Original Date: 04/24/2007
Revision Date: / /

AutoGen is a software tool that automates the planning and execution of control programs for robotic welding of ship structures. This capability will dramatically improve the ability of welding robots to play a major role in ship construction, improving quality and productivity.

One of the challenges facing the U.S. Navy is achieving desired fleet size and readiness, which is partially due to the high cost of manufacturing warships required for today’s complex designs. The most common process in shipbuilding is welding. Stiffeners are welded to plate steel to form panels. Panels are assembled into subunits; and subunits are combined into larger units to form a complete ship. Welding is important at every step. Mechanized systems play a role in the panel phase, but robots are required beyond that. Few robots are used in U.S. shipbuilding because of accuracy requirements, installation cost, and personnel training. The major obstacle, however, is robotic programming. Almost every subunit is different and must be programmed individually; therefore, programming time is a key factor for the success of welding robots in shipbuilding.

In the automotive industry, the amount of time to program a robot is about twenty times the amount of time to perform the weld; however, the programming cost is spread over millions of identical welds using the same program. In military shipbuilding where every unit is different, programming cost is not afforded the large duplicate volumes but must be spread over one unit or only several at most. Therefore, the human time to program must be less than the human time to weld for the robot to be cost-effective. One of the major initiatives by the National Shipbuilding Research Program and the Office of Naval Research Manufacturing Technology Centers of Excellence is to utilize robotics to reduce the cost and improve the reliability of welding. Since 2000 the University of New Orleans, College of Engineering (UNO COE) has participated in the Automatic Generation (AutoGen) project. AutoGen is a software tool that automates the planning and execution of control programs for robotic welding of ship structures. The software works by a process-centered evaluation of the geometric context of computer representations of ship designs, including component part juxtapositions and manufacturing plans for the work piece. AutoGen identifies each weld and appropriately characterizes it, assigning suitable procedures using the same decision processes as skilled craft and welding engineers. It constructs the robot motions necessary to accomplish welds and determines the correct process control values for each path.

When provided with valid data, AutoGen generates robot control programs completely without manual intervention or edits. Foundational concepts defining AutoGen provide for a general solution not limited by product scale or geometric complexity and bounded only by the accuracy and the dexterity of the combined robot and torch. AutoGen can handle compound curved surfaces and compute collision-free and singularity-free trajectories for the torch manipulator. AutoGen understands the geometry of the part and can plan welds on work pieces presented before the robot in different orientations. Weld planning proceeds according to local process conditions and requirements computed by direct reasoning from the geometric contents of the work piece design information and the context defined by the manufacturing plan. AutoGen completes construction of robot control programs by applying computed local conditions to weld process response surfaces and the kinematics and control language specifications of the available robot. The solution is a direct consequence of the requirement. AutoGen casts all of its entities as computational geometries. All of the core issues of joint design, torch motion, and weld process and placement are expressly and directly coupled to each other through the method of geometric reasoning.

The logical object design of AutoGen provides for ready expansion and affordable modification, including incorporation of equipment from different manufacturers and possibly different kinds or types than in the original. The input data required can be electronically extracted from currently available shipyard sources, although some internal consistency of information representation and completeness issues may emerge with testing. The architecture of the software is intentionally open to modification of most of the performance rules. The software has open, well-defined hardware and software interfaces. These are readily adapted and configured for implementation by individual shipyards and to robots from different manufacturers. The AutoGen planning capability can be applied to a wide range of process problems. This provides opportunity for a large domain of potential applications in a variety of industries.
Initial work from 2000 through 2003 achieved much of the anticipated AutoGen functionality as running code. In November 2003 an AutoGen demonstration was carried out for several shipbuilding personnel at the UNO ShipWorks Robotics Laboratory (SWRL) located within the Northrop Grumman Ship Systems Avondale Operations facilities. AutoGen automatically generated control programs and welding operation on shipyard work pieces appropriately scaled for the available ABB/IRB/1400 robot.

Work in 2005 ported AutoGen to automatically compute control programs for the SWRL ISU/Motoman UP6 gantry robot. Work under this funding also ported AutoGen to produce control programs coupled with the PAWS offline system for the General Dynamics Electric Boat gantry mounted with a semi-independent pair of Staubli manipulators. This work encompassed careful calibration of the coupled motions of two significantly differing machines and addressed the particulars of the separate controllers. A preliminary version of combined robot and gantry motion (i.e., greater than six degrees of freedom) was demonstrated at Electric Boat.

AutoGen development introduced computation of directives for multi-degree-of-freedom coordinated motion, enhancements in algorithms for selecting manipulator-base placement, torch motions, and weave functions. AutoGen development realized new functionalities using electrical ground-seeking touch sensing for measurement with the robot to determine the actual lay down of work pieces and the positions of components within the assemblages. The UNO demonstrated welds through a three-plane corner, joints with continuously varying welding positions around a ring onto a canted plate, and reflexive presentation connections on a collard Tee-section standing on plate and intersecting coped transverse structure. At Sandia National Laboratories (SNL), the free-space motion planner was recast for parallel computation.

Through these developments, UNO and its partners identified further work necessary to the approaching commercialization of the software and its production implementation in ship manufacture. Experience operating AutoGen defined the need for extensions in robot motion design and planning. Robot welding through continuously varying torch positions showed the necessity for refinement, including subordinate process domains and handed- weave patterns. The current software must be restructured from a laboratory experiment to industrial architectural standards. Finally, the UNO experience showed the need for memory management to handle larger work-piece models for a graphical user interface and utilities to handle common tasks.

The goal of 2007 research is to commercialize AutoGen for shipyard robotic welding applications. The automatic planning capabilities of AutoGen are required because manual robotic planning of the thousands of subassemblies is not feasible. This capability will dramatically improve the ability of welding robots to play a major role in ship construction, improving quality and productivity. The focus of the research is in two major areas: