Overhead car console assembly verification
VISION SENSORS FACILITATE “BUILD-IN-SEQUENCE” STRATEGY FOR HEADLINERS
In-line checks confirm presence, position, color and orientation of parts for JIT assembly of X400 Jaguar headliners
Summary: Cognex In-Sight® 1000C vision systems are being used in a highly integrated assembly line making headliners for the X400 Jaguar car. Johnson Controls Automotive has installed the line to meet the demands of a new contract, and a special feature is that automated vision systems are used as an integral part of the assembly process, checking for the presence, position and orientation of components such as grab handles, labels and foam spacers. Electrical checks confirm that console lamps are working properly and customer color choices are also confirmed correct before shipment. Headliners are built-in-sequence to JIT production demands from the Jaguar line, so the line is highly time- and quality-critical. Image archiving supports full traceability.
The new X400 Jaguar headliner line is at the JCA Speke plant on Merseyside. Headliners (interior linings for the roof) are assembled in response to a 'broadcast' call from Jaguar, which manifests itself at the JCA factory as a sequence of bar code labels, which automatically supervise the assembly line. The task is to assemble the headliners in the same sequence as the cars they belong to, ensuring that color schemes, fittings and various optional extras match exactly the customer's order. The aim is for finished headliners to be transported to the Jaguar line 'Just in Time' (JIT) to be installed in the correct car as it arrives at the assembly point. A production rate of 40 headliners per hour is targeted, which means that each headliner takes approximately 90 seconds to complete.
"This 'build in sequence' strategy poses critical production challenges," says Nick Bradburn, Manufacturing Engineer. "The starting point for us was the need to replace an old line which relied on mechanically-positioned optical sensors for checking the product and which therefore needed lots of maintenance." JCA was seeking to improve their delivery integrity and build quality in keeping with their continuous improvement policy and their aim to exceed customer expectations. This would also handle the increased policy warranty responsibilities of the car industry and underwrite the vital 'JIT Sequenced Environment'.
The challenge is 'rigorous' according to Bradburn. "We only assemble two types of headliner - one for saloons and the other for estates - but the task is complicated by the multiple choices of colors and fittings (Intrusion Sensor, Park Aid, Moonroof etc.) available to the customer. When other variants are included - such as international differences in the Garage Door Opener (GDO) and SRX options frequencies - it means that hundreds of headliner variants are possible - all of which must be proved correct and tested in the correct build sequence at high production rates."
The assembly/test process
The first two tables of the new 4-table assembly line concentrate on adding parts such as locator clips, visors, foam blocks and some electrical items to both sides of the headliner substrate -the hidden (or B) side first and then the visible interior (or A) side next. At each table the bar code label is scanned first and the correct parts are delivered using a KANBAN system. The processes here are purely manual with no vision checking.
Tables 3 and 4 differ in having overhead gantries about 3 meters high above each table, on which four Cognex In-Sight color vision sensors connected via Ethernet in a Vision Area Network (VAN) look vertically down on the headliner. Here, scanning the bar code also tells the inspection system what test parameters to use and instructions are passed serially to a master camera in each VAN, which then passes on the data to the other three cameras over the Ethernet.
Each Cognex In-Sight color camera is a full-featured CCD camera with on-board processing capability that can undertake its own set of 'inspection' tasks separately from the central controller. The 'C' indicates that it is fitted with additional hardware to enable the Hue, Saturation and Intensity of its target to be assessed, allowing it to confirm that the correct color parts are fitted. The VAN is also connected to a web server that transmits captured images to a central storage and archiving facility.
In detail, the headliner is placed on Table 3 with its (hidden) Side B uppermost. First, the presence and orientation of foam blocks is confirmed, then the locator clips. The grab handle spacers are checked to ensure there are zinc on the front and plastic on the rear as well as whether or not there is a Moonroof.
The major item fitted on Table 3 is the Master Plan Console (MPC) - the module above center of the driver's head in the finished car, which contains courtesy lamps, switches, microphone, GDO and similar items. Once again there are many different configurations, so the correct MPC is selected by scanning the next kanban part, the match between headliner and MPC is then confirmed / declined on the HMI. Once the correct MPC is in fitted, the overhead Cognex cameras are triggered and the position, orientation and content of the MPC are checked, with the results showing on a nearby monitor.
Once a 'pass' has been confirmed, the headliner is placed on Table 4 with its (visible) Side A uppermost. Scanning the bar code label tells the test system whether the headliner is for USA, Canada or the rest of the world and gives correct options for the `eCheck', which appear as a series of bulleted check points on a nearby monitor screen. The first job is to plug the main wiring loom into a fixed test system that checks continuity of the electrical items. A 'green' bullet indicates a pass and the operator next triggers the vision system manually to look for the presence of 'lamps' in the off state, and to perform some other checks including looking for the correct air bag label on the underside of the visor. The Cognex cameras are automatically re-configured with a smaller aperture during this test because of the label's reflectivity.
Once the label test has generated a 'green' bullet, an 'eCheck' is performed to prove the functionality of the MPC. These tests include turning all lights 'on' and allowing the vision system to compare this image with the previous 'lamps off' image to prove that lamps are working. The GDO, Intrusion Sensor, microphone, Moonroof, Park Aid and SRX are also tested, by manually activating them. A green 'bullet' confirms the correct operation of all items and this allows the operator to proceed to the first full vision check. At this point the cameras switch to a larger aperture so that their color sensitivity is greater, to allow the color of the unit, grab handles and visors to be confirmed, also that grab handles are correctly placed front and rear. Once all green 'bullets' are lit, a pass stamp is attached to the B Side and the operator knows that the Head Liner is ready to be shipped to the Jaguar plant.
The entire assembly/test process including the workstations and tables are under PLC control. Once triggered, the four cameras capture a full image of the headliner in less than 5 mS, with each camera 'responsible' for one quadrant.
As with all camera-based inspection systems, good illumination is extremely important. In the JCA installation, three fluorescent strips, interspersed by the four cameras, illuminate the work piece from above. These strips are high-frequency fluorescents to give best color sensitivity and to avoid interference 'flicker' lines developing in the image (the camera and strips operate at 50Hz and 60Hz respectively), which would affect threshold counts and distort the results. Further benefits of high-frequency strips are that they reach working temperature much faster and they tend to fail immediately, which avoids slowly deteriorating image captures.
Programming the cameras to look for the hundreds of combinations of product was straightforward because of the simple, spreadsheet approach use in configuring each Cognex Insight vision sensor. "It's just a matter of specifying step-by-step what you want each system to do," says Bradburn. "For position and orientation tests you mark the area you want using the relevant on-screen tool and then you configure the parameters needed to identify that shape or its content. The Cognex approach makes this very easy. For the color tests, you need to be sure that the system can distinguish that color from any others it may encounter - which requires some experimentation - but after that's done it's straightforward to 'see' the wrong color items and flag them as mistakes." Bradburn expects to refine the tests continuously to make this even more effective over time - which is easily done because of the simple programming technique.
As each set of images is captured for analysis, it is transmitted over the factory Ethernet network to a dedicated server, where it is stored for future use. "This means that we now have 100% traceability," says Bradburn," so if any quality concerns arise we can go back to the captured images of the headliner to confirm root cause. Similarly, we can use the archive to double-check on any production issues."
Each camera is equipped with Cognex's full library of image processing and grey-scale analysis tools, including PatFind®, a supertool for locating parts. Programming is done from a remote operator station using a 'spreadsheet' approach that enables each step in the vision process to be specified in a simple way. Combining the flexibility of a programming language with point-and-click simplicity, the vision spreadsheet enables the right tool to be selected for each part of the process. To make life even easier, the vision spreadsheet is transparent, so the image of the part being checked is visible during set up. This allows the user to see vision tools being applied to the image without having to switch between screens.
Parameters are chosen from drop-down menus, and the results of each test are automatically inserted into adjacent 'cells' in the spreadsheet. Cells can easily be linked together to perform required tasks, making 'proof of concept' prototyping much faster and enabling applications to be modified 'on the fly' during product changeovers. The interface also provides context-sensitive help in various languages.
"Machine vision operations are based on data," says Cognex, "and spreadsheets offer one of the most robust, efficient means of working with large data sets. Also, spreadsheets can include hundreds of specialized functions, options and operations, ranging from advanced mathematics to annotated graphics. Therefore, they have a great deal of built-in capability. Spreadsheets also provide a familiar work environment, which helps minimize the learning curve for inexperienced users and eliminates the need to learn a programming language."