Industrial Applications of High-Performance Computing

279

Chapter 20

An Italian Twist on

an America’s Cup

Sailboat Hull Design

Raffaele Ponzini and Andrea Penza

20.1 INTRODUCTION

To win the America’s Cup [1], one must defeat the other teams not only

in regattas but also in design studies, a strongly technological challenge

where a major role is covered by computational uid dynamics (CFD).

As is common in most computer-aided engineering (CAE) industrial

applications, soware licenses are a major cost item; moreover, it is

well known that in-silico data acquisition (i.e., data obtained from

computational models) is becoming the elective environment for the

design development process, involving multiphysics CAE tools and a

dramatic increase in requested high-performance computing (HPC)

resources. As a byproduct, budgets related to “computing” activities are

increasingly at risk of becoming the real bottleneck in the technological

challenge. Open-source CFD soware might become a key point to

CONTENTS

20.1 Introduction 279

20.2 Adoption of HPC: Insights into Portability and Scalability 280

20.3 HPC Performance Evaluation and Benets 282

20.4 Future Developments 282

References 283

280 ◾ Industrial Applications of High-Performance Computing

overcome such a bottleneck; nevertheless, large campaigns are required

to validate digital models against physical models to ensure reliability

for industrial-use cases of interest. e problem is complex in its entirety

and it is crucial to have access to advanced computing resources that are

able to process the large amount of data produced by such design studies

considered as a whole and by each single numerical simulation. For this

purpose, the design team of the Luna Rossa Challenge [2], besides taking

advantage of large amount of HPC resources for design data production,

both in terms of computing and remote visualization, undertook

with CINECA in 2012, a 12-month feasibility study evaluating the

OpenFOAM (Open Source Field Operation and Manipulation) toolbox

[3].

20.2 ADOPTION OF HPC: INSIGHTS INTO

PORTABILITY AND SCALABILITY

e validation project presented herein has been designed to study

both state-of-the-art AC72 hulls, those that ended up competing in the

San Francisco Bay waters in August 2013, and well-known benchmark

hulls, fully validated through computational tools and towing tank

experiments. To provide fruitful answers, we tested both the accuracy

of the open-source solution against other commercial CFD codes and

its computational eciency concerning the time-to-solution at dierent

degrees of parallelism.

As a rst attempt, we studied the hull hydrodynamic of a reference hull

known in literature as DTMB–5415, a model conceived by the U.S. David

Taylor Model Basin (DTMB) in the early 1980s as a preliminary design

for a surface combatant ship with a sonar dome bow and transom stern.

For intellectual property reasons, all the results detailed in this article

concern this hull type. Suce to say that we found that the same general

conclusions may be applied also to the racing hulls in design.

e Navier–Stokes physics equations that govern the motion of uids

are particularly complex to solve. Only in very rare cases, and with simpli-

ed geometry and conditions, is it possible to obtain an analytical solu-

tion. erefore, digital numerical methods are considered the only feasible

way to solve such complex equations in real-life geometry and ow con-

ditions. e library used for this validation work, OpenFOAM, is based

on breaking down, or discretization, of the problem obtained through

the so-called nite-volume method. e equations are strongly nonlinear

and coupled, which makes the search for a solution even more complex,