Variant generation

Variant generation and quality assessment: Meshes and pressure distribution of different volute designs

Variation. In the quest of identifying the optimal design, variation certainly has top priority. For how could you possibly pick the best option without insight into the variance available? Variant generation means looking for the most suitable design in the design space as is defined by your design requirements. You determine factors like size, maximum load, speed and draught—variables, constraints and objective functions, which your design, as representation of the final ship hull, propeller or pump, has to meet. Based on your design frame, the FRIENDSHIP-Framework produces a range of suitable designs with the aid of variation algorithms. Such are, for example, exhaustive search, ensemble investigation and Sobol (design of experiments). Maybe the best part in the variant generation within the FRIENDSHIP-Framework is the resource output. The system generates hundreds of variants in overnight studies. You start say the exhaustive search when leaving your office, you get back in the morning—and voilà! Graphical overviews, tables and autgenerated pdf display the results of the night's work. The computational time can be reduced further with distributed computing, a technical improvement, which we have made available from release 2.1. Distributed or remote computing employs idle hardware and helps streamline the process chain.

Additional features of the variant generation  within the FRIENDSHIP-Framework are:

• Intuitive resource manager,  which distributes simulation analysis on cluster systems. Both synchronous and parallel computations can be operated.
• CFD calculations are configured and triggered from within the FRIENDSHIP-Framework. Externally computed results are brought back for assessment.
• Constraint management: Equality and inequality constraints can be defined. Considered constraints are monitored and marked for evaluation.
• Variant management: The entire data management for manual and automatic variant creation and assessment is provided.

Once you have selected the most suitable design candidates, subsequent optimization helps you to get the best from your model.