Seeing is better than sampling. (WIP).

The Place: The DaimlerChrysler (DCX) Warren Stamping Plant (Warren, MI).

The Application: Producing closure panels, including doors, hoods, lift gates, and tailgates.

The Issue: These are components that customers have direct contact with. They must be properly manufactured to meet safety requirements. There needs to be good seals produced on edges so as to keep materials (e.g., water, road salt) out of the inside of these components because rust-through warranty expenses are costly.

An Example: In process, pieces are stamped in the pressroom, then delivered to the assembly area. There, the outer closer panel is placed in a nest, where a 3-mm wide bead of hem adhesive is applied by a robot near the outside edge of the panel, adjacent to a flange that will be hemmed on the edge of the inner panel.

The Situation: They had been using manual inspection at the plant. An operator would pull a panel from the line (once an hour, once a shift, or at some other frequency). Because they are operating at cycle that produces as many as 240 panels per hour, it was possible too many components could be produced before errors were caught. In one instance, 3,000 faulty panels made it through to the final assembly line--an expensive miss. (A destructive test, wherein panels are cut apart, is still performed once per shift for purposes of checking adhesive wet-out, despite the change in method noted immediately below.)

The Solution: A vision system is deployed for 100% panel inspection. Once the robot has complexed its work, a programmable controller (PLC) triggers the system, a Passport DSL from PPT Vision (Eden Prairie, MN). Depending on the size of the panel, from four to 12 cameras are deployed. LED strobe lighting illuminates the part. Ambient light, excessive oil or other things that may cause a change to panel reflectivity are overcome. The vision system processes a pass/fail response within 0.5 seconds. If the Bead Evaluation And Detection System (BEAD) detects a problem, an alarm and tower light are activated; an image of the failed area is locked in place on the control panel; and a signal is sent to the line PLC to keep the panel from indexing to the next station. In order to get things back up and running, an operator must confirm the failure and hit a reject switch so that the problematic panel is removed from the line. Failed images and other data are stored for root-cause-analysis and process improvement.

Comment: "The cost savings have been enormous, one catch usually gives us our return on investment," said Dale Frendburg, Adhesive Process Specialist at Warren Stamping.

Future: There is the possibility that in the case of a reject, the process will be automatic (i.e., a failed part: rejected). The information from the system will be hooked into the plant-wide Factory Information System. They are looking at having the robot reapply adhesive where there are gaps, inspecting parts at the press line, and performing dimensional checks of panels. "There are endless possibilities for the use of machine vision," noted Cindy Renault, Quality Manager at DCX Warren Stamping.

Details: Write in 134 on the Reader Service Card.