Numerical Optimisation

To find the best solutions to complex problems Prototech applies numerical optimisation algorithms in combination with simulation and modelling.

Numerical Optimisation

In the illustration above, a winglet is optimised for a wind turbine rotor blade. Here, the best performing winglet is obtained by constructing a Kriging surrogate model which is refined and optimised using an expected improvement infill criterion and a hybrid genetic/gradient optimisation algorithm. The turbine performance is simulated by solving the incompressible Navier-Stokes equations and the turbulent flow is predicted using a Reynolds-Stress Model. The optimised winglet increases the power production for the rotor blade by about 8%.

At Prototech different state-of-the art optimisation algorithms are used depending on the problem;

  • When the solution space is smooth, unimodal and gradient information is available local gradient-based search methods are the most efficient.
  • When the solution space is smooth, unimodal and gradient information is not available local gradient-free search methods are the most efficient
  • If the solution space is strongly multi-modal or if numerical noise exists in the solution, global optimisation algorithms are the most efficient.
  • For problems where the simulation time is very long, surrogate models are the most efficient approach.

Prototech and CMR has also a broad modelling and simulation expertise with in-depth application knowledge to effectively describe and understand different systems and processes. Furher information can be found here.