
Prodim logoi
Hexagon Logo 2
Control logo
Creaform logo
Vision 2016 ad
Nikon logo