Columns

AIA:Robotic Dispensing Puts Hush in SUV

The Durango HB sport utility vehicle from DaimlerChrysler (Auburn Hills, MI) is built with acoustical dampening foam in several body cavities to lesson road and engine noise. Typically, this foam is dispensed manually, using a heavy dispensing gun placed against access holes to the insides of doors, fenders and other hollow regions.

To increase throughput and ensure correct foam dispensing, DaimlerChrysler installed a paint-shop-type robotic workcell that uses three articulated IRB 4400 L-i10 robots equipped with S4Cplus+ robot controllers from ABB Inc. (Auburn Hills, MI). The cell also uses VIAMES integrated image processing controllers and 12 dual-camera, location measurement stereovision sensors from ISRA Vision Systems Inc. (Lansing, MI) to provide vision guidance for the robots. The stereovision sensors come with a halogen stripe light, composed of a halogen light and optical grating. These stripes provide additional image features, which enable the vision system to guide the robots even when manipulating highly reflective parts, such as car bodies.

The Durango HB is available in two body styles, both with four doors and a tailgate, but only one with a sunroof. During assembly, the vehicle body comes down the production line via high-speed belt transfer, until it reaches a skid stop station in the foam dispensing workcell. The skid supporting the truck body is lifted hydraulically into place, and a programmable logic controller (PLC) sends a start message to the machine vision controller along with an identification code telling the vision system which model type is in the workcell. Using this identification data, the image processing controllers know where to direct the robots for the particular model in the cell.

To help direct the robots, the image processing controllers use normalized correlation algorithms to identify the access holes, with the light stripes creating additional features for 3D data extraction on the curved, reflective body panels. DaimlerChrysler uses normalized correlation, as opposed to geometric pattern matching, because it works better in changing plant lighting conditions or when there is substantial transient "noise" caused by debris and dust. It is also faster-an important consideration when linking a number of cameras.

When enough features have been located, the machine vision controller calculates the offsets between similar features in separate images from each of the cameras. This data then provides the basis for a triangulation function that generates 3D data on the location of the truck panels within six degrees of freedom. The robot positions itself to a zero point in rough alignment to the truck panel. Then the vision controller provides offset coordinates so it can guide a GX15 foam dispensing gun from Gusmer Corp. (Lakewood, NJ) directly to the hole.

Each hole is 12 millimeters in diameter, and the dispensing nozzle tip is 11 millimeters in diameter, so the system has to locate the gun within 0.5 millimeter to make a good fit. The dispensing gun nozzle is spring loaded with a force-feedback sensor tied into the robot controller, so a tight seal can be verified before dispensing the foam into the cavity. This cuts down on spillage and waste.

The gun also has a meter that monitors the amount of foam being dispensed, with the controller telling the gun how much to dispense for the specific body cavity being filled. Once the correct amount has been dispensed, the robot withdraws the nozzle, and the controller queries the next sensor, which provides coordinates for the next hole. After each car is completed, the PLC directs the robot to clean the gun tip at a nearby tip clean station.

Looking for quick answers on assembly and manufacturing topics? Try Ask ASM, our new smart AI search tool. Ask ASM