Continuing in the direction of providing an end-to-end solution to our customers, we have recently partnered with Friendship Systems, developers of a slick optimization software called CAESES. More and more engineers are using computational power for optimizing their designs and conducting parametric studies and design of experiments (DOE). The availability of computational resources (cloud computing included) and advances in the field of optimization algorithms make it increasingly feasible to use optimization software in a design cycle.
While CAESES finds application in a multitude of industries, including shipbuilding, turbine design, and the energy sector, I will be discussing a case study on optimizing a sediment scour problem using CAESES and FLOW-3D. CAESES offers advanced optimization algorithms based on genetic algorithms and artificial neural networks. It also offers simpler and faster algorithms for quick and easy problems. CAESES can be run in parallel for machines with multiple cores, which addresses the inherent problem of computational cost involved in optimization studies. Also, a seamless connection can be established between FLOW-3D and CAESES, such that once the problem statement is fed into CAESES along with any associated files, the remaining process is absolutely automatic.
Sediment scour optimization
For sediment scour modeling problems, maximum scour depth and height are of interest to the modeler. The three main design parameters that affect the results of a sediment transport problem in FLOW-3D are the critical Shield’s number, bedload transport coefficient and the entrainment transport coefficient. Unless an optimal mesh is used, the values of these parameters need to be calibrated (from the default values) to get the correct results. The design space of a standard sediment scour problem is described below. The problem is selected from the Chatterjee report and can be set up in CAESES to obtain a set of design parameters to obtain the maximum and minimum scour elevations values that match the experimental values.
The Chatterjee report provides the experimental scour depth profiles at various times (figure below). The profile at 8 minutes was chosen for the case study. The minimum and maximum elevations (objectives) for the 8 minute profile are -0.05 m and 0.05 m, respectively.
The case study has a very coarse (non-optimal) mesh. Therefore, the default parameters in FLOW-3D do not give good results. However, once the parameters are optimized, very good results can be obtained even for a coarse mesh. Notice that the optimized parameters give excellent results of -0.049 m and 0.0505 m for minimum and maximum elevations respectively. A comparison of results is shown below.
FLOW-3D in combination with CAESES is a powerful combination for conducting such optimization and DOE studies. As mentioned earlier, the list of applications for such studies is long and more examples will be covered in an upcoming webinar on August 25, 2016 at 1 pm EDT.