ship design and the computer - repository.tudelft.nl
TRANSCRIPT
ARCH1EF
SHIP DESIGN AND THE COMPUTER
Horst Nowacki
s% TV p".A.(4, 11, IP
.... A 1.1am == ......>
THE
DEPARTMENT OF NAVA
ARCH/TEC
Lab. v. Scheepsbouwkunde
No. fahnische HogeschoolDecember 1969 Delft
THE UNIVERSITY OF MICHIGAN
COLLEGE OF ENGINEERING
ARINE ENGINEERING
SHIP DESIGN AND THE COMPUTER
BY
HORST NOWACKI
-DEPARTMENT OF NAVAL ARCHITECTUREAND
MARINE ENGINEERINGTHE UNIVERSITY OF MICHIGAN
ANN ARBOR, MICHIGAN
EASTERN CANADIAN SECTIONSOCIETY OF NAVAL ARCHITECTS
ANDMARINE ENGINEERS
MONTREAL
DECEMBER 2, 1969
" I 1Mdeling Sch
ibliotheek van de
nische Hogeschooeepvaartkunde
DAT UMI
DOCUMENTATIE 42
ABSTRACT
The paper discusses current trends in computer-aidedship design, and the effects computer use may have on thescope and style of ship design.
INTRODUCTION
Ship design is without doubt receiving strong, fresh
impulses from the rapid advances of computer technology.
We are witnessing a process of mutual adaptation: Computer
systems and new computer methods are providing a growing
variety of the capabilities essential to ship design. In
ship design, on the other hand, a new, somewhat more rational
and systematic style is developing where the computer is
used as a tool.
This paper will attempt to evaluate the present stage
of developments, and to discuss where current trends seem
to be leading. The purpose is not making an exact predic-
tion, but rather stimulating a few thoughts and a clarifying
discussion before this professional audience of what lies
ahead for ship design.
DESIGN OBJECTIVE
The goal. of z design process may be defined in terms
of input and output as follows:
Given a functional requirement (say, a certain amount
of cargo shall be transported from A to B); given further
a number of constraints of technical, physical, or legal
nature (stability, strength, ship safety, classification
rules; etc.).
Sought an optimal technical solution judged on the
basis of a concrete measure of merit.
This comprehensive definition of design, which results
from applying the thinking of systems analysis to design,
goes probably beyond the notion of ship design many of us
have today. Up to now the designer has frequently been
satisfied demonstrating the feasibility rather than the
optimality of a technical solution. Moreover, merit compar-
isons in the course of a design have not always been based
on a single, clearly expressed measure of merit.
Computer-aided design, too, is far from the ideal of
completely rational, optimal solutions. But the following
will show that the use of computers creates the tendency to
more ambitious, comprehensive design objectives.
1
The economic success of a technical venture is generally
measured by the ratio of benefit and effort. To the owner
of a merchant ship, profit is the desired benefit, and the
size of the investment measures the effort. Their ratio,
the profitability, is to be maximized.
We need not discuss in detail how to express profita-
bility in ship design. There are many, more or less equi-
valent possibilities (yield, required freight rate, etc.),
but there is no reason to be dogmatic since, within reason-
able variation of the criterion, the technical solutions
obtained are fairly equivalent, as Benford has frequently
stated, References (1) and (2).
To illustrate how a suitable measure of merit provides
a common perspective to all design considerations, let us
look at the Ship Merit Factor, SMF, recently proposed by
Cheng, Reference (3). One of the versions he gives for
SMF is
VWpSMF = k
1 1 WP 1 RV= 8760 fs fw fv r-T-r- W R P
where
SMF = ship merit factor
k = 8760 fs f f /(1 + f) = service constantw v
W = payload
V = design speed in knots
C = AAC = average annual cost
8760 = 24 365 hours/year
RATIONAL DECISION CRITERIA
( 1 )
fs = utilization factor, percent of annual
service hours
fw = load factor, percent of designed payload
fv = operating speed factor, percent of design
speed
f = port time factor, port time/sea time
C' = C/PB = specific operating cost in dollars
per horsepower-year
PB = power delivered by prime mover
W /W = payload-displacement ratio
R/W = drag-displacement ratio
RV/PB = no nH ript s = propulsive efficiency
,n ,nR,nS = open water, hull, relative rotative, and
shaft transmission efficiencies
The Ship Merit Factor is defined as the inverse of the
well-known Required Freight Rate (RFR). The essential point
is that equation (1) allows judging the contributions of
various design measures in different areas to the success ofthe whole, and selecting the most promising changes. It is
evident, for example, that good cargo handling and a fast
turnaround are as significant as design improvements in the
more traditional domains of resistance, propulsion, and
strength. Of course, each gain must be related to the
corresponding cost, that is the effect on C'.
CLOSED LOOP OPTIMIZATION VERSUS OPEN LOOP DESIGN
The computer enables us to generate and compare a
much greater number of design variants than possible before.
The question arises naturally how to go about a systematic
evaluation of this variety and how to pick a best solution.
The effort of evaluating all conceivable variations of the
design exhaustively is generally prohibitive even when com-
puters are used.
This is why several relatively efficient optimization1
methods have been applied to design problems in recent years, 11
which tend to converge to the optimum automatically, and 1
fairly fast and selectively, on the basis of a given measure 1
of merit function. (Linear, and nonlinear programming,
random search, direct search, and other search methods). 1
Applications of optimization methods to early stage ship !
design (selection of principal dimensions), and ship struc-
tural optimization (midship section) are probably well-known,
References (4), (5), and (6).
At the University of Michigan, we have recently success-
fully adapted the direct search technique, Reference (7), and
the SUMT method by Kowalik, Moe, and Lund (Sequential Uncon-
strained minimization Technique, References (8), (9)) to the
problem of preliminary ship design. Meyer-Detring (10) has 1
presented a tanker design application, and a variety of other 1
studies are underway.
The use of optimization methods in closed form requires
that the nature and scope of the intended design variations1
be known when the program is written, that is before the 1
design has begun. In the typical operation of a batch1
computer, the user has no control over the computation during the
execution of a run. Consequently, where batch computers are I
used in design optimization the designer is forced to formulate
- 4 -
explicitly, and ahead of time, his design objectives, the
measure of merit, and the range of design variations.
This may have some beneficial effects upon the ratio-
nality of our design decisions. But it is also clear that
in this kind of computer use we have to break with the
traditional style of design,. This may be illustrated by
the flow chart of Figure 1 for the example of early stage
ship design.
In this figure, the arrows and incomplete loops shall
indicate the "open" stiucture of conventional design pro-
cedures. The sequence of steps is actually very flexible
and hardly predictable. It will be decided by the designer
on the basis of intermediate results in such a manner that
one obtains the most suitable solution by gradual trial and
error improvement, and with the least possible effort. The
designer always reserves the right to learn from the inter-
mediate results and to enter new thoughts into the design as
it proceeds.
The "open" logical structure and freedom of traditional
manual design thus is not ip harmony with the "closed" format
of optimization by a single computer run. Consequently,
optimization methods have been successfully applied only to
certain subproblems of ship design, for example the deter-
mination of principal dimensions.
A new development is coming about now with the intro-
duction of time-sharing computer systems which permit
continuous access to the computer, even during execution of
the program, and hence facilitate a free dialog between
computer and user. Although this technical development is
still in its early stages it can already be concluded that
time-sharing and dialog are very suitable media for design
tasks.
This suggests a compromise in the future computer use
in design such that the designer controls the gradual step
by step development of the complete design, at the same
GIVEN: ROUTE, PAYLOAD, (SPEED)FIND: SIZE, PROPORTIONS, (SPEED)
V
ESTIMATE A, SPEED, ANDPRINCIPAL DIMENSIONS
POWER ESTIMATE: SHP
-WEIGHT CHECK 1
LINES AND ARRANGEMENTS
FREEBOARD
CHECK OF _CAPACITIES
STABILITY, TRIM, AND MOTIONS
STRENGTH
MEASURE OF MERIT
I OPTIMUM I
FIG. I FLOW CHART OF PRELIMINARY DESIGN
THE SYSTEMS APPROACH IN SHIP DESIGN
In the present context let us define a system as a
number of objects or activities, serving a common purpose
and being judged by a common measure of merit. Where ship
design decisions are subjected to rational criteria, such
as profitability, it follows naturally that the ship and
its parts are viewed as a system. Moreover, many ship
types cannot be designed effectively today without devot-
ing much attention to the interaction between ship and shore
based handling and distribution facilities, and to several
other details of ship operations. In this connection the
ship is viewed as part of a more comprehensive system.
For these reasons the organization of ship design
studies follows the systems approach tore and more fre-
quently. Such systems studies tend to be far more elaborate
than used to be the case in conventional design, which makes
computer use all the more imperative. Figure 2 illustrates
the possible scope of systems design studies, and the variables
and parameters that may be involved. The figure shows that
beside the technical ship design variables many other details
of the system have to be determined in mutual harmony where
they are under the system designer's control (routing, sched-
uling, cargo handling, details of ship operations). Moreover,
there are many external influences on the system that are of
uncertain nature and best represented statistically (cargo
availability, weather and seaway, other causes for delays).
Dealing with such influences properly, necessitates stochastic
decision models.
The systems approach not only causes these additional
complexities, but fortunately also provides the means for
problem solution. Generally the problem is first decomposed
into its simplest elements following a formal pattern. Never-
theless the analyst will benefit from a thorough technical
-8
TI
CARGO PARAMETERS:STOWAGE FACTORS,
AVAILABILITY,SPECIAL REQUIREMENTS
TECHNICAL DESIGNVARIABLES OF THE SHIP
DISPLACEMENT, SPEED,LENGTH, BEAM, DRAFT, DEPTH,C, C, , ETC.
PORT PARAMETERS:TERMINAL FACILITIES,
CARGOHANDLING METHODS,PHYSICAL PORT LIMITATIONS,
DELAYS, AND OTHERUNCERTAINTIES
Is- II
MATHEMATICALMODEL OF SYSTEM
MEASURE OF MERIT
IOPTIMIZED SYSTEM
VOYAGE PARAMETERS:TRADE ROUTE,
UNCERTAINTIES, AND DELAYSEN ROUTE, ETC.
COST PARAMETERS:SHIP BUILDING AND
OPERATING COST,COST FOR TERMINALS,AND OTHER PARTS OF
THE SYSTEM
FIG. 2: FORMAT OF SYSTEMS STUDY IN SHIP DESIGN
understanding of the system. It remains an essential
engineering talent to discover how to reduce a complex
problem to many simpler, more tractable ones.
In the following step the system is optimized by
means of standardized, flexible, and powerful optimiza-
tion methods, often provided as subroutines by computer
systems. The crucial role Of the computer in this kind
of work should be obvious.
- 10 -
DATA STRUCTURES
The growing scope of design studies puts no small
demands on computer storage and cost economy. It is
therefore important to organize the handling of data in
the computer efficiently. Modern computer systems offer
many features facilitating efficient data structures. The
following are of particular importance.
Random access:
Disk, data cell, and similar secondary storage devicesallow direct storage and retrieval of each data elementas opposed to the time-consuming scan required in se-quential storage devices (tapes).
Dynamic storage allocation:
In classical FORTRAN style computer software the alloca-tion of storage is done by the programmer rigidly beforeprogram execution. In certain modern software environ-ments the system itself is responsible for the allocationof storage and dynamic updating during execution. Thisdynamic allocation feature avoids the waste of idle storagespace.
List structures:
Contrary to the usual array structure of logically con-nected data elements in successive storage locations inthe computer memory, list structures allow placing thedata elements at arbitrary locations in storage, providingthe logical connections by pointers associated with eachdata element. The logical and the physical sequence ofdata in storage are thus independent of each other. Thisfeature is essential to update data scopes convenientlywithout relocating any data elements.
Time sharing:
User control during program execution may lead to essen-tial cost savings in debugging and whenever intermediatechecks are advisable to keep large programs from takingoff tangentially.
The advances in computer technology and systems programming
discussed in the foregoing have been sufficient to meet the
growing demands of the design-oriented user.
COMPUTER GRAPHICS
The picture is an essential ingredient of engineering
design. Pictures convey ideas spontaneously, much more
immediately than numbers. A design idea is often born as
a ,visual concept, and must be displayed graphically to be
communicated easily and uniquely.
Many computer systems today are strictly analytically
oriented. The user must encode digitally all geometric
input data, and decode the digital output for geometric
meaning. At this level there exists a conspicuous mis-
match between computer and user.
This deficiency is being removed by computer graphics
providing graphic input and output devices. Digital plotters
have been playing a useful role for some time (passive
graphics, just output), whereas active graphics by cathode
ray tube display and light pen control is beginning to
emerge from its experimental stages.
The processing of geometric data is extremely inten-
sive in the shipbuilding industry, and it is obvious that
computer graphics are going to play an eminent part in all
phases of shipbuilding from early stage design through
production, Figure 3.
Graphic output on plotters is already an essential feature
in lines fairing applications and as a check medium for nu- i
merically controlled flame-cutting equipment. The Norwegian
AUTOKON system, widely usedthroughout the world, uses
graphics intensively in both applications, and demonstrates
the key role of graphics in any all-round design-production
system. The CASDOS ship detailing system, under development
for the U. S. Navy by Arthur D. Little, Inc., also depends
heavily on its original solutions in graphics. More details
about the two systems below.
CHECK, OPTIMIZE 1
WORKING DRAWING
PRODUCTION PROGRAM
FIG. 3: ROLE OF GRAPHICS IN THE COURSE OF DESIGN
The use of computer graphics in the shipbuilding
industry has been directed to ship production primarily;
but other industries have already made significant
progress in design applications of graphics. This is
certainly true for the developments of the U.S. aircraft
and automotive industries, References (11) and (12).
The foregoing shall not give the impression that all
problems of computer graphics have now been resolved. The
cost of many installations is not yet too encouraging
either. But in the course of further technological progress
computer graphics should gradually advance to a standard
tool in the ship design office.
- 14 -
DESIGN AND PRODUCTION
Design and production form a continuum in data pro-
cessing. The output from the design phase is input to
the production process, and it is expedient to simplify
the transition by identical data formats.
Consequently it has proven successful to integrate
design- and production-oriented computer software wherever
large program packages had to be developed for automation
purposes. In other words, design-oriented computer ap-plications have been thriving best in an environment of
computer systems thinking motivated by production tech-
nology.
The major efforts of the automobile and aircraft
industries spanning design and production have already
been mentioned. In our field the best known examples are:
AUTOKON, References (13) and (14):
This software system encompasses hull form definition
(fairing), part definition for a great number of struc-
tural members of the ship, and preparation of numerical
control tapes for flame-cutting equipment.
CASDOS, References (15)1 (16), and (17):
The objective of this system is to generate from contract
plans and specifications, with due regard to the peculiar-
ities of a given shipyard, all the necessary information
for building a ship: Working drawings, bills of materials,
and NC tapes. The system is based on man-computer dialog,
but the role of the man is largely creative and critically
selective, while a great amount of the routine work involved
in design detailing is automated. CASDOS is in trial oper-
ation at the Puget Sound Naval Shipyard at the present time.
- 15 -
PROBLEM-ORIENTED LANGUAGES
Widespread computer use in design offices is often
hampered by the fact the engineering staff is not suff-
iciently familiar with the whole potential of the computer
and the details of programming. Learning a programming
language like FORTRAN or ALGOL is perhaps not too difficult
but requires a conscious effort in intellectual skills not
too germane to design so that we must not expect every
member of a design team to become conversant with such
programming languages.
Whenever one is interested in making the computer a
useful, everyday tool for all engineers in a large organi-
zation, rather than relying on the services of a few
computer specialists, introduction of more problem-oriented
languages will be the expedient solution. They will enable
the user to talk to the computer in his own technical term-
inology. CASDOS and AUTOKON point in this direction. In
the AUTOKON code, for example, a single, compact program-
ming statement is sufficient to defihe, practically in
shipbuilding terminology, all the details necessary for the
automatic flamecutting of a floor plate.
The organizational effort required for such languages
is considerable. But they nonetheless represent a natural
solution whenever a major programming system is to be made
available to a large engineering organization. Problem-
oriented languages are conducive to efficient sharing of
the work between the systems programmer and the user.
- 16 -
OUTLOOK
In the preceding, several significant tendencies
have been pointed out with regard to the effects the rapid
advances of computer technology may have upon the scope
and style of ship design. It depends on the initiative and
foresight of our profession to what extent and how soon we
shall exploit the new potential. There is no doubt that
the successful innovations of other fields will gradually
be adopted in ours. But there are encouraging signs that
a little more will happen, namely a thorough reevaluation
of the methods of ship design, and perhaps the development
of a special style of computer use tailored to a modernized
interpretation of the old art of ship design.
- 17 -
ACKNOWLEDGMENTS
The thoughts presented in this paper summarize in a
very condensed way the material I collected last winter
for a new graduate course in "Computer-Aided Ship Design"
at The University of Michigan, Reference (18). A grant
the Department of Naval Architecture and Marine Engineering
had received from the Bethlehem Steel Corporation helped
me during the preparation of my notes. I further want to
acknowledge gratefully the inspiration and advice from my
colleagues, especially Harry Benford whose influence this
paper cannot deny, and the constant encouragement I derived
from the interest of my students.
REFERENCES
Benford, H., "Principles of Engineering Economy inShip Design," Transactions SNAME, 1963
Benford, H., "Fundamentals of Ship Design Economics,"Lecture Notes, The University of Michigan, Departmentof Naval Architecture and Marine Engineering, AnnArbor, Michigan, 1968
Cheng, H. M., "Performance Comparisons for MarineVehicles, SNAME New York Metropolitan Section,September 1968
Murphy, R. D., Sabat, D. J., Taylor, R. J., "LeastCost Ship Characteristics by Computer Techniques,"Marine Technology, April 1968
Mandel, P., Leopold, R., "Optimization Methods Appliedto Ship Design," Transactions SNAME, 1966
Moe, J., Lund, S., "Cost and Weight Minimization ofStructures with Special Emphasis on LongitudinalStrength Members of Tankers," De Ingenieur, nos. 47and 49, 1967, The Hague, Holland
Hooke, R., Jeeves, T. A., "Direct Search Solution ofNumerical and Statistical Problems," Journal of theAssociation for Computing Machines, Vol. 8, April1962
Kowalik, J., "Nonlinear Programming Procedures andDesign Optimization," Acta Polytechnica Scandinavica,no. Ma 13, Troudheim, 1966
Kavlie, D., Kowalik, J., Lund, S., Moe, J., "DesignOptimization Using a General Nonlinear ProgrammingMethod," European Shipbuilding, no. 4, 1966
Meyer-Detring, D., "Tanker Preliminary Design Economics,"SNAME Southeast Section, Miami, September 1969
Chasen, S. H., "Experience in the Application ofInteractive Computer Graphics," Section 11, LectureNotes, Computer-Aided Ship Design, Intensive ShortCourse, The University of Michigan, Department ofNaval Architecture and Marine Engineering, Ann Arbor,Michigan, May 1968
Herzog, B., "Computer Graphics: An Introduction,"Section 7, Lecture Notes, Computer-Aided Ship Design,Intensive Short Course, The University of Michigan,Department of Naval Architecture and Marine Engineering,Ann Arbor, Michigan, May 1968
- 19 -
Hysing, T., "From Basic Design to Flamecutting,"note issued by Central Institute for IndustrialResearch, Oslo, January 1968
Sorensen, P., "Autokon II, A Preliminary OutlineDescription," Shipping Research Services, Oslo,July 1969
Nachtsheim, J. J., Romberg, B. W., O'Brien, J. B.,"Computer Aided Structural Detailing of Ships,"Transactions SNAME, 1967
Cohen, J. B., Gardner, G. 0., Romberg, B. W., "DesignAutomation in Ship Detailing," Proceedings A.C.M.,National Meeting, 1967
Romberg, B. W., "A Computer System for StructuralDetailing of Naval Ships," SNAME Chesapeake andHampton Roads Sections, September 1968
Nowacki H., "Computer-Aided Ship Design," LectureNotes, The University of Michigan, Department ofNaval Architecture and Marine Engineering, Ann Arbor,April 1969
- 20 -