Intelligent Bending Workstations


Main Participants: Satyandra K. Gupta, D.A. Bourne, K. Kim, and S.S. Krishnan.

Sponsors: This project was sponsored by Amada.

Keywords: Automated Process Planning and heet Metal Bending.

Motivation

In order to offer flexibility, better quality control, higher degree of automation, and improved productivity, machine tool manufacturers are combining material processing, material handling, and part positioning systems into single integrated manufacturing cells. Programming such integrated cells manually is a time consuming task and can become a major bottleneck in effectively using such cells. Process planning and part programming time directly affect the lot sizes that can be economically produced on these cells. We believe that automated process planning systems can significantly enhance the throughput of such integrated cells and dramatically lower the economic lot sizes.

Depending upon the level of process plan details, the process planning systems can be divided into two different types: macro planners and micro planners. Macro planners deal with the higher level process planning decisions such as selection of machines, selection of operation types, ordering operations, selection of tools etc. Micro planners deal with the lower level planning decisions such as selection of operation parameters, NC code generation etc. To create a completely automated process planning system, we need both capabilities. Traditionally, these two types of planners were developed independently and were interfaced later. Due to strong interactions among various components of an integrated manufacturing cell, macro planning and micro planning functions need to be tightly integrated into a single system.

Main Results and Their Anticipated Benefits

We have developed an automated process planning system for a robotic sheet-metal bending press-brake. Our system is based on the generative approach and performs both macro as well as micro planning. Once a CAD design is given for a new part, the system determines: the operation sequence, the tools and robot grippers needed, the tool layout, the grasp positions, the gage and the robot motion plans for making the part. These plans are sent to the press-brakes controller, which executes them and then returns gaging information back to the planning system for plan improvement. A second plan is then formulated, which reduces the gaging time by incorporating the reduced uncertainty in the part location.

Our system is based on a distributed architecture. We have a separate planner for each specialized component of the robotic press-brake. These specialized planners collaborate with a central operation planner to perform the process planning. Currently, our system consists of a central operation planner and three specialized planners: tooling, grasping, and moving. The central operation planner proposes various alternative partial sequences and each specialized planner evaluates them based on its objective function. A distributed architecture allows us to encapsulate specialized planning knowledge of each component into a separate module and provides an opportunity for using a different representation and problem solving technique for each planning module. This architecture also provides a highly modular environment for adding more specialized planners to the system. Our system presents a significant improvement over the state-of-the-art. After the release of final CAD file, using our system, we can produce the first part in less than an hour. For full production automation of sheet metal bending, the resulting planning and execution time is reduced for the first part significantly.

In our system, there is one central planner that sends out queries to specialized planners. The central planner keeps track of the query results and develops a near optimal plan. Specialized planners act as servers, which solve problems in grasping, tooling and moving for given partial operation sequences. Note that all of the planners communicate in the Feature Exchange Language (FEL), which is a human readable, extendible language. This FEL syntax is regular, human readable and easily processed by each module.

In our system, the part's design is presented to the planning system, which automatically plans all aspects of the setup and the execution steps for making the part. A person is then guided step-by-step in the setup process and the plan is sent to the controller. The controller has a built-in interpreter for executing the plan on the bending machine. The part is loaded by a separate loading-unloading robot, and the bending robot starts to bend the part bend-by-bend. At some point, the robot may interfere with a bend-line, and as a result the robot hands the part to a repositioning gripper, so that the robot can alter its grasp position. The bending, and regrasping are continued as needed until the part is complete, at which time the unloading robot grasps the finished part and stacks it. The results of this production run are used to produce a better and faster plan, since most of the gage information can be reused making successive parts (i.e., the robot is not accurate but it is repeatable). This modified plan is then used to make the rest of the parts in the batch.

Our system has been implemented using the C++ programming language. For geometric modeling and reasoning we have used NOODLES geometric kernel. For graphical interface we have used HOOPS graphics library. All message passing among planners is accomplished by Feature Exchange Language.

Our system architecture offers the following advantages:

Related Publications

The following paper provides more details on the above-described results.

Contact

For additional information and to obtain copies of the above paper please contact:

Dr. David Bourne
Robotics Institute
Carnegie Mellon University
5000 Forbes Avenue Pittsburgh, PA
Email: db@ri.cmu.edu