Vision-Guided Robot Automates Machining Process
Foundry automation challenges
Some of the most difficult, dirty, and dangerous jobs are located in foundries, yet foundries have been relatively slow to embrace automation. One reason is that they often manufacture a variety of parts which increases the difficulty of implementing automation. Another reason is that parts typically come out of the casting process with relatively large variations in size and geometry, making it difficult to locate them using conventional, hard automation methods.
In this particular application, clutch housings for over-the-road trucks are cast in eight different variants. Each casting has a large sprue that needs to be removed and a stainless steel filtration screen embedded in the sprue that needs to be separated because it cannot be re-melted.
This application would have been difficult to automate with traditional machine vision systems because they rely upon analyzing pixel-grid values, a process known as correlation. This method locates objects by comparing a grey-level model, or reference image, of the object to the image. Typically, when the object changes in rotation or scale, the vision system can no longer detect the object. But, recent innovations in vision technology have made it possible for robots to automate operations with greater amounts of variability than could be accommodated in the past. Odyssey Machine Company worked with RAF Automation, a Cognex distributor based in Solon, Ohio, to develop a flexible automation system capable of handling variations in pan type and location. The key to the automation system is its ability to immediately identify the pan type and define its precise location on the infeed conveyor that brings castings to the machining workcell.
Object location technology key to success
Odyssey Machine Company and RAF Automation based their solution on Cognex's PatMax® software technology for accurate location of objects despite changes in object angle, size, and appearance. PatMax uses geometric information in place of pixel grid-based correlation by applying a three-step geometric measurement process to an object. PatMax first identifies and isolates the key individual features within an object image and measures characteristics such as shape, dimensions, angle, arcs, and shading. It then correlates the spatial relationships between the key features of the trained image to the runtime image, encompassing both distance and relative angle. By analyzing the geometric information from both the features and spatial relationships, PatMax is able to precisely and repeatably determine the object's position without regard to the object's angle, size, or appearance.
The solution used the traditional flat-belt infeed conveyor, which was modified by Odyssey to allow the parts to move consistently into the vision cell. The castings are placed with the sprue face down on the conveyor, which holds their y axis (perpendicular to the direction of the travel of the conveyor) position within about ½-inch. The eight different castings each require different robot programs. The first critical step is to identify the part number of the next casting coming down the conveyor in random order. Odyssey Machine selected the Cognex In-Sight Micro 1100 vision system for this task. "We picked the In-Sight Series first and foremost because of Cognex's unique PatMax object location technology," LeRoux said. "In addition to PatMax, the In-Sight Micro 1100 provides other Cognex tools in the smallest package available, making it ideal for mounting in very tight spaces."
Communicating with robot and PLC
The robot used in this application is a Fanuc S900W that the customer had previously retired, but was rebuilt by Fanuc for this application. The workcell runs under the control of an Allen-Bradley programmable logic controller (PLC) which is connected to the vision system with an Ethernet cable. The firmware of the Cognex vision system contains the protocols needed to communicate with each popular type of PLC including Allen-Bradley, Siemens, Mitsubishi, and more. This communications suite is known as Cognex Connect™ and is a standard capability for all In-Sight vision systems. This vision system is connected to the RJ2 robot control with digital input/outputs to communicate with the robot controller. The In-Sight Micro vision system sends the part identification information and the coordinate information on the part to the PLC which sends information directly to the robot controller.
Odyssey Machine Company engineers taught the vision system to recognize each individual part by shooting a reference image (stored in memory) of the part on the conveyor. When a new part comes down the line, a standard conveyor photo-eye detects the presence of the part, stops the conveyor and triggers the vision system to capture an image. The In-Sight Micro system compares the geometric information to the reference images stored in memory, selects the correct pan number and calibrates the match with each image as a percentage score. "We have yet to see a situation where the score was less than 95% for the correct part or more than 65% for the wrong part," LeRoux said. The PatMax algorithm also provides the location of the part, including the x axis and y axis (in the direction of conveyor travel) coordinates and theta, the angle of rotation in the x-y plane.
Calibrating the robot and vision system
The robot's coordinate system needs to be synchronized to the vision system, but only once unless the conveyer is moved in relation to the vision system Odyssey Machine Company engineers calibrated the robot's coordinate system by placing physical registers on the conveyer. They captured an image of the registers and the vision system determined the coordinates of each register in its own coordinate system. The engineers then jogged the robot to each physical register and noted its position in the robot's coordinate system. Then they determined the offset between the two coordinate systems and adjusted the robot control so that it uses the same coordinate system as the In-Sight Micro system.
Next, the robot moves to the location of the part, twists its wrist to the correct angle and picks up the part. The robot places the part in a fixture on a three-stage linear saw with carbide-tipped blades. The first stage of the saw cuts off the end of the sprue just below the stainless steel filter and drops the cut end into a meltable scrap bin. The second stage cuts off the stainless steel filter and drops it into a different bin from which the filter is recovered. The third stage cuts off the rest of the sprue and also drops it into the meltable scrap bin. The removal of the sprue leaves a hole in the part.
The process continues as the robot picks the part out of the fixture, by the hole, and moves it to the second fixture in the workcell. This fixture uses a compliant air spindle from ATI Automation that drives a high-speed steel helical end mill at 30,000 rpm. The robot moves the part around the end mill which removes the flash. The radial compliance of the spindle compensates for differences in the amount of flash on each part.
"The new robot workcell has successfully automated what used to be a difficult manual task with the danger of repetitive stress injury," LeRoux concluded. "The Cognex 1100 Micro vision system flawlessly identifies the part number and the x, y and theta coordinates of the part. The robotic workcell eliminates two of the three people previously required to operate the workcell. One person is still required to load the finished castings onto a rack. The entire cell costs about $400,000 including rebuilding the robot, the cost of the vision system and the metal cutting equipment. The automation portion of the cell is about $75,000. The total cost of employing a person in a foundry is also about $75,000 so by freeing up two people the robot quickly pays for itself."