Computer simulation in shipbuilding
HPC (High-Performance Computing) methods are successfully used by T-Services engineers for computer simulation of flows around ships and ship models taking into account the free surface and wave generation. Several simulation stages can be specified:
- Development and analysis of solid-state and surface geometrical models of ships using various engineering packages;
- Mesh generation for hydrodynamic simulation in various CAD/CAE packages. Different approaches and mesh topologies can be used. For example, block hexa- and tetrahedral meshes, meshes with polyhedral cells etc. can be utilized for different geometries;
- Modern CAE/CFD packages and HPC technologies must be tuned for simulation of external hydrodynamics and towing resistance of a ship. This tuning includes specification of initial and boundary conditions, turbulence models, estimation of the accuracy and convergence of the simulation, including guaranteed accuracy in force and moments calculations;
- Computer resources use and its relation with simulation accuracy for typical ship hydrodynamic problems is a problem for separate study.
T-Services are developing a new system for computer simulation in shipbuilding based on HPC technologies. The system includes high-performance computer equipped with codes and mathematical libraries that are necessary for numerical simulations of ship hydrodynamics and screws.
In this article the process of high-performance simulation of a ship hydrodynamics is illustrated by calculations of towing resistance of two ship models. The geometry of these models (“support” and “dry cargo” ships) was given by Design bureau “Vympel”, Nizhniy Novgorod. The simulations were carried out on a supercomputer system developed by T-Platforms. The standard T-Blade unit with 5U height contains ten 2-processor high-performance blades based on 2.33GHz Intel(R) Xeon(R) E5410. The blades are connected with high-performance infiniband 4x interconnect.
Supercomputer technology use for ships hydrodynamic modeling is obvious because personal computer resources are insufficient. During all four stages mentioned above the use of supercomputer is necessary.
Geometrical characteristics of ship models were given in an STL file. This file format is widely used for 3D object prototyping and storage. The technology uses the so-called stereolithography method when an object is stored as a list of triangular faces.
The surface of one of the ships was described with more than 300 000 triangles. The opening and processing of such file in a CAD system for consequent mesh generation requires several hours or even tens of hours on modern high-performance processors due to absence of efficient parallel algorithms of CAD model import. Moreover, working with such large datasets requires powerful video cards and large amount of memory.
Surface and volume mesh generation requires good qualification of an engineer, powerful computer and is very time consuming. High quality meshes can contain several millions or tens of millions cells. For simulated ship models the hexa meshes with following characteristics were generated:
|
Model «support» ship |
Model «dry cargo» ship |
||
| Number of cells | 4320072 | Number of cells | 5981730 |
| Mesh quality | 0.333 | Mesh quality | 0.356 |
| Minimal cell angle | 19.62 | Minimal cell angle | 20.88 |
| Aspect ratio | 1230 | Aspect ratio | 149 |
*) Mesh quality is defined as: for triangular meshes it is defined as a minimum of the ratio of height to a base side; for tetrahedral meshes – by elongation of elements; for quadrangular, hexa and prismatic meshes mesh quality is defined as a ration of two determinants – minimal and maximal jacobeans. Theoretically the best mesh quality is equal to 1. Usually meshes that are widely used in simulations have qualities of 0.2-0.3 and higher.
The whole cycle of development of a new ship model that includes numerical simulation and calculation of integral characteristics of a ship with high precision usually must be done in 15-20 days. During this time it is necessary to study ship hydrodynamics with 10-15 different speeds of towing and/or other far field parameters. We performed flow simulations in 9 regimes in the velocity range from 0.7 to 2.1 m/s and for 2 different initial intensities of turbulent pulsations (0.1 and 5 %). It is impossible to use a personal computer for such simulations due to limitations of memory and processor power. Only use of a supercomputer with parameters that are mentioned above allowed performing simulations in realistic time span (Fig. 1-5).
|
|
|
|
Fig. 1. Mesh on the surface of a «support» ship |
Fig. 2. Mesh on the surface of a «dry cargo» ship |
| Model «support» ship | Model «dry cargo» ship |
Fig. 3. Pressure distribution of the side surface and
the bottom of the ship (far field velocity 2.1 m/s)
| Fig. 4. Pressure distribution on a free surface for «support» ship (far field velocity 2.1 m/s) | Fig. 5. Shape of the free surface for «dry cargo» ship (far field velocity 2.1 m/s) |
Fig. 6 shows the dependence of towing resistance (in Newton) on the speed of towing for “support” and “dry carrier” ship models. The experimental data obtained in Krylov Shipbuilding Research Institute is shown by solid lines. The results of numerical simulations are shown by markers – hollow dots are for the turbulent pulsation energy of 5 %, solid dots – for 0.1%. Performed parametric studies reveal a strong influence of the initial turbulent pulsation energy on the value of towing resistance (up to 25-35% depending on the rowing speed).
The best agreement with experimental data is achieved for the initial turbulent pulsation energy of 0.1%. This value must be close to experimental conditions for towing in stagnant water..
| Model «support» ship | Model «dry cargo» ship |
- experimental data on stagnant water (Krylov Shipbuilding Research Institute),
- simulations (initial turbulent pulsation energy 0.1 %),
- simulations (initial turbulent pulsation energy 5 %),
The utilization of supercomputer resources and HPC technologies allows to achieve high accuracy in determination of towing resistance and to reveal main local flow patterns. Calculated values are corresponding with experimental data on ship towing on stagnated water for the low level initial turbulent pulsation energy.
The results show the high accuracy in physical and computational domains received in reasonable timeframe. It means that in the nearest future HPC technologies could become a standard tool for the prototyping and research of different kind of ships.



