Skip to content. | Skip to navigation

Log in

Subscribe to our newsletter


Case Study


Hydrodynamic Optimization of the Aftbody of a Container Vessel Using Viscous Flow CFD Methods

The quality of a ship's wake field is crucial for the occurrence of cavitation in the propulsion system, induction of vibration and pressure pulses on the hull structure, but also for the propulsive efficiency. When it comes to optimizing a hull shape with respect to the wake quality, the assessment of design variants demands the application of RANS solvers, since the viscous flow effects are predominant. Key prerequisites for an efficient design process are thus the integration of viscous flow solvers in an automated process and the reduction of computation time.

In the present case study the applicability of viscous flow, i.e. RANS methods for the simulation-driven design of aftbody shapes, was investigated. A fully parametric model of a container vessel was used and the target of the optimization was an improvement of the ship's wake field quality. The FRIENDSHIP-Framework served as modeling, as well as integration and optimization workbench, while SHIPFLOW's RANS solver Chapman was used to compute the wake field. The case study was conducted as a complement to the R&D project OptiHULL, a collaboration between the Thyssen-Krupp Nordseewerke shipyard, FutureShip and FRIENDSHIP SYSTEMS. The OptiHULL project aimed at the efficient parametrization, assessment and optimization of aftbody shapes. It started in May 2008 and was finalized in June 2009.

To describe the hull shapes, a new fully parametric surface model (see Fig. 1), was developed and employed. The complete hull consists of separate models for aftbody, parallel midbody and forebody, linked to each other by a set of global parameters and a main frame description. The transfer of the geometry for the grid generation to SHIPFLOW is attained through the export of offset data. Using the FRIENDSHIP-Framework's feature technology, a routine was written that automatically generates new suitable offset groups for SHIPFLOW each time the surface geometry is updated. The export is integrated in the configuration for the SHIPFLOW computation inside the FRIENDSHIP-Framework.


[Fig.1] Surface model of the ship hull

The three design variables chosen for the aftbody investigation were the fullness parameters for curves diag1, diag2, the lower and middle diagonal, respectively, and the fourth part of the centerplane curve, which defines the lower contour of the bulb between the points bulbTip and aftBase (see Fig. 2 ).


[Fig. 2] Basic curves for the aftbody model

As a first step the converged solution for the baseline variant was computed with the RANS solver Chapman, following SHIPFLOW's zonal approach: potential flow and boundary layer solution are computed for the complete ship and used as input for the viscous flow solution of the aftbody flow, which starts from midships. The computed wake field was compared to model test results, showing good agreement. In the next step a Sobol design engine was set up for a design of experiments with 50 variants. The Chapman solver was set to restart and the baseline results were chosen as initial solution for the new computations. In comparison to the 3000 iterations of the baseline solution the number of iterations for each variant could then be set to a considerably lower value, thus reducing the total computation time to ~70 hours on an average office PC.

In Fig. 3 the wake quality measures provided by SHIPFLOW are plotted against the three design variables, diag1fullness, diag2fullness and cpcPart4fullness, going from left to right. The upper diagrams show the maximum wake variation on one radius as function of the design variables, the lower diagrams the wake fraction. The maximum wake variation is an indicator for the uniformity of the wake field, therefore minimizing the wake variation is advantageous for the propeller design. The Taylor wake fraction is an indicator of the flow retardation through viscous effects, meaning that the resistance and the propeller loading increases with the wake fraction. On the other hand a higher wake fraction is beneficial for the hull efficiency. Following these arguments it was decided to regard a design as "good" if it has a low wake variation and/or a high wake fraction.

 optimizationwakevariation1_.png  optimizationwakevariation2  optimizationwakevariation3
 optimizationwakevariation4  optimizationwakevariation5  optimizationwakevariation6

 [Fig. 3] Plots of quality measures as derived within the FRIENDSHIP-Framework

The plots show that cpcPart4Fullness has only little influence on the results, while the impact of the fullness of the two diagonals is clearly more pronounced. An increasing fullness of the lowest diagonal reduces the maximum wake variation and raises the wake fraction. The fullness of the middle diagonal has an inverse effect on the wake variation, while the effect on the wake fraction is not very distinct. In general it can be said that the tendency for improvement would be to decrease the fullness of the lower bulb contour and the middle diagonal, while increasing the fullness of the lower diagonal. The latter modification being the most beneficial, also with regard to a gain in displacement volume.

Four designs were selected out of the generated variants; the ones with the lowest and highest maximum wake variation and wake fraction, respectively. These variants are shown in comparison in Fig. 4. The variant with the lowest wake variation and the one with the highest wake fraction show a similar value of the respective other parameter, leading to the conclusion that, at least in this case, either one of the two quality measures could be suitable as target function for a further optimization.


[Fig. 4] Comparison of the four selected variants

The computed wake fields of the four selected variants are shown in Fig. 5. They look similar for the best and worst designs by means of the wake variation and wake fraction, respectively. One can see that the increase of fullness of the lower diagonal leads to a more circular distribution of velocities.


Des004: lowest max. wake variation

wake des029 small

Des029: highest max. wake variation

wake des014 small

Des014: highest wake fraction

wake des033 small

Des033: lowest wake fraction

 [Fig. 5] Wake fields of the four selected variants, colored by axial velocity. Surface streamlines in the propeller plane show the location of vortices

This test case showed that viscous flow methods can be effectively used for the hydrodynamic design and optimization of aftbody shapes, especially in connection with a fully parametric hull model. Typical obstacles of RANS computations like volume grid generation and adaptation are fully automated, and the computation time can be regarded as very reasonable for the reward of getting a good feeling of the potentials for improvement, without the need to resort to expensive model tests.

The benefits that can be achieved using this methodology are: reduced risk through well founded decisions during the design process, better performance of the final design, reduced costs for standard model tests and a lower time to market with an increased level of knowledge.

For more information please contact Mr. Mattia Brenner.

Document Actions