Paul L. Falkenstein, Certified LabVIEW Developer - Coleman Technologies Inc. (now Sciotex)
Michael D. Coleman, Ph.D. - Coleman Technologies, Inc.
Coleman Technologies, Inc. (CTI), now known as NI Partner Sciotex, was contracted by a major dinnerware manufacturer to provide a conveyor-based visual inspection system for detecting four types of defects occurring in many different styles of dinner plates and bowls. The manufacturer selected CTI based on the relatively low cost of our proposed system and our extensive experience in creating solutions for challenging machine vision applications.
Manufacturing dinner plates can introduce several types of defects. Warp is defined as the variability of the plate height around its circumference. Trim defects include both "trim gouge,” where there is an indentation at one or more angular positions, and “trim bulge,” where the circumference bows out from a perfect circle. The other defect, glass adhesion, occurs when small molten glass balls adhere to the top or bottom surfaces in the region adjacent to the plate rim.
The system requirements called for detecting each of the four defect types when the defect size is >250 micron (0.01 in.) for plates up to 12 in. in diameter. Figure 1 shows a highly magnified view of a plate with both trim bulge and glass adhesion defects that were detected and highlighted by PQIS. The system throughput is up to 60 plates per minute and inspection should run 24 hours per day, seven days per week, with minimal down time.
Figure 2 shows the PQIS layout. The system has three sections of 12 in.-wide conveyors positioned end to end with approximately 0.25 in. gaps between conveyors. All conveyors are driven by a single variable-speed drive capable of 65 ft/min.
Positioned below the first gap is a 4,096 pixel line-scan camera and lens that acquires images of the tops of the inverted ware shapes. Images are acquired via NI Vision Acquisition Software. Directly above the second gap is another 4,096 pixel line-scan camera and lens that acquires images of the plate bottoms. Both cameras are connected via Camera Link protocol to an NI PCIe-1430 dual Camera Link acquisition device. Each camera has an independent electronic photo sensor to trigger collection through a Camera Link I/O extension device.
Illumination at each gap between conveyors is provided by focused LED line lights angled off axis from the cameras. Using off-axis lighting results in a dark image background outside of the plates allowing for high edge contrasts. Direct on-axis illumination would cause blooming in the CCD sensor, destroying detailed information about the plate perimeter. Transmitting light through the translucent plates for illumination highlights the defects in the glass. Adhered glass scatters the transmitted light and appears as a change in light intensity. Advanced processing of the light intensities yields a defect detection of greater than 95 percent with no human intervention.
The plate warp is measured at another station on one of the conveyors. A set of red laser lines is used to create a laser line incident on the bottom surface of the inverted plate or bowl. Vertical displacements of this line are viewed in respect to a reference reading and are measured using an area scan camera. Triangulation is used to generate a point cloud model of the sample as it passes under the camera. The software corrects for lens perspective and laser alignment to produce models with 100 µm height tolerances and cross-section resolution of less than 150 µm. The two lasers can generate more than 250,000 points per second. Point clouds are analyzed to determine sample warp. Figure 3 shows a typical warp model generated by PQIS.
Rapidly Developing Advanced Image Analysis and Control Software With LabVIEW
We used LabVIEW and the Vision Development Module to develop the PQIS software. These tools helped us quickly develop the complex vision analysis routines required, as well as provide the users with an easy-to-use, informative user interface. Figure 4 shows a screen shot of the PQIS main application window.
It presents images of the top and the bottom of each plate, an interactive 3D plot of the plate rim warp measurement, and a summary table of cumulative run statistics including the number of failures of each type. As shown in Figure 1, the PQIS highlights the plate edge and adhesion defects found for each plate view.
System administrators can use the PQIS software to configure the pass/fail criteria for each defect. Additionally, the software provides utilities for system alignment and diagnostics. New plate designs can be defined through the provided editor and added to the system as needed. The system uses an NI PCI-6514 industrial digital I/O device to pass results and status to an external programmable logic controller that automatically sorts plates as they exit the PQIS. All results are stored in a database after each sample is analyzed. The database allows for concurrent process trending through SQL queries. Process trending and statistics are invaluable to both process engineering and quality assurance departments.
The first generation PQIS has been operating for more than a year and the manufacturer has already seen a return on their investment because we eliminated the need to manually inspect each plate and improved product quality and yields. We recently installed the second generation PQIS and validated its ability to detect defects with accuracy that is orders of magnitude superior to manual inspection. Developing the advanced image analysis routines employed in these systems would likely have been prohibitively expensive if we attempted to develop the software with any tools besides LabVIEW and the Vision Development Module.
Paul L. Falkenstein, Certified LabVIEW Developer
Coleman Technologies Inc. (now Sciotex)
6 Dickinson Dr, Unit 113
Chadds Ford, PA 19317
An NI Partner is a business entity independent from NI and has no agency or joint-venture relationship and does not form part of any business associations with NI.