Publications

Abstract not provided.

Abstract not provided.

AUTOmated GENeration of Control Programs for Robotic Welding of Ship Structure (AUTOGEN) is software that automates the planning and compiling of control programs for robotic welding of ship structure. The software works by evaluating computer representations of the ship design and the manufacturing plan. Based on this evaluation, AUTOGEN internally identifies and appropriately characterizes each weld. Then it constructs the robot motions necessary to accomplish the welds and determines for each the correct assignment of process control values. AUTOGEN generates these robot control programs completely without manual intervention or edits except to correct wrong or missing input data. Most ship structure assemblies are unique or at best manufactured only a few times. Accordingly, the high cost inherent in all previous methods of preparing complex control programs has made robot welding of ship structures economically unattractive to the U.S. shipbuilding industry. AUTOGEN eliminates the cost of creating robot control programs. With programming costs eliminated, capitalization of robots to weld ship structures becomes economically viable. Robot welding of ship structures will result in reduced ship costs, uniform product quality, and enhanced worker safety. Sandia National Laboratories and Northrop Grumman Ship Systems worked with the National Ship-building Research Program to develop a means of automated path and process generation for robotic welding. This effort resulted in the AUTOGEN program, which has successfully demonstrated automated path generation and robot control. Although the current implementation of AUTOGEN is optimized for welding applications, the path and process planning capability has applicability to a number of industrial applications, including painting, riveting, and adhesive delivery.

This paper documents work performed to convert scanned range data to CAD solid model representation. The work successfully developed surface fitting algorithms for quadric surfaces (e.g. plane, cone, cylinder, and sphere), and a segmentation algorithm based entirely on surface type, rather than on a differential metric like Gaussian curvature. Extraction of all CAD-required parameters for quadric surface representation was completed. Approximate face boundaries derived from the original point cloud were constructed. Work to extrapolate surfaces, compute exact edges and solid connectivity was begun, but left incomplete due to funding reductions. The surface fitting algorithms are robust in the face of noise and degenerate surface forms.