Not counting fasteners,, a Metallic aircraft fuselage comprises thousands of parts made mainly from sheet and extrusion stock. They are formed into unique shapes to create lightweight, strong, aerodynamic craft. In forming, tools force raw stock into the shapes desired. These dies, punches, form blocks, and so forth account for much of the initial production costs of a part.
Manufacturers have long desired single universal tools that could make many different shapes. Patents for reconfigurable” discrete-die” tooling-dies composed of movable elements such as pins or plates-date back nearly to the inception of the patent office.
In the late 1970s, David Hardt of the Massachusetts Institute of Technology explored the mechanical design and shape control algorithms of discrete-die tooling. He built a hydraulic press with matched 12-square-inch reconfigurable discrete-die tooling composed of 0.25-inch square steel pins. Computer-controlled servo motors moved the pins individually along their axes. Hydraulic clamps locked the pins in position by applying pressure sideways.
One of Hardt’s many contributions to the field was the concept of self-supporting pins-pins arranged in a densely packed array. Earlier designs had failed because many lacked stiffness in their freestanding discrete elements.
Later, C. Robert Crowe at the Advanced Research Projects Agency (now DARPA) funded a collaboration of Grumman Aerospace (now Northrop Grumman Corp. of Bethpage, N.Y.) , MIT, and the Cyril Bath Co. of Monroe, N.C, for the purpose of translating laboratory concepts into a full production ‘ device. The reconfigurable tool that resulted has a working volume of 42 x 72 x 12 inches. Used for stretch forming, the tool employs 2,688 movable pins, each 1.125 inches square by 21 inches long.
The tool has proved itself during trials in a production environment. The forces needed to deform sheet metal are provided by the motion of the stretch press jaws and table, which are actuated hydraulically. The tool does not move during forming; it merely supports the imposed loads.
Pins on this forming die, locked in the shape of a saddle, produce a toroidal sheet with both negative and positive curves. The shape tests die performance.
During trials, the die table pushes the die and polymer blanket into the sheet as the stretch press jaws pull the work sideways. The jaws rotate to maintain tangency.
Subject to Interpolation
An obvious challenge is forming smooth skins over a tool whose surface is composed of 2,688 pins withrounded ends. Some reconfigurable tools, such as one devised by North Sails for sail making, use pivoting flat plates as their working surfaces. This approach fails to support the loads generated in deforming metal.
MIT researchers found that a dense array of pins with hemispherical tips provides the most universally acceptable tool surface because it places the fewest restrictions on the overall shape of the surface envelope of the tool. But, this pin end shape can dimple parts that are formed directly on its surface.
Experiments on dimple suppression at MIT and Northrop Grumlnan showed that many polymeric materials could be used as blankets to interpolate the locally uneven tooling surface into a smooth overall part shape. Early experimenters heated thermoplastic materials to match the shape of the tool. The blanket would then be used for forming at room temperature. Heating and preforming were later found to be unnecessary. The current interpolators simply deform during forming and springback to their original shapes afterward.
Through both experimentation and finite element modeling, investigators discovered that the key variables in dimple suppression by polymeric blankets were thickness and compressive modulus. Because pressure distribution that is locally uneven leads to dimpling of sheet metal, a good interpolator must reduce this pressure variation to a level that the sheet can tolerate without forming a dimple. Resistance to dimpling is conferred by the sheet’s local bending resistance and by the membrane forces imposed as the press stretches the sheet.
Using finite element analysis, MIT’s Simona Socrate and Mary Boyce, along with Northrop Grumman’s Lembit Kutt, investigated the ability of a polymer blanket to smooth out the local pressure. Kutt ran finite element analysis of the entire pin die, interpolator, and sheet metalassembly during a stretch forming operation. In this manner, Kutt gathered information about the mechanical behavior of the interpolator.
Using additional single-pin models with boundary conditions set for an infinite array of pins, Kutt found that the ratio between the highest pressure seen by the sheet metal (directly over the center of a pin tip) and the lowest (over the corner of the pin) could be reduced to 1.07 with a 1-inch-thick interpolator having a compressive modulus of 1,100 psi. (A perfect interpolator could be made from a fluid, but how to contain it remains a question.) Typical aircraft sheet, say 0.063-inch 2024-0 aluminum, supports this pressure difference without dimpling.
Finite element analysis on a range of parameters examined interpolator modulus, thickness, Poisson’s ratio, local contact friction, and assumed boundary conditions. By defining a parameter space in which dimpling was not expected, the researchers produced a guide for interpolator selection. Experimentation has confirmed these predictions, and dimpling is not a problem in part production.
Finite element model stretch wraps a O.063-inch sheet over a 36-pin die to form a spherical cap. A l-inch-thick interpolator separates them.
Another challenge was designing a tool with 2,688 moving elements that could withstand the rigors of a productionshop. Besides enduring the loads imparted during fonning, the tool had to bear rough handling as it was installed and removed from the press.
Repositioning the pins had to happen rapidly and precisely. Each pin could be off by no more than 0.002 inch. Full reconfiguration of the die could take no more than 15 minutes.
A computer control system would instruct the pins to assume an arbitrary contour specified by a pin setting file derived from a CAD file of the part shape.
Investigators examined a number of concepts for fixing the pins. They abandoned the clamped-rod concept used in the MIT tool after discovering that the sidewise hydraulic clamping pressure needed for a tool this size would be prohibitive.
Investigators at Rensselaer Polytechnic Institute in Troy, N.Y., tried supporting each pin hydraulically through individual cylinders and on/off valves under open loop control. They found the system to be inaccurate and insufficiently stiff. They considered individual hydraulic servo valve control of each pin too expensive and complex.
The researchers discovered that pins could withstand the forming loads if a 5/ 8-inch-diameter threaded rod supported each one. Threaded rods could also position pins vertically.
Investigators explored three ways of positioning the threaded rods. In one scheme, 16 high-speed, high-precision servo motors with position encoders would be mounted on an x-y table underneath the tool. The motors would rapidly position individual pins (taking about 10 seconds) and then the x-y table would shift quickly to another set of pins to repeat the process. Because the toolhad to be removed from the stretch press and placed on a setting table for each shape change, the idea was discarded for being slow.
Another concept involved driving each pin’s lead screw through a shaft and worm gear. Shafts running through the bottom of the tool and turned by a large externalmotor would engage each pin via electromagnetic clutches. The vast number of wires needed for control-ling every clutch eventually led to this concept’s demise, as did difficulty in makingthe system modular.
The chosen concept called for installation of small dc motors under each lead screw. Each motor uses an integral 84: 1 gear-reducing head and a position encoder. Eight pin-motor combinations were assembled to make one module. The tool holds 336 modules.
Each module contains a microprocessor and motor controllers mounted on a printed circuit board. Each module engages a 10-pin electrical connector in the base of the containment box as it is installed. Modules are installed and removed vertically, allowing for any given module to be removed independently for service or maintenance.
The electrical connector provides power at 5 V and 16 V for the logic and motors, respectively, and RS-232 lines for communication with the host computer. The modules are daisy-chained on the RS-232 line, and the initiation command from the host causes each module to number itself sequentially, regardless of where it may have been positioned.
The local microprocessor controls all local functions, receives instructions from the host computer, turns individual motors on and off, and continuously stores the position of each pin. It communicates with the host computer only when receiving commands or transmitting a reposition accomplished signal. It provides various error signals. The arrangement permits efficient control of all 2,688 motors with minimal communication to the host computer. One RS-232 line is adequate for communication between the tool and the control computer.
An additional advantage of the computer control system is the ability to reconfigure the tool in stages, reducing the total power required by the tool during the procedure. Currently, the software divides the tool into three zones, and moves all of the modules in a zone at one time. Reconfiguration takes less than 12 minutes.
All those pins and motors moving at once would draw tremendous current. To improve power management, engineers split the reconfiguration sequence into three steps, while keeping total reset time below 15 minutes.
An eight-pin module makes up the fundamental building block of the 336 module tool. Any module can be replaced independently.
In forming large sheet metal parts, elastic spring back can cause unacceptable part shape errors if the part is formed on a tool whose shape mirrors that of the finished part.
Two approaches to tool shape correction have been developed for use with the reconfigurable tool. In one, an iterative technique calculates the correct die shape based on two initial guesses. In the other, a predictive technique uses finite element analysis to model and correct for elastic springback.
Two parts are formed with the iterative process, one in the net shape desired and one distorted by approximately 10 percent. All four shapes-the two parts and the two dies that make them-are compared. Then the correct tool shape is calculated.
Walter Norfleet of MIT has concluded that a technique based on identification of the part shape’s local sensitivity to each individual pin’s position is more efficient than an incremental correction method, or a method in which the correction is calculated in Fourier space. The Fourier approach, in which a series of superimposed sine waves of various frequencies describes tool shape, couples the entire part shape to motions of individual pins.
In the predictive approach, a model of the forming process is prepared and an analysis performed over a net shape tool. In the fully loaded configuration, the residual moments that would act in the sheet to cause elastic springback are reversed and applied in a forward manner, deforming the tool and correcting it for spring back. The procedure, devised by Boyce and Apostolos Karafillis at MIT, is called “springforward.”
Because this procedure requires detailed knowledge of specialized software, Elias Anagnostou of Northrop Grumman developed a simplified front end. It prepares the finite element models for analysis using Abaqus from Hibbitt, Karlsson & Sorensen in Pawtucket, R.I., submits the job, and then helps to postprocess the results. Anagnostou created a simplified user interface using Patran command language from MSc.Software of Los Angeles.
After the user makes a series of choices from pull-down menu, the software imports the part shape file from a CAD system, prepares a finite element model of the tool, material, and process, and submits the job to Abaqus. Once Abaqus finishes its calculations, which usually takes several hours, it returns the results to Patran, which displays them graphically. The software eliminates the need for multiple tool tryouts on the production floor.
Two major aerospace firms have completed shop floor trials of the reconfigurable tools. Another tool is under construction for an aircraft rework and maintenance center.
Cost and benefit analysis has shown the attractiveness of reconfigurable tooling based on the initial investment alone. An even greater benefit could be realized through lowered assembly costs and shortened turn around time for small production lots. Small-lot production shops are likely to profit by replacing fixed-shape forming tools with reconfigurable tools.