Generic Tool for Optimization (GT Opt)
MACROS Generic Tool for Optimization provides a variety of innovative methods to conduct optimization of your model or multiple models in case of multidisciplinary optimization. Depending on your needs the tool automatically invokes particular algorithm which seems to be the most suitable for your particular problem.
Thus for overwhelming majority of users the optimization process is automated and does not need detailed supervision. But that does not imply that you have no means to influence optimization process. To the contrary, MACROS Generic Tool for Optimization is highly configurable: you can change virtually all relevant aspects including algorithms selection policy and parameters of each method.
Combined with the above unmanned operational mode, Generic Tool for Optimization appears to be unique "Swiss-Army Knife" -like tool to solve your problem.
The pool of in-house developed state-of-the-art algorithms available in Generic Tool for Optimization covers virtually all categories of engineering optimization problems.
Single objective optimization
Single-objective constrained optimization is suitable for both analytic models and problems formulated on top of complex numerical simulation process. Moreover, depending on problem type and properties the tool automatically corrects the selection of best suited algorithms which range from dedicated sequential quadratically constrained quadratic programming for smooth underlying problems to more simple but robust derivative-free methods for noisy engineering applications.
Multi-objective constrained optimization
Here the tool offers unique innovative algorithms to find Pareto frontier and ensure Pareto optimality of obtained solutions. Implemented methods are more efficient than the commonly used approaches, in particular because they almost never evaluate designs far away from optimality. Instead (and this is the key feature) Generic tool for Optimization pinpoints a few optimal solutions and then diffuses along Pareto frontier always staying near optimality surface.
Robust single- and multi-objective optimization
Robust single- and multi-objective constrained optimization based on original state-of-the-art stochastic approach developed by DATADVANCE.
The heart of the method is careful adjustment of current number of sampled uncertain parameters: away from optimality it is more than enough to consider only a small number of random realizations, but their number must increase once optimal solution is approached. Moreover, the unique feature of this family of algorithms is that they not only provide the solution itself, corresponding uncertainty estimates of objective/constraints values at found solution are provided as well. A particular advantage of our robust optimization algorithms for engineering applications is that the tool never need to know explicit distribution law of the uncertain parameters, it is enough to provide their distribution only empirically.
Surrogate based optimization
Surrogate based optimization (SBO) implemented using our in-house developed Generic tool for Approximation. This is a particular example of joint power of various MACROS Generic Tools: in SBO framework optimization and approximation tools are used together to produce highly efficient global optimization methodology.
In more details, our SBO implementation is modeled around expected improvement (EI) approach, but enhances it many different aspects. In particular, it combines globalization properties of usual EI with fast convergence of usual optimization methods. Moreover, multi-objective constrained problems are dealt with equal success as well due to the usage of modern filtering optimization philosophy.