Practical Ship Hydrodynamics

Introduction 21
performed on workstations or fast PCs with high-resolution screens. User
experience and a degree of automation mainly determine time (and labour
costs) in this step. Experienced staff can expect grid generation to be the
major part (30% to 90%) of the man time involved in CFD analyses.
Completely automatic procedures for complex geometries such as ships
are not available and do not seem realistic for the future. Staff training and
software development/adaptation are the main fixed costs which usually
surpass depreciation of hardware and software by an order of magnitude.
2. Computation
The computation involves almost no man time. Computations for inviscid
CFD can usually run on PCs; viscous CFD may prefer more powerful
computer environments up to parallel computers/supercomputers depending
on the problem size. Often workstations are the preferred choice. Computing
costs usually account for less than 1% of total costs. Also software licences
for the flow code are often negligible compared to other costs, especially
those for generating the grid.
3. Postprocessing
The graphics require fast PCs with colour screens, laser printers and colour
plotters/printers. The necessary software is commercially available. Postprocessing
requires some time (typically 10% to 20% of the total time), but
can be performed after short training by staff without special qualifications.
User friendliness of the programs determines time and thus labour costs
in this step. Use of postprocessing programs should be kept as simple as
possible minimizing user input. Interpretation of results still requires expertize
and is a lengthy process. You pay thus for the interpretation, not the
number of colour plots.
The high fixed costs for training and user-defined macros to accelerate the
CFD process lead to considerable economies of scale. This is often not realized
by management. Experience shows that many shipyards buy CFD software,
because the hardware is available or not expensive, and the software licence
costs may be as much as a few CFD analyses at a consulting company. The
vendors are naturally only too happy to sell the software. Then the problems
and the disillusion start. Usually no initial training is given by the vendor (or
bought by the shipyard). Typical beginners’ mistakes are the consequence:
ž Time is lost in program handling.
ž Unsuitable grids are used requiring repeated analyses or resulting in useless
results.
By the time the management realizes the problems, it is usually too late. The
software licences are all bought, the design engineer has now already invested
(lost) so much time struggling with the code. Nobody wants to admit defeat. So
the CFD analyses are continued in-house with the occasional outsourcing when
problems and time pressures become too large. As a general rule, outsourcing
is recommended for shipyards and design offices with fewer than five projects
per year. In-house CFD makes sense starting from ten projects per year. These
numbers may shift if CFD codes become more user-friendly, i.e. run almost
automatically. However, for finite-element analyses of structures we have seen
a development that after an initial period where shipyards performed the analyses
in-house the pendulum swung the other way with shipyards now using
almost exclusively outsourcing as the sensible option.