By Joe Greco
CAD and CAE systems do a great job of helping us explore a variety of possible solutions to design and engineering problems. But using this software to find the best solution can mean many long hours of setting up what-if scenarios and waiting for the results. Fortunately, a powerful set of behavioral optimization tools can spell relief for those overloaded with options. With these programs, users specify what they're aiming for, then sit back while the computer finds which design variation does the best job of meeting their goals.
While the idea of having the computer do the tedious number crunching sounds like engineering nirvana, getting to this stage isn't for the faint-hearted. Properly specifying design goals requires engineering experience, and using the programs sometimes requires programming ability. But if a user with even a moderate amount of expertise is willing to invest some time in these programs, they can go a long way toward making life a lot easier.
There are two ways behavioral optimization tools work. They can find answers to specific questions, such as "What changes are needed to align a cam's center of gravity to its center of mass?" Or they can optimize specific parameters, such as the shape of a tank, when other parameters such as the tank's volume are given.
To test the current capabilities of behavioral optimization software, I chose four of the leading programs and put them through their paces. The first two, GrafiCalc from Geomate and MechSoft.com from MechSoft.com Inc., are standalone products that are integrated to varying degrees with traditional MCAD software. The last two are modules of high-end MCAD standards: PTC's Behavioral Modeling Extension for Pro/Engineer and Dassault Systemes' Product Engineering Optimizer for Catia. Note that while this quartet doesn't represent all the behavioral optimization packages available today, these are currently among the most popular.
Geomate's offering is the only software in this group that is a 2D product, but that doesn't mean it isn't powerful and extremely useful. With GrafiCalc, variables such as area, length, angle, perimeter, moment of inertia, volume, cost, weight, clearance, distance, tolerance, stress, and deflection are specified as behavioral goals. Then GrafiCalc automatically evaluates various design solutions and finds the ones that best fulfill the desired requirements.
Users have the option of either importing 2D geometry via DXF from the sketcher of their MCAD software or using the program's built-in sketch tools. In either case, the idea is to use GrafiCalc before committing to any 3D solid geometry. Once the behavioral optimization is carried out in GrafiCalc, the resolved 2D information can be brought back into the MCAD program. If the user needs to maintain an associated link between the GrafiCalc geometry and the MCAD program, this can be done via Excel using Windows DDE.
 |
| With GrafiCalc, the user calls up the Goal Seek option to specify target values and variables (near right). The program then performs the optimization and displays the results (far right). |
I found working with GrafiCalc to be fairly easy. For instance, without much practice, I was able to have GrafiCalc optimize a profile I had drawn, by requesting that it alter a dimension in order to have the total area equal a value I had specified. This was done by first launching the Goal Seek command and then simply clicking in the appropriate locations on the screen to define three requirements. The first requirement was the Actual value, which was the existing amount of area of the profile. The second requirement was the Target value, which was the amount of desired area that I wanted the profile to be. The final requirement was specifying which dimension on the profile was to change in order to have the area of the profile equal the target area.
While this isn't the most difficult example of behavioral-optimization, I was impressed with how easy it was to set up and how the visual aspects of the program al lowed me to avoid writing any complex equations. However, there are a few areas where the software could be improved. For example, it could benefit from more Windows-like keyboard commands, as it uses antiquated shortcuts such as Alt-backspace for Undo. Despite this, GrafiCalc, which costs only $295, is an excellent engineering tool for someone who needs behavioral optimization but doesn't necessarily possess a great deal of engineering expertise.
 |
| MechSoft.com allows users to work with CAD models, like this gear assembly created in SolidWorks, and perform optimizations such as finding minimum mass while maintaining size and safety requirements. |
MechSoft.com allows engineers to use its library of standard parts to build intelligent assemblies based on real-world engineering calculations and then use behavioral optimization routines to enhance the design.
The software is built around a feature called the Group Manager, which contains an overview of the functional assembly and the physical calculations that drive it. For example, using the library of standard parts, a user can place a pulley and a conveyor belt into the Group Manager or these items can originate in a CAD system. In any case, the links that define the interaction of the parts with each other are understood by the program, and then the calculations that make the pulley-belt system work are chosen from a list of standard mechanical engineering calculations. In addition, for special conditions, user-defined calculations created using Microsoft Excel can be employed.
The optimization comes in when the user requests that the software solve a specific problem such as "find the pulley system that can move the most material while still using a motor with less than 5 horsepower and a belt that is less than 18 inches wide." The user can specify a certain amount of time to find the answers or ask the software to generate a certain number of improvements. Then the simulation is run and the results are plotted for evaluation.
 |
| MechSoft.com's Optimization dialog box provides the user with a record and list of possible solutions from which to choose. |
I used a version of the MechSoft.com product that is integrated with SolidWorks 2001 and opened an assembly file that contained two gears mated together. My goal was to optimize the gearbox in order to find the minimal mass, while still meeting other required safety criteria and dimensional requirements. To be gin, I opened the Group Manager, which displayed how the two parts were linked together and how they were associated with a calculation. The graphical user interface that MechSoft.com employs is helpful, but the program does require a thorough understanding of basic engineering principles.
With the physics of the assembly defined, I launched the Behavioral Optimizer module and specified the constraints, requesting that certain power and speed requirements were met. Then I defined the minimal mass goal, and requested that the program find solutions until 100 consecutive models were created without improvement.
 |
| Pro/Engineer's Behavioral Modeling Extension was used to optimize this racing engine manifold. The goal was to minimize the difference between the length of the pipes. |
I found running the Behavioral Optimizer and reviewing the optimum solutions that I requested to be fairly straightforward. However, MechSoft.com offers so many different features and so much information in each dialog box that it's easy to get a little lost. But overall, it was obvious how the software could save an engineer a tremendous amount of time.
 |
| The Behavioral Modeling Extension explored more than 1000 possible manifold design solutions. It then displayed them in a scatter graph so the user could visualize options and tradeoffs. |
PTC's Behavioral Modeling Extension works with the 2000i and 2001 versions of Pro/E. It provides a package of behavioral routines that enables users to explore design ideas by automatically optimizing geometry. According to PTC, this extension has been implemented at the core of its system kernel, meaning all Pro/E applications can work with the engineering solutions created with it. At the same time, there are links to external applications.
Behavioral features inside Pro/E can capture important design measurements that can be used to drive a design. An example of Pro/E's behavioral capabilities is the ability to achieve optimal engine performance by having the desired constant flow-rate through an intake manifold drive the defining cross-sections of the pipe. Even the angle of reflectivity off a surface can be measured and captured within a behavioral feature, and that value can be used to drive the surface's curvature.
 |
| To use Catia's Product Engineering Optimizer, the user sets a target goal for a model, in this case a volume of 400 cubic centimeters, then hits the Edit List button to add parameters. |
Users specify how many design iterations they would like to see and how they would like to see them. For instance, you could ask Pro/E to display a scatter plot showing its 200 best solutions for keeping an automobile's drag coefficient low while maintaining its buckling strength. The resulting color graphs and plots update if the design changes. In fact, with Version 2001 of the software, optimizations are automatically rerun if the design changes, as opposed to the previous version, which required them to be manually run.
Behavioral optimization can also be found in other PTC products such as Shrinkwrap, which can be used to predict the intended motion behavior of an assembly, and Pro/Mechanica, which is used for mechanical dynamic analysis. PTC plans to integrate the Behavioral Modeling Extension with its Routed System tools, which will allow for pipe and cable lengths to be optimized based on various analyses. The company also plans to incorporate behavioral optimization techniques in its plastic, mold design, and weldment applications.
 |
| As the Product Engineering Optimizer optimizes the model, it displays the updated geometry as well as a feature tree, which shows the formulas that were needed to create the optimization. |
An optional module of Dassault Systemes' Catia, Product Engineering Optimizer allows users to optimize designs according to a product's weight, total volume, cost, or a number of other conditions. As with the Pro/E module, this product can create scatter plots showing different scenarios of how various user-defined constraints balance against each other, for example, a vehicle's aerodynamics versus its weight versus its strength. It can then take the best solutions and link pricing data to them to help users determine the appropriate materials and tooling processes.
I had the chance to use Catia's Product Engineering Optimizer to modify a 3D shape in order to change its original volume to one that I had specified. It was easy to set up and run the simulations. But some of the work had been done for me, because in the sample file that I was using, some complex formulas had already been written.
According to the company, future plans for behavioral optimization include immediate feedback on whether or not a given part, once placed, will work based on conditions such as stress or heat. Essentially, the goal would be to have Catia parts understand how they should behave in the real world.
As mentioned, these are not the only products available for behavioral optimization. Other products include Concept from MCS.Software, which is used, for example, to take a rough shape, such as a bracket, and optimize it for weight. In addition, LMS International makes software for optimizing acoustics, structural vibration, and multibody dynamics, while SRAC develops products that can optimize both parts and assemblies for structural efficiency.
Behavioral optimization is where engineering software is headed, and these applications, while extremely useful, provide only a glimpse of things to come. For example, when combined with knowledge-based systems, a program could anticipate what an engineer is designing, and based on past experiences, automatically add parts, such as motors, that function in the context of that design. In the end, this will mean more productive engineers, and less expensive, more efficient products for consumers.
Joe Greco is a frequent contributor to Computer Graphics World who specializes in writing about computer aided design. He can be reached at Joe3d@home.com.
Dassault Systemes · www.dsweb.com
Geomate Corp · www.graficalc.com
LMS · www.lmsintl.com
MechSoft.com · www.MechSoft.com
MSC Software · www.mscsoftware.com
PTC · www.ptc.com
SRAC · www.srac.com