In this presentation, steps to generate reliable Response Surface Models based on Computational Fluid Dynamics (CFD) simulations are explained with two applications: Missile aerodynamics and store separation. For this purpose, commercially available software, i.e. modeFRONTIER of ESTECO, FLUENT of ANSYS, and MATLAB/Simulink of MathWorks, are utilized. Starting from an initial distribution of design points that are decided by a Uniform Latin Hypercube algorithm, a Sequential (or Adaptive) Sampling method is followed to improve the final Response Surface Models. The quality of the Response Surface Models are reported by monitoring the convergence histories of mean normalized errors that are computed using previously simulated design points. In the first application, aerodynamic forces and moments of a missile are obtained with Response Surface Models and compared with the results of Computational Fluid Dynamics analyses for some cases. And in the second application, one of the well-known safe separation test cases, EGLIN Experiment, was simulated by using aerodynamic forces and moments under the wing obtained from Response Surface Models. In order to obtain aerodynamic forces and moments under the wing, a way similar to the Grid Method was followed to generate Response Surface Models within the predefined domain. The outputs of Response Surface Models are used by a 6-DoF code. As a result, the trajectory, attitude, linear velocities and angular velocities of the model were predicted and compared with the experimental data and the results of unsteady CFD analysis.

Safety-critical real-time embedded systems software saw great improvements by the use of model-driven development. Unified Modelling Language (UML) - once appeared to meet the need for a common modeling language - was augmented with composite structures and ports and is now more suitable for modeling of software for such systems. Industry responded to updates in the UML by incorporating support for UML ports and composite structures. IBM Rhapsody for example, is widely used by the embedded software community and supports UML ports. Modeling using UML ports works great, until the code is generated and tried to be manually reviewed. The problem is, when realizing UML ports in object oriented languages, the standard approach is to implement them using dedicated objects. However, this direct approach causes runtime overhead and results in significant amount of additional code generated by ports. Pin-sharp performance requirements and painstaking code reviews to ensure safety are indispensable facts of developing avionics software. In order to meet the performance requirements and to decrease costs and increase accuracy of code reviews, more efficient approaches are needed for realization of UML ports in modern programming languages. This presentation presents a solution, which introduces relatively less runtime overhead and results in dramatically more compact source code. We use a model to model transformation, where a software model that uses UML ports is transformed into a model that uses associations instead of objects to efficiently implement UML port semantics with less lines of code. Preliminary results from a case study demonstrates that the theory does work in practice, too.