The use of sealants and adhesives is increasing as automotive designers try to improve fuel efficiency by reducing vehicle weight. Safety, aesthetics and durability are all critical factors and, when taken together, promote greater use of structural adhesives — such as epoxies and polyurethanes — than of mechanical fasteners. Adhesives and sealants also are used to make cars quieter, better insulated and resistant to corrosion.With adhesives/sealants being critical to automobile assembly, it is important to deliver the appropriate amount of mastic material to the correct location every time. In the past, engineers designed fixtures to hold each part to be sealed in a precise location, but this was expensive and did not provide flexibility.When machine vision was introduced, systems were large and could be used only in fixed locations. As a result, the field of view was limited, and the systems were not flexible enough to handle designs with multiple parts. Today, smaller hardware and software that require less lighting and fewer cameras have led to smaller vision systems. Yet one challenge for the automotive manufacturing industry is to create flexible and efficient systems that can produce multiple vehicle styles on the same line, that enable line changeovers on the fly and that operate at minimal cost. New vision technologies are beginning to revolutionize the field of robotic dispensing, moving automakers much closer to the goal of flexible manufacturing.Traditional vision technologiesTraditionally, vision systems for robotic sealant/adhesive dispensing have consisted of two or more three-dimensional cameras that locate the part to be sealed, with more cameras added to handle large parts or those with many subcomponents. Such systems also required associated lights, cables and protection shields. The weight of the vision systems prohibited them from being mounted on the robot, as did the fact that many such robots have to navigate into tight corners of the substrate. Thus, the vision cameras were fixed on the floor or on some rigid structure (Figure 1). Figure 1. Four floor-mounted cameras and four lights are used to locate a vehicle floor pan.Traditional vision systems can’t “see” diverse styles of substrates because of this rigid mounting and the cameras’ limited fields of view. The cameras are fixed at a location with a clear view of features of the part to be dispensed and, depending on the size of the part, two or more may be required. If the features of interest are outside the field of view of the cameras, more cameras must be added. In some North American auto plants, more than 20 cameras are required in a single robot station to handle multiple vehicle styles (Figure 2). Initial and operating costs are high for such a complex system.Figure 2. Multiple floor-mounted cameras are used for processing multiple vehicle styles.Another commonly used method for locating substrates is laser triangulation, whereby a laser locates the part to be sealed. Lasers are small enough to be mounted on a robot and flexible enough to handle various substrate styles because the robot can move lasers to the points of interest. However, lasers have some disadvantages.Because of limited field of view, they typically must make a “search pattern” to locate enough features for a true 3-D calculation. The search causes longer process cycle times, which are not desirable in automotive manufacturing. Another disadvantage involves safety. Lasers used for locating features are hazardous to the eye, and the area of the manufacturing facility that houses the laser must be isolated to protect workers’ eyes — resulting in additional cost and inconvenience.Figure 3. Smaller cameras can be mounted on a robot to locate parts before they are sealed.Newer technologiesNew technologies are changing the way robotic vision is used in automotive manufacturing. For example, TrueView Single-Camera 3-D from ABB Inc. of Auburn Hills, Mich., uses only one conventional CCD camera to gather orientation for all six degrees of movement. Efforts to miniaturize the hardware have resulted in an extremely small and lightweight package that can be safely carried by the robot along with tools associated with sealant/adhesive dispensing. The vision system’s software processes in less than one second, enabling fully automated recalibration in less than five minutes. This speed greatly eases system installation and startup, and single one-button recalibration minimizes the challenges of machine vision applications.Figure 4. Robots can reach into small spaces, with vision systems guiding them to dispense sealant to the proper area.The small camera allows for mounting the vision right on the same tool as the dispensing applicator. The robot can then handle different substrates simply by moving the camera around, and the small package enables the sealant/adhesive applicator to reach through small spaces such as automobile windows to work on interior seam sealing (Figure 4).The vision system can locate parts that do not have clear features, such as a roof without holes, bolts or lines. To do this, a low-power laser pattern is projected on the part (Figure 5). Automobiles have many such parts, so this capability has opened up new applications in sealing and bonding.Figure 5. Using a vision system, an 8-mm-wide applicator is inserted into the 12-mm-wide roof ditch to seal the seam between the roof panel and the body side panel of the vehicle.The camera obtains full 3-D information using one shot from one camera by looking at multiple features within the shot (about five features is the norm). By comparing the features’ sizes and relationship with one another, it obtains the positional data of the part of interest. By comparing the shape of the features — for example, a circular feature that now appears as an oval, it determines the angular data of the part. Because there is only one camera, data crunching and operator setup are simpler than with standard “stereo pair” sensors. The disadvantage is that it assumes that the features are always identical. So if the camera is dependent on a circular hole for identification and one part receives one coat of paint while another receives five, the hole dimensions will change depending on coating thickness, causing the camera to misinterpret the data.Real-time inspectionFuture advances in vision systems will enable real-time guidance and inspection.The robot must wait up to six seconds after the application part is presented for vision to perform its guidance functions. The ideal situation would be for guidance to happen in real time, as the robot is working on the part. With more sophisticated software and ever-increasing processor speeds, realization of this goal is probably only three to five years away. Once it is accomplished, robotic automation productivity will improve dramatically. Today, robot inspection of the sealant/adhesive applied is not as popular as guidance because of the time requirements and the difficulty in measuring the location and amount of the mastic material. Aggressive research is being done in this area and, in five to eight years, real-time vision inspection of sealants and adhesives will finally become a reality.The small, robot-mountable true 3-D vision systems have opened up opportunities in robot dispensing. In the near future, however, automakers will concentrate on using real-time vision guidance for dispensing sealant and adhesives precisely to a part and then inspecting it in real time, dramatically increasing productivity and quality.Meet the authorTsunou Chang is sealer/adhesive technology manager at ABB Inc. in Auburn Hills, Mich.; e-mail: email@example.com.