Optimizer for STK/Analyzer

• Gradient-based optimizer
• Adaptive Surrogate optimizer
• Darwin Genetic optimizer
• Variable Influence Profiler
• Prediction Profiler

Optimization Algorithms

• Gradient-based optimizer. Local search for optima in problems with smooth, continuously varying objective and constraint functions.

• Adaptive surrogate optimizer. Part of Boeing's Design Explorer software, the Adaptive Surrogate optimizer solves complex problems characterized by long running analyses. It intelligently uses non-physics-based mathematical models (Kriging models) to reduce the number of required scenario executions. A global search algorithm, it is not likely to get stuck in local optima. The software also performs well in noisy design spaces and is robust in the face of scenario failures.

• Darwin genetic optimizer. Darwin Genetic algorithms are ideally suited for design problems with discretely valued design variables (e.g. integer variables). Because they do not require objective or constraint gradient information, genetic algorithms effectively search discontinuous and "noisy" design spaces. Compared to gradient-based optimization algorithms, genetic optimizers are much more likely to find the best design. They can find many near-optimal designs too, thus providing more design alternatives.

  Darwin Genetic algorithms accommodate up to three objectives for your design problem. For multi-objective problems, Darwin generates a Pareto trade-off curve. The design points on the Pareto curve are all optimal in that they represent a point that would be impossible to improve without degrading one objective over the other(s).

Design Space Tools

• Variable Influence Profiler. Performs sensitivity analysis on a scenario. Users can generate variable importance plots that illustrate the relative impact of selected parameters on a scenario, and generate main and interaction effects plots to view and understand overall trends. By focusing on those variables that are most influential (and by placing less emphasis on those variables with little influence), the complexity of the design task is significantly reduced. Gaining a better understanding of variable relationships and overall design trends encourages more effective design problem formation and location, and more accurate optimization results.

• Prediction Profiler. Interactively explores the design space. Users can efficiently predict output variable values for any combination of input variables, visualize slices of the resulting /n/-dimensional design space (including the effects of constraints), and manually explore and search the design space for good designs. This tool can be used as a "manual optimizer" to locate good designs, or it can be used to find good starting points for a more formal optimization algorithm. It may also be used as a post-processor to visualize the design space around an optimum point generated by an optimization algorithm.