BOSS quattro is our application to manage and take advantage of parametric models. BOSS quattro features several engines, each of them having a specific purpose:
- Parametric analyses are the simplest way to evaluate the influence of given parameters on the behaviour of your structure. It just launches analyses for each set of parameters values defined by users. Results can then be plotted as a function of the parameters.
Design of experiments has the same purpose as the parametric analyses. The main difference is that the successive values of the parameters are cleverly chosen using specific algorithms, in order to properly cover the design space while minimizing the amount of analyses.
Both methods can be used to build response surfaces. This is a way to interpolate the results obtained by running analyses and it ends up with a approximate description of the results over the complete design space. These approximations provide fast analysis tools. They can be used for rapid sizing studies in the early design stages.
- Multidisciplinary optimization is the way to compute the optimal set of parameters according to the criteria you define. BOSS quattro features several family of optimization algorithms:
o Gradient based methods (GCM, Conlin, SQP...)
o Derivative free methods (Genetic algorithms, Surrogate-based Optimization...)
- Sensitivity analysis is used in the gradient based optimization methods but it is also meaningful when applied out of an iterative optimization process. For given values of the parameters, it is describing the influence of each parameter on a list results.
- Statistic analyses (Monte Carlo) are used to run parametric studies with the sets of parameters built using statistical distributions (Gauss for example). One of the main use is to study the influence of production tolerances and make sure that the behaviour of the structure remains acceptable over the full range of potential values for the variables.
- Updating: Also called “reverse engineering” or "least square", this engine is used to fit modeling results to experimental data. It adjusts the values of the parameters to obtain this match. Material properties are the perfect example for this application.
It is worth mentioning that BOSS quattro is an open platform. This openness can be noticed at two different levels:
- While it has a preferred link to SAMCEF solvers, BOSS quattro can eventually use almost any kind of applications where parameters are input and results output. The most common FEA packages, as well as MS Excel or even your own in-house tool can be easily coupled to BOSS quattro. Several tasks, potentially linked to different disciplines (structure, fluid, heat transfer, costs...) can be run in parallel and BOSS quattro will find the optimum common to all of them.
- It is possible to link your own optimization algorithms to BOSS quattro. You are as such not obliged to let aside your favorite tool that is so well suited for your application. BOSS quattro is in that case providing you a professional infrastructure to manage your models, results, task flows... while using your own optimization solver.