Copyright © 1995, Mechanical Engineering.
In today's high-tech engineering environments, prototypes can take a
number of forms. Among a growing segment of design engineers, virtual
prototyping--visualizing and testing computer-aided design (CAD) models
on a computer before they are physically created--is becoming an
increasingly popular way to refine design assumptions and improve new
products. Virtual prototyping is accomplished by running a computer
model through iterative dynamic simulations before making a physical or
rapid prototype.
A physical prototype, rapid or not, can require a lot of manual tooling, skilled hand assembly, delicate testing instrumentation, and time spent interpreting prototype data. Engineers must incorporate what was learned by revising the design, making a new prototype, and repeating the entire process. The time associated with making more than one prototype, especially with design revisions between each iteration, can tie up engineers and equipment for days or weeks at a time in creating mock-ups that may not be nearly as informative as they need to be.
Virtual prototyping, which performs all of the above steps on a personal computer, runs more variations than a rapid prototyping system and permits the design of tests not feasible in the laboratory. For instance, an engineer can now measure tiny components that ordinarily would be difficult to instrument without affecting test conditions. Forces that are difficult or impossible to simulate in the laboratory (such as zero gravity) can also be applied to the design. This versatility permits CAD model modification and refinement to be accomplished much more rapidly than with conventional design tests.
Rapid prototyping (RP) systems produce relatively quick, life-size models that engineers and other decision makers can touch, feel, and hold in their hands. These physical models are especially helpful to less technically oriented people, who need to see how a part will look or how subassemblies may fit together to produce a finished product. When aesthetic evaluation of a model is in order--including such practical matters as its physical fit with other components, its proposed package design, or even its use as a marketing sample--rapid prototyping is a very important method for addressing design needs.
Before creating a rapid prototype, however, engineers should first consider improving results by analyzing the kinematic and dynamic properties of their design through the use of new, inexpensive software programs designed to apply the laws of physics to their computer-generated model. While not appropriate for every application, any CAD model involving moving parts (or models that will be affected by external forces) will benefit from kinematic/dynamic analysis.
Certain rapid prototype models can be used for motion simulation, but
they are rare exceptions. That's because the physical assembly of one
or more rapid prototype components--linking objects with ball joints,
pins, or sliders--is time consuming and tedious. In a virtual
prototyping environment, however, attachments can be accomplished with
mouse clicks. Kinematic/dynamic analysis can be thought of as another
form of rapid prototyping. It is certainly faster to model a design,
construct a simulation, and render and animate the results for others
to review by importing and exporting data from one application to
another.
Engineers who have adopted kinematic/dynamic analysis have discovered that they don't use it merely to verify the soundness of their initial design; the feedback received from a good dynamic simulation gives users important information that can lead to modifying the parameters and improving the design in ways that might otherwise have been overlooked. Such flexibility can help counteract the tendency to overdesign a product to the point where alternatives are viewed with chagrin if not downright hostility. Kinematic/dynamic analysis doesn't just verify original assumptions--it questions them.
Flinn uses Working Model 2.0 from Knowledge Revolution, in San Mateo, Calif., a desktop software package that applies the laws of physics to CAD models. The program mathematically calculates the movement of objects with assigned masses and depicts their motion over time upon encountering other objects or forces. The forces can be measured using on-screen tables that measure displacement, velocity, and acceleration while the simulation is running. Once the first run has finished, the user can replay the simulation again and again at faster speeds (because the dynamic calculations have already been performed). Or the simulation can be played back frame by frame.
If the results demonstrate a need for altering a design specification, the user makes the modification and then runs the simulation again to see the new result. Conversely, the user can modify a test to see new conditions act upon the design without altering the specifications. Relatively quick and easy changes can be made to the CAD model, the simulation, or both, allowing the engineer to concentrate on solving the engineering problem, not on the simulation tool itself. "We were able to complete a simulation of a transmission column shift system and have results in only a fraction of the time it would have taken using physical prototypes or standard Unix-based kinematics software," said Flinn.
"A lot of the analysis portion of prototype testing takes place after the simulation, when you sit and look at a velocity curve or some other piece of data and decide what's required next," said Sergei Fedorjaczenko, a design engineer at Carlingswitch Inc., in Plainville, Conn. This manufacturer of magnetic circuit breakers and electrical switches uses kinematic/dynamic analysis to perform simulations for tests that are measured in milliseconds.
Fedorjaczenko likens examining a tested circuit-breaker prototype to an
autopsy. "In the event of a failure, we try to see what failed first
and why. These events usually involve an interruption of current and
voltage, which means an event lasting no longer than 10 to 20
milliseconds. Very often, the conclusions of this autopsy are
speculative and not very precise. You get a feel for it, but you're
still not quite sure."
Fedorjaczenko uses the Working Model software to check whether Carlingswitch's simulations are running correctly. Tables and other data provided by the simulation itself let him evaluate what needs to be changed in order to improve results.
"It takes only about five or 10 minutes to change a variable and run a particular simulation again, so you're able to do a lot of what-ifs," Fedorjaczenko said. "We go through literally dozens of these what-ifs in a matter of hours."
Different RP systems use different materials, such as thermoplastics and powdered metals, that can be expensive. Other systems create hollow models, which are fine for appearance checks or packaging concerns but cannot be used for kinematic/dynamic analysis. With a good kinematic/dynamic analysis program, the designer can assign to a design the physical properties of many different materials, such as steel, ice, plastic, clay, or any custom material imaginable, and perform motion analysis as if a working prototype existed. The virtual prototype can be analyzed using calculations carried out to as many as 20 decimal places. This analysis improves the accuracy of predictions of how the model will behave and ultimately makes the model more valuable in the creation of a rapid prototype.
To perform kinematic/dynamic analysis, the mass of the object and the distribution of that mass are key parameters, which is why a hollow rapid-prototype model is not helpful. Or, in cases where the rapid prototype material is a laser-setting resin and the intended part is to be made from heavier material, the designer will be unable to perform any meaningful dynamic analysis. A prototype should help an engineer make design decisions, but by resorting too soon to rapid prototyping, the engineer will lack a reality-based prototype and may miss exploring some attractive design alternatives.
RP systems are limited in the size of the prototype that they can produce. It is difficult to produce very small parts (say, less than an inch) at full scale, because of the resolution problems of RP systems. Similarly, parts larger than the workspace of an RP system are difficult to make, because large parts must be sectioned, with each piece built separately. The separate parts are then assembled into a full-scale rapid prototype, but the limitation of such a jigsaw-puzzle approach becomes readily apparent for products more than a few feet in size. Computer simulations are, of course, not constrained by the physical size of a design.
Rapid prototyping is not an effective way for designers to explore as many options as with simulation software, because RP is not a practical method for performing what-if analysis. Some rapid prototype models take hours or days to create, which discourages multiple iterations. Although, with several newer systems, the price of creating rapid prototypes is dropping, they still cost a significant amount in both time and money. A kinematic/dynamic analysis program used in the early stages of design helps refine concepts, performs optimization, and troubleshoots poor designs by revealing their dynamic flaws. When an engineer is convinced that a design can perform as predicted and satisfy the requirements of the initial problem, then a rapid prototype can be the ideal validation-- especially after completing 100 computer iterations.
The output from an RP system is unlikely to provide FEA data, because in most instances the rapid prototype material isn't the same as that of the finished product. Kinematic/dynamic analysis, however, can provide reliable FEA data. It is also more convenient, as it is already in a format ready to interface with FEA programs. In addition, many more data points can be gathered from kinematic/dynamic analysis than would be practical using an instrumented rapid prototype. Such an analysis, therefore, increases the accuracy of load measurements.
Some RP vendors are addressing the cost issue by developing less expensive "benchtop" systems that help take the hesitancy out of implementing a rapid prototype. This is the approach taken by Stratasys Inc., in Eden Prairie, Minn.
"As a rule, most engineers are not risk takers," said Scott Crump, chief executive officer of Stratasys, which manufactures an RP system that generates thermoplastic anti-lock brake system models from CAD designs. "Engineers want to minimize the tooling risk and eventually see 3-D output." However, Crump said, the industry "is still in a mainframe mentality" when it comes to rapid prototyping. According to Crump, rapid prototyping has blossomed only in the last four years with the advent of faster and less expensive systems.
For some specific applications, a rapid prototype can be used as a real part: for example, a space holder that's not highly engineered or machined. But rapid prototyping is generally intended for creating life-size physical mock-ups, which still have fundamental value in many applications. Some rapid prototypes, for example, are used to create molds from which real parts are created, such as medical implants. Rapid prototyping is also an excellent engineering check for final tooling. At times, good rapid prototype models are even better than conventional tooling molds.
An engineer's productivity will clearly be improved when more analysis work can be done on a design before rapid prototyping, taking it farther and through more conceptual iterations before engaging the manufacturing engineer and tying up RP equipment. Rapid prototyping as a resource is used most efficiently when it is used with confidence that the product design is as close as possible to reality. The use of kinematic/dynamic analysis software ensures that when a rapid prototype is finally made, it will be the excellent design that was originally sought.