There are several general-purpose multiobjective optimization software packages currently available in the commercial sector. In general, commercial optimization software tools provide intuitive user interfaces with shorter learning curves than academic codes. These software programs share similar
major characteristics, while some of them also offer unique capabilities. Examples of such software include the following:
One of the key advantages of commercial codes is their ability to seamlessly couple with multiple third-party (commercial and in-house) engineering tools. Unlike the academic codes, these commercial software packages are designed to work with a wide variety of modeling and simulation software programs. Each of them can be thought of as an integration platform that enables the automation of a complex design simulation, and in the meantime offers optimization capabilities to support decision making. For example, the workflow environment in modeFRONTIER
® is able to formalize, drive, and manage individual steps, such as CAD modeling, computation fluid dynamics (CFD), finite element analysis (FEA), and so on, composing an engineering process through the integration of the most popular engineering solvers, while the data can be transferred automatically from one simulation to the next. Such communication between software is usually guaranteed by application protocol interfaces or automatic file exchange.
Figure 5.27 shows a screen shot of the Workflow Editor in modeFRONTIER
®, and some of the third-party software supported in modeFRONTIER
®, including CAD, CAE, and generic applications, are listed in
Table 5.4. Additional wizard style tools are also available in modeFRONTIER
® to achieve the coupling with other commercial tools or in-house codes. Some commercial optimization software, such as OPTIMUS
® and BOSS Quattro
®, also provide native interfaces with major CAE and FEA systems, broadening the basic capabilities of those engineering tools.
In addition to a strong connection with third-party applications, each of the commercial software packages mentioned above also bundles a collection of advanced optimization techniques for both
single and multiobjective problems, ranging from gradient-based methods to genetic algorithms. For example, the multiobjective algorithms available in modeFRONTIER
® include multiobjective genetic algorithm (MOGA), adaptive range MOGA, multiobjective simulated annealing (MOSA), nondominated sorting genetic algorithm (NSGA II), multiobjective game theory, evolutionary strategies methodologies, and normal boundary intersection. Moreover, in modeFRONTIER
®, different algorithms can even be combined by the user in order to obtain hybrid approaches. For example, it is possible to combine the robustness of a genetic algorithm together with the accuracy of a
gradient-based method, using the former for initial screening and the latter for refinements, so that the efficiency of the design cycle can be further enhanced. During optimization, these commercial software codes continuously update product parameters and extract relevant outputs from individual steps within the workflow until an optimal design is obtained. Because running a large number of simulations for optimization can be computationally expensive in practical engineering design, response surface methods (
Jones 2001;
Poles et al. 2008) are employed in most commercial software packages to accelerate the optimization process.
Table 5.4
Examples of Third-Party Applications Supported in modeFRONTIER®
CAD | SolidWorks, CATIA, Creo, NX, Spaceclaim |
CAE | ANSYS Workbench, ABAQUS, Adams, Virtual.Lab, Image.Lab, ANSA |
Generic applications | Excel, OpenOffice, SCIBAL, MATLAB, LABVIEW, MATHCAD |
Another advantage of commercial optimization software over academic tools is that they often provide interactive result viewers to help users select the most suitable design. The alternatives obtained during optimization can be displayed in the forms of tables, charts, or two-dimensional (2D) and three-dimensional (3D) plots. As an example, the visualization of 2D and 3D Pareto fronts in OPTIMUS
® is shown in
Figure 5.28. Usually, if the problem definition exceeds three dimensions, a 2D or 3D subspace can be selected, and parallel coordinate plots can be used to determine best designs out of the set of Pareto designs. Information other than the optimal solutions, such as the design space and sensitivity, can also be extracted and analyzed using the visualization tools incorporated in individual commercial optimization software (for example, see
Figure 5.29).
Note that additional capabilities other than those discussed above are available in most commercial optimization software packages. For example, in modeFRONTIER
®, OPTIMUS
®, optiSLang
®, and Isight
®, robustness analysis and reliability analysis can be performed based on statistical methods, such as the Monte Carlo method, to deal with uncertainties that impact the optimization process. Also, modeFRONTIER
®, BOSS Quattro
®, and Isight
® allow users to run multiobjective optimization in parallel by evaluating more than one simulation at the same time using several local or remote processors.