Quality control and quality assurance may soon be synonymous with machine vision. Demand for error-free assembly keeps rising as the pressure on production time mounts. In our Six Sigma/lean industrial world, there is less and less room for mistakes. And many companies now turn to the greater flexibility and lower price points of machine vision as quality solutions.
The automotive and electronics industries pioneered the use of machine vision for inspection and verification. "For electronics assembly, machine vision plays a role in almost every step of the assembly process," notes Kyle Voosen, machine vision product manager at National Instruments. "Here the rule of 10 applies, where it costs 10 times more to correct an assembly error at each subsequent step of a manufacturing line. So, if a socket is placed on top of bad solder, but the problem isn''t found until ICT on the actual chip, then the correction will be 100 times more expensive than if the poor solder paste was detected at the very beginning." (For more, click here.)
Component Positioning
And machine vision continues to spread throughout the automotive sector. VW/Audi recently commissioned a DVT system in Brazil to "inspect the marriage of the car body with chassis," says DVT media relations manager Richard Daigle. The incorrect positioning of components may damage the car body, and in some cases tolerances run as low as 2 mm.
The machine vision solution does not require a PC on the line, is Profibus-linked, and includes technical support and free software and upgrades. Three DVT Legend 530 cameras inspect pins for positioning of the chassis over the car body, at the front, middle, and rear, with wide-angle lenses. The front camera inspects two shock absorbers and the transmission support, the motor support, and two guide pins; the middle camera inspects gearshift position; and the rear camera inspects four pins.
Up to seven models can go down the line on any given day, each requiring different positions. Model changes are communicated through Profibus, which pulls up the VW-determined product file matching the car body present via RF. The camera then inspects the positions of the pin and manual transmission gearshift, at line speed. Assembly will not proceed down the line unless cameras permit. Output averages 700 cars per day, 24 hours a day, three shifts, five-six days a week. (For more, click here.)
Pattern Matching
Wineman Technology used National Instruments'' LabView to help an automotive fabricator identify aluminum wheel rims for sorting. "If the wrong wheel is placed into a lathe," explains NI''s Voosen, "$50,000 of damage to the machine and possible human injuries are possible."
Wineman project engineer Tara Crower designed a system "utilizing Vision Builder''s pattern-matching abilities" to sort different wheel rim types from the caster and "send them to the proper assembly line for machining. Any errors that would cause a wheel to go down the wrong lane would result in catastrophic damage to machine tooling." With the operation''s high volume, the cycle time for ID " had to be less than five seconds." And the system had to "interact with a PLC and another PC controlling the robot for part transfer."
Wineman developed a two-camera system -- one camera positioned to monitor the top of the wheel, the other viewing a profile of the wheel from the side. Continues Crower: "Because of the similarity in many wheel types, proper lighting was very important for this system so that the pattern-matching algorithm in LabVIEW could identify the smallest detail on the wheel. High frequency fluorescent lights were used to illuminate the top view of the wheel. The side view of the wheel was illuminated using an LED backlight. A standalone PC cabinet was located away from the vision stand, which controls the entire system. The PC contains a National Instruments PCI-1409 image acquisition card."
The software requires minimal operator interaction. "The main screen is a production screen that processes the parts as they enter the station," Crower explains. "A picture of the part currently in the stand is displayed on the screen along with a picture of the part it has been identified as. Using LabVIEW''s vision tools for pattern matching, the parts are analyzed. Rims are first sorted by size to reduce the number of parts checked for pattern matching. This helps to reduce the cycle time for each part. Parts are added very easily with the part setup wizard. The user is taken through a step-by-step process to add a part. Three different regions of interest are set up for each wheel rim. These three regions are then compared to identify the rim in the vision stand. Communication between the PLC and the PC is accomplished using HighwayVIEW."
The solution significantly improved quality and efficiency. ID accuracy increased from 70% to 99%. "Many wheel rims with flash that were previously unidentified can now be identified," adds Crower. "Wheels with an excessive amount of flash may not be identified and will be sent down an escape lane. The previous system required three hours plus to add a new wheel to the system. With the new system a wheel can be added in less than five minutes." (For more, click here.)
Visual Inspection
Spanish Integrator Ikusmen Vision Artificial installed a Cognex visual inspection system for Spain''s Dinalot, a manufacturer of small machine parts for motors. The system replaces manual inspection, with a process completed in nine seconds.
The system includes two Cognex In-Sight 1000 vision sensors, an XY Cartesian robot and a rack-mounted industrial PC. The PC configures the vision sensors; controls the sensors and the robot; and displays the inspection images and measurements.
Parts arrive at the inspection station from a cleaning bath, are picked up by the robot for inspection, and placed on the control surface. The first sensor locates the central notch on the part and calculates its position. The robot then moves the camera so that it is above the nominal position of each hole, and the sensor calculates the angle of each hole relative to the central notch and measures their diameters, to a level of precision of +/- 0.02 mm for the diameters and +/- 0.1° for the angles. The second sensor inspects the quality of the threading. Cognex''s solution provides 100% inline inspection of parts. (For more, click here.)