Robot vision system locates radioactive pucks

May 1, 2002
The end of the Cold War and the subsequent downsizing of nuclear arsenals led to the existence of surplus plutonium, which must be disposed of so that it is inaccessible and unattractive for weapons use.

Andrew Wilson
[email protected]

The end of the Cold War and the subsequent downsizing of nuclear arsenals led to the existence of surplus plutonium, which must be disposed of so that it is inaccessible and unattractive for weapons use. At the US Department of Energy Savannah River site (Aiken, SC), one of the approaches that has been studied involves the immobilization of plutonium within titanate-based ceramic forms.

At the Westinghouse Savannah River Co., engineers developed a prototype machine-vision-based robotics system as part of a system to immobilize plutonium in titanate-based ceramic pucks (left). During operation, images of the pucks are imaged, thresholded, and related to two-dimensional Cartesian coordinates (right). The data determine the single location of a puck for robotic placement into a canister.
Click here to enlarge image

These ceramic "pucks" are placed in sealed stainless-steel cans and put into long cylindrical magazines. Then the magazines are latched to racks inside canisters and filled with a mixture of high-level radioactive waste and molten glass. Once immobilized, the pucks are transported in batches using a linear transport system for loading into the stainless-steel cans. Due to radiation hazards, puck-location uncertainty, and accuracy requirements, robotics provides an ideal can-loading method. A vision system obtains puck-location information and improves accuracy during loading.

To provide the can-loading robot with the locations of the pucks and to assist in puck alignment prior to can-loading, Eric Kriikku and his engineering colleagues at the Savannah River Technology Center, the site's applied-research-and-development laboratory, developed a prototype robotics vision system. This system connects two RS-170, 640 X 480-pixel, gray-scale CCD cameras from Watec (Las Vegas, NV) to a Pulsar frame grabber from Matrox (Dorval, QC, Canada). The frame grabber is housed in a 333-MHz PC from Dell Computer (Round Rock, TX) running Windows 98 software.

Located above the robot, the first camera's field of view (FOV) reveals all the pucks on a transfer tray simultaneously. Outline and center markings are superimposed by the image-processing software to indicate successful puck identification. Custom image-processing software, written in C++ and running on the host PC, updates the image 15 times per second. Data obtained allow coarse puck locations to be determined that are accurate enough to position a 2-in. suction cup near the center of a 2.65-in. puck. Image coordinates from this camera are related to robot coordinates using a vision-system calibration step.

Located 8 in. from the can, the second camera looks upward to obtain high-accuracy puck location with respect to the robot suction-cup center. Once the robot receives the puck location (using data from the first camera), it picks up the puck and positions it over the second camera. Because the second camera's FOV is slightly larger than a single puck, location is established within ±0.008 in. With these data, the vision system measures the offset between the puck center and the suction-cup center and instructs the robot to adjust its position to make up the offset.

In the pick-and-place system, a robot and a robot-control system from Motionex (now part of Cross Automation; Gastonia, NC) use three dc servomotors, each controlled by a dc controller from Emerson Motion Control (Eden Prairie, MN). To provide the data to position the dc motors, positioning data are sent from the vision system over a serial interface to a separate industrial PC from Xycom Automation (Saline, MI) running Windows NT software.

Under Windows NT, the industrial PC uses SteepleChase Visual Logic Controller (VLC) software from Entivity (Ann Arbor, MI) to program both the control logic and the robot motion. After serial positioning data are received from the vision computer over an RS-232 serial interface, data are sent from an XMP series PCI-based multiaxis control card from Motion Engineering (Santa Barbara, CA). To provide the user interface and motion-control functions, Kriikku combined the VLC software with the Citect ToolBox from Ci Technologies (Charlotte, NC), which interfaces with the other tasks to execute commands and refresh screen data.

Originally constructed during the early 1950s to produce tritium and plutonium-239 used in the fabrication of nuclear weapons, the Savannah River site now processes and stores nuclear materials in support of national defense and nonproliferation efforts, while developing and deploying technologies to improve the environment and treat wastes. Westinghouse Savannah River Co. operates the 310-sq mile site for the Department of Energy.

Voice Your Opinion

To join the conversation, and become an exclusive member of Vision Systems Design, create an account today!