Journal of Ship Production and Design, Vol. 31, No. 2, May 2015, pp.

7987
http://dx.doi.org/10.5957/JSPD.31.2.130044

New Estimation Methods for the Steel Weight of European Inland

Dry Bulk Ships
Robert G. Hekkenberg and Hans Hopman
Delft University of TechnologyMaritime & Transport Technology, Delft, The Netherlands

There are hardly any methods available for the estimation of the steel weight of
inland ships in the earliest design stages. The few weight estimation methods that
are available are out of date, limited to a narrow band of main dimensions and not
sensitive enough to be used for innovative designs. In this article, which summarizes key results from the PhD thesis of Hekkenberg (2013), new methods are
derived for the estimation of the steel weight of inland dry bulk ships. The estimation
methods that are presented are derived from large systematic series of computergenerated ship designs that comply with Lloyds Registers rules regarding their structure and with the European rules for freeboard. The structure and weight of these
designs are validated by a comparison of modeled scantlings with the scantlings of
midship sections of existing ships. Further validation is done by a comparison of the
modeled overall steel weight with the steel weight of actual ships and the weight
estimates of existing methods. The designs are used to derive two types of estimation
methods: the first is a set of simple formulas that are valid for inland ships with
common sizes and L/B ratios and the second is a more complex set of formulas that
allows estimation of the steel weight of inland ships with draughts ranging from 1.5 to
4.5 m, lengths of 40185 m, beams of 525 m, and L/B ratios of 420.
Keywords: barges; design (vessels); computers in design

1. Introduction
THE EUROPEAN inland shipping sector is going through significant changes. There is a rapid increase in the average and
maximum size of inland ships and at the same time there are
worries about low water levels that lead to the assumption that
current deep-draught inland ships may eventually be replaced by
ships that are better adapted to shallow water. This expectation
also exists in the ship design community as is underlined by the
presentation by Thill (2009) of DST, the German Development
Center for Ship Technology and Transport Systems. Thill predicts
the need for lighter and wider shallow draught vessels to cope
with low water levels in the Rhine (Thill 2009). As a result, it is
likely that an increasing number of ships with dimensions, L/B
ratios, and B/T ratios that deviate significantly from common
values will be developed.
Manuscript received by JSPD Committee May 15, 2014; accepted
June 4, 2014.
MAY 2015

In inland ship design, early-stage weight estimates often rely

heavily on the experience of the designer or require a detailed
structural analysis. There are no adequate estimation methods for
the estimation of steel weight in contrast to seagoing ships for
which several such methods exist. The advent of inland ships
with deviating main dimensions makes it hard or impossible to
rely on experience any more, so an alternative is looked for in the
form of new estimation methods.
These estimation methods should enable designers to quickly
make an acceptable estimate of the steel weight of European
inland ships with a large range of lengths, beams, and draughts.
To be future-proof, it is also considered worthwhile to be able
to predict the steel weight of vessels that exceed the current
length limit of 135 m that is imposed by the Central Commission for Navigation on the Rhine (Central Commission
for Navigation on the Rhine 2010). In the remainder of this
article, which summarizes key results from the PhD thesis of
Hekkenberg (2013), these estimation methods and their development are discussed.

2158-2874/15/3102-0079$00.00/0 JOURNAL OF SHIP PRODUCTION AND DESIGN

79

2. Literature review
Steel weight of inland ships constitutes one of the main gaps in
academic knowledge on the technical characteristics on inland
ships. Hengst (1995, p. 119), who provides a broad overview of
the inland shipping sector, directly refers to the detailed calculations of classification societies for the determination of a vessels
structure and provides no estimation methods for the weight of
inland ships. The various classification societies provide detailed
calculations for all main structural elements of inland ships, but
for use as quick reference formulas, Germanischer Lloyd and
Bureau Veritas (Germanischer Lloyd 2006, Part B, Chapter 4,
Section 2, paragraph 5.2) only state a steel weight of 0.15

length
beam
depth (LBD) for vessels with a depth of 3.7 m
or less and 0.1
LBD for vessels with a larger depth. As a result
of the obvious weight discontinuity resulting from this approach
and the fact that no differentiation by vessel type (i.e., tank ship,
container ship, or dry bulk ship) is provided, this method is considered to be too inaccurate for practical use. Rules by Lloyds
Register (2008) do not provide quick weight estimates like those
by Germanischer Lloyd (2006) but only provide detailed requirements for steel structures of inland ships. As a result, these rules
require a detailed and time-consuming analysis of the steel structure before a weight estimate can be made.
This lack of available quick weight estimation methods is
further underlined by Hofman (2006), who describes a method
to determine the optimal characteristics for an inland container
vessel for a specified waterway. Hofman concludes that there is
no proper way to determine the light weight for those ships because
direct calculations of scantlings, based on rules alone, lead to an
underestimation of light weight. Hofman also states that previous
methods by Heuser (1986) and Michalski (2005) are inadequate
for his purposes because the method by Heuser . . .is a rough and
relatively old approximation, not strictly related to container
vessels, whereas the work by Michalski only covers a small range
of vessel sizes.
Heuser (1986) does, however, provide one of the few weight
estimations that are valid for inland ships. It only uses LBD as
a variable and does not subdivide light weight into steel weight
and other weight elements.
The only known paper in which the effects of length, beam,
and draught on the weight of inland ships are systematically
explored stems from 1978 is by Schellenberger (1978). The
paper provides data on the steel weight and cost of hull construction of dry cargo vessels and dry cargo push barges with
lengths between 75 and 125 m, beams between 8 and 15 m, and
depths of 34.5 m. The calculations are made according to an
obsolete set of classification rules and are only based on the
calculated scantlings of the midship rather than on the weight
of the entire hull.
This lack of up-to-date inland navigation-specific calculation
methods would not be a particular problem if there were a large
degree of similarity between seagoing and inland ships; one could
use methods for seagoing ships as a basis and derive values for
inland ships from them. Unfortunately, this large degree of similarity between the steel structures of seagoing and inland ships
does not exist. As a result of the vastly different environmental
conditions (i.e., mainly no significant waves), near absence of
freeboard, and different L/B, B/T, and D/T ratios, steel weight
estimates for seagoing ships cannot be used for inland ships.
80

MAY 2015

Table 1 Comparison of inland ship steel weight with estimation

method of Watson and Gilfillan (W&G)

Ship
A
B
C

W&G
Coaster

W&G
Bulk

W&G
Container

463 T
1091 T

478 T
1128 T

556 T
1309 T

W&G
Tanker

Actual

953 T

375 T
845 T
700 T

In Table 1, it is demonstrated how the differences between

seagoing ships and inland ships affect the validity of weight estimation methods that are intended for seagoing ships when they
are applied to inland ships. In the table, the results of the steel
weight estimation method by Watson and Gilfillan for coasters,
bulk carriers, container ships, and tankers (Watson 1998, p. 82)
are compared with known steel weights of the bare hulls of three
existing inland ships:
Ship A: a 86
9.6
3.5-m (L
B
D) dry cargo vessel;
Ship B: a 135
11.45
4.25-m (L
B
D) dry cargo
vessel; and
Ship C: a 110
11.4
5-m (L
B
D) tank vessel.
From Table 1, it is clear that for each of the ships, the structural
weight is always overestimated and as a result, the method is not
applicable for inland ships. It is of course possible to tune the
method by Watson and Gilfillan to inland ships on the basis of
weight data of existing ships, but the amount of publicly available
data is limited and by definition excludes innovative designs.
From this review of weight estimation methods, it is concluded
that the estimation of the steel weight of inland ships represents
an important gap in knowledge that needs to be addressed, especially because steel weight has a direct influence on both the
cargo-carrying capacity of the ship and on the building cost of
the hull. These two aspects in turn have a strong impact on the
cost of transport by inland ship.

3. Approach
The range of main dimensions of existing European inland
ships is limited, whereas the estimation methods to be developed
are not only intended for common inland ships, but also for inland
ships with nonstandard main dimensions. Therefore, an analysis
of the steel weight of existing ships is an insufficient basis for the
development of the new estimation methods. This in turn leads to
the need for an alternative approach. In this approach, a design
model is developed with which conceptual designs can be generated automatically. This model is used to generate a large series
of designs with systematically varied main dimensions. In this
series, draught is varied from 1.5 to 4.5 m, length is varied from
40 to 185 m, beam is varied from 5 to 25 m, and L/B ratio is
bounded between 4 and 20, as is shown in Fig. 1. It should be
noted, however, that this range of dimensions goes well beyond
the range of dimensions of existing inland ships as a result of
which model outcomes can only be validated partly and the results
at the edges of the explored design space should be used with
significant caution.
The structures and steel weights of the modeled ships are validated by means of a comparison of modeled scantlings and weight
JOURNAL OF SHIP PRODUCTION AND DESIGN

Fig. 1 Overview of design points in the systematic series

with actual scantlings and/or weight for a number of existing ships

with main dimensions that are common in Europe. As a final step,
the steel weights from all ships are analyzed and converted into
formulas by means of an ordinary least squares regression analysis.
To facilitate the generation of the ship designs, a model is
created through a combination of computer-aided design program
Rhinoceros 3D and Excel. Rhinoceros is used for all geometryrelated operations, whereas Excel is used to calculate scantlings
of all main structural members according to the structural rules
of Lloyds Register (2008). This leads to a three-dimensional (3D)
representation of the design, which includes all main structural
members, as is shown in Fig. 2.

4. Design model
The core of the model with which the ship designs are generated is formed by the equations of the structural rules for inland
ships by Lloyds Register (2008), which will not be iterated here
as a result of their sheer number. These equations are applied to
ships with the following structural arrangement. The sections
forward and aft of the hold are transversely framed with a frame
spacing of 500 mm and a web frame spacing of 3000 mm. The
midship, if transversely framed, has a frame spacing of 600 mm
and a web frame spacing of 1800 mm. In case the midship is
longitudinally framed, the number of girders in the double bottom
is selected such that the spacing between them is as close as
possible to 3 m under the boundary condition that there is a

Fig. 2 Modeled ship (cut in half for illustration purposes)

MAY 2015

center girder and there are girders below the longitudinal bulkheads of the double hull. Longitudinals are fitted between these
girders at a maximum spacing of 600 mm. This maximum spacing
is also maintained in the sides, whereas the actual spacing is
determined on the basis of the height of the sides above the tank
top. With this structural arrangement and the classification rules,
the scantlings of the main structural members are determined. The
length and/or area of each structural element is obtained from
the 3D representation of the ship, thereby allowing the calculation
of its weight and center of gravity.
Still, even with an accurate calculation of the scantlings of
the main structure, the weight of the vessel will be underestimated
as a result of the absence of structural details like brackets and
additional weight resulting from, e.g., welds and paint. Furthermore, uncertainty in the actual weight of structural members
occurs as a result of the fact that lightening holes in web frames,
floors, plate stringers, and nonwatertight girders are not modeled
and that local reinforcements to support machinery other than
the main engine and generator sets are also not modeled. Finally,
simplification of the hull form, especially the absence of tunnels
in the aftship, may also lead to errors in structural weight.
This has been compensated for by applying the following
weight additions and subtractions to the calculated weight of all
structural items:
All structural elements in the midship and foreship have
been allocated a 10% weight addition.
All structural elements in the aftship have been allocated
a 15% weight addition.
A 25% weight reduction is applied to the weight of web
frames, floors, plate stringers, and nonwatertight girders to
account for lightening holes.
A 30% weight reduction is applied to plate frames in tanks.

Such values are common values within naval architecture and it will
be shown later that these assumptions result in realistic steel weights.
Length, beam, and draught are user inputs, whereas the
required depth is determined by the requirements of Appendix II,
Chapter 4, Article 4 of the European guideline for inland navigation (European Parliament and Council of the European Union
2006). For dry bulk ships, this results in a safety distance of no
less than 30 cm and a freeboard of no less than 15 cm, assuming
that the accommodation does not contribute to the vessels stability
in a positive way.
JOURNAL OF SHIP PRODUCTION AND DESIGN

81

Because the details of the shape and general arrangement of

the ship are of secondary importance for the estimation of the
overall steel weight, the algorithm according to which they are
generated is not elaborated here in detail but may be found in
Hekkenberg (2013). The most important parameters are, however, briefly discussed. In all cases, the stern form is a pram
shape with a length of 1.5 B with an upper limit of 0.4 L and a
local block coefficient of 0.66. The bow is ship-shaped with a
length that is the smallest of 1.25 B or 4.2 T. These values
correspond well with suggestions by Heuser (1987). The distance
between the stern of the ship and the engine room bulkhead is
determined by the largest of either the required length to fit the
main engine and generator set in the engine room with sufficient
free space around them or the required length for the accommodation. The single-tier accommodation starts 2.5 m forward of
the stern and its length is determined by a required floor space
of 100 m2 and a width limitation caused by 1-m wide walkways
that run alongside the accommodation. The length of the accommodation is, however, also limited to a maximum of 20% of the
vessels length. The length of the section forward of the hold is
determined by the prescribed length of the forepeak and the
required space to fit the bow thruster and bow thruster engine in
the bow thruster room with sufficient free space around them.
The length of the hold is determined by subtracting the length of
fore and aft sections from the total length of the ship.

5. Model validation
Validation of the correctness of the outcomes of the developed model is done in two ways. It is checked if the structures
of ships that are created in the model match those of actual
ships, both in terms of scantlings of the midship section and
in terms of overall steel weight. It should be noted, however,
that only limited validation of the results of the model is possible because no inland dry bulk ships longer than 135 m exist
in Europe, whereas there is hardly any modern vessels below
70 m and the L/B and L/T ratios of existing vessels are in a
significantly narrower band than is explored in this article.
5.1. Validation of midship section scantlings
Validation of calculated scantlings is done through a comparison with the midship sections of two actual dry bulk ships:
1. A 86
9.6-m ship with a transversely framed double bottom
and sides; and
2. A 104.5
11.44-m ship with a transversely framed double
bottom and sides.
5.1.1. Ship 1: length 86 m, beam 9.6 m, T 2.85 m,
depth 3.0 m. The vessel is a ship with a frame spacing of
500 mm and a web frame placed every six frames. It has a
double bottom height of 485 mm and sides that are 635 mm wide.
Comparison of the actual scantlings of the ship with modeled
values yields the results as shown in Table 2.
When reviewing the comparison between actual and modeled
scantlings, several things are important to note:
1. Deck plating, coaming, and sheer strake have similar thicknesses in model and reality. In both cases these are the
elements of which thickness is increased to withstand the
82

MAY 2015

Table 2 Comparison of modeled and real scantlings of a 86
9.6-m

dry bulk (DB) vessel
Real
Bottom plating
Inner bottom plating
Bilge
DB floors
DB girders
Inner side plating
Side plating
Sheer strake
Side frames
Inner side frames
Deck plating
Coaming
Web frames

10
12
12
8
8
9
9
25
HP140
8
HP140
8
18
20
8

Model
mm
mm
mm
mm
mm
mm
mm
mm
Profile
Profile
mm
mm
mm

7.5
12
9.5
8
8
7
7
19
HP120
8
HP120
8
19
21
8

mm
mm
mm
mm
mm
mm
mm
mm
Profile
Profile
mm
mm
mm

longitudinal bending moment. That they have similar thickness implies that the model responds correctly to the need
to increase scantlings to withstand these moments.
2. Plating thickness in the side and bottom is thicker in reality
than in the model. This is however a designers choice: for
robustness, shell plating on real inland ships is never thinner
than 9 or 10 mm despite the fact that the rules allow use of
thinner material. The underestimation of calculated scantlings is in line with the earlier observations of Hofman
(2006). To eliminate unwanted differences between model
values and real values, the model has the option for use of a
user-defined minimum thickness of side and bottom plating.
In all results that are shown in this article, this minimum
value is set at 9 mm.
3. A similar idea is probably valid for the stiffeners in the side:
according to the rules, they can be made lighter than they
are in practice. This is not compensated for.
5.1.2. Ship 2: length 104.5 m, beam 11.44 m, T 3.1 m,
depth 3.5 m. The ship, for which the main frame is shown in
Figure 3, has a frame spacing of 620 mm and a web frame placed
every five frame spaces. The double bottom is 500 mm high and
the sides are 635 mm wide. Comparison of model values with
actual scantlings yields the results as shown in Table 3.
When reviewing the comparison between actual and modeled
scantlings, the following observations can be made:
1. Deck plating, coaming, and sheer strake have similar thicknesses in model and reality. In both cases these are the
elements of which thickness is increased to withstand longitudinal bending moment. That they have similar thickness
implies that the model responds correctly to the need to
increase scantlings to withstand these moments.
2. There is a slight discrepancy between modeled and real
bottom plating thickness as well as double bottom (DB)
girder thickness. The latter is the result of a design choice
rather than a strict requirement and the first is likely to be
caused by minor differences in calculated bending moments.
3. Inner bottom plating thickness differs. This thickness is
on occasion considered negotiable with class and subject to a number of designer choices. Note that in this
case the designer opted for a thinner inner bottom than
JOURNAL OF SHIP PRODUCTION AND DESIGN

Fig. 3 Midship section of Ship 2; drawing courtesy of Mercurius Shipping Group

a different designer did for the inner bottom of Ship 1,

which is smaller.
4. Frames in the side and inner side differ slightly. The difference in stiffener type in the side is caused by a calculated
required section modulus that is marginally higher than
that of the stiffener used in the real ship. The difference in
stiffener type in the inner side is caused by the designers
choice to use identical frames in the side and inner side,
whereas the inner side frames selected by the model are
lighter frames with a similar section modulus. The reason
Table 3 Comparison of modeled and real scantlings of a 104.5

11.44-m dry bulk (DB) ship
Element

Real Value

Bottom plating
Inner bottom plating
Bilge
DB floors
DB girders
Inner side plating
Side plating
Sheer strake
Side frames
Inner side frames
ADN stringers

11
11
13
8
10
10
10
20
100
75
9
100
75
9
100
8 with
100
15
face plate
22
25
8

Deck plating
Coaming
Web frames

MAY 2015

Model Value
mm
mm
mm
mm
mm
mm
mm
mm
profile
profile

mm
mm
mm

11.5
12
13.5
8
8
8.5
8.5
21.5
130
65
8
HP160
7
100
10 with
100
15
face plate
21.5
23.5
7.5

mm
mm
mm
mm
mm
mm
mm
mm
profile
profile

mm
mm
mm

this frame was not also used on the side shell is a requirement for stringers on the shell posed by ADN regulations.
5. Plating thickness in the side is slightly thicker in reality
than in the model. This is however again a designers
choice, as was also discussed for Ship 1.
Based on this, the results from the model for both ships are
deemed acceptable because general results match those of actual
ships and discrepancies can mainly be explained by deliberate
deviations from rule requirements. What is important, however,
is to realize that these design choices exist, making ships in
practice heavier than they need to be if built strictly according
to the rules. Ignoring such designer choices leads to a weight
underestimation of approximately 4% for the first ship in this
validation. This underestimation gets more significant for small
ships: for a ship with LBT 40
5
2 m, the underestimation
increases to approximately 9%. As is apparent from Ship 2, for
larger ships, required plate thickness quickly increases, thereby
resolving the discrepancy between required and actual scantlings.
5.2. Validation of overall steel weight
Validating the overall weight of the modeled steel structure of
a ship is difficult as a result of the impact of the many possible
design choices in terms of frame spacing, double hull width,
double bottom height, choice of framing system, vessel layout,
etc. However, based on a number of reference vessels and estimates from Germanischer Lloyd (2006), a reasonable validation
can be done.
Germanischer Lloyd states a steel weight of 0.15 * LBD for
vessels with a depth of 3.7 m or less and 0.1 * LBD for
vessels with a larger depth. (Germanischer Lloyd 2006, Part B,
JOURNAL OF SHIP PRODUCTION AND DESIGN

83

Table 4 Weight data of existing dry bulk vessels

LBD

135
11.45
4.25

86
9.6
3.5

86
9.6
3.75

Weight aftship
Weight midship
Weight foreship
Weight total
Weight/LBD

150
620
75
845
0.1286

57
270
48
375
0.1298

76
271
53
400
0.1292

verse framing system, a frame space in the midship of 0.62 m,

and a frame space at the fore and aft end of 0.5 m, results match
reality closely, although both fore and aft sections are relatively
light, as is shown in Table 5. In both cases, this may be because
no accommodation weight is included in the model. Weights of
front and aft sections are however very similar to that of another
real 86
9.6
3.5-m dry cargo vessel, which is the second
ship in Table 4.

LBD, length
beam
depth.

6. Trends in steel weight

Table 5 Comparison of real and modeled weight

LBD
Weight aftship
Weight midship
Weight foreship
Weight total
Weight/LBD

Dry Bulk (real)

Dry Bulk (model)

86
9.6
3.75
76
271
53
400
0.1292

86
9.6
3.75
58
271
41
370
0.120

LBD, length
beam
depth.

Chapter 4, Section 2, paragraph 5.2). Looking at a number of

actual vessels in Table 4, it is found that their weights are indeed
close to those values.
When modeling the 86
9.6
3.75-m ship with a double
bottom height of 0.6 m, a double hull width of 0.635 m, a trans-

In Fig. 4, the steel weights of all designs in two series of

longitudinally and transversely framed designs that are generated
with the design model are plotted against LBD. In this same
figure, the weight estimates as provided by Germanischer Lloyd
(2006), being 10% or 15% of LBD, are shown. What becomes
apparent is that a number of ship designs are substantially heavier
than these estimates. This underlines the limited validity of these
estimates for nonstandard ships.
When the steel weight of transversely framed ships is plotted
against LBT in a 3D graph in Fig. 5, it becomes apparent that
the lightest ships are relatively short and wide ships with a high
draught. For these short ships, longitudinal bending moments are
not yet high enough to increase plate thicknesses beyond the
minimum value of 9 mm. The reason for the use of LBT instead
of LBD in Fig. 5 and in the weight estimation methods is that
this allows for a more straightforward analysis of lightweight

Fig. 4 Steel weight versus length
beam
depth (LBD)

Fig. 5 Steel weight per m3 of LBT

84

MAY 2015

JOURNAL OF SHIP PRODUCTION AND DESIGN

versus displacement. The use of T rather than D as an estimation

value is justified by the fact that the two are closely linked for
the ship type under consideration. As was discussed before, the
required depth is determined by the requirements of Appendix II,
Chapter 4, Article 4 of the European guideline for inland navigation (European Parliament and Council of the European Union
2006), which leads to a minimum required freeboard of only
15 cm for all dry bulk ships under the assumption that the
accommodation does not contribute to the vessels stability. In
practice, this small freeboard is indeed commonly used and it
is also applied to all created designs.
The generated data can also be used to estimate what Watson
and Gilfillans well-known K-factor (Watson 1998, p. 82) would
be for inland dry bulk ships. For E-numbers between 260 and
5500, K-values ranging from just over 0.021 to just under
0.044 are obtained, which is a significantly larger spread than for
any other ship type in Watson and Gilfillans method. Limiting
the data set to ships that do not exceed the current maximum
allowed length of 135 m does not diminish this spread in K-values.
This again underlines the need for new, alternative weight estimation methods for inland ships.

Table 6 Simple estimationcoefficients

Transverse Framing
T (m)

c1

c2

R2

1.5
2
2.5
3
3.5
4
4.5

1.80E-05
8.33E-06
7.18E-06
3.00E-06
1.73E-06
1.78E-06
1.61E-06

2.37E-01
1.89E-01
1.51E-01
1.37E-01
1.25E-01
1.16E-01
1.09E-01

0.980
0.986
0.987
0.988
0.988
0.991
0.988

Longitudinal Framing
T (m)

c1

c2

R2

1.5
2
2.5
3
3.5
4
4.5

2.62E-05
1.49E-05
9.86E-06
6.04E-06
3.72E-06
2.58E-06
1.50E-06

2.11E-01
1.67E-01
1.37E-01
1.22E-01
1.14E-01
1.07E-01
1.03E-01

0.978
0.984
0.988
0.989
0.989
0.992
0.992

7. New estimation methods

Each dot in Fig. 5 represents the weight of a ship design. All of
these weights are analyzed by means of an ordinary least squares
regression analysis to derive accurate and easy-to-use estimation methods from them. The steel weight of ships with a length
below 135 m, a constant draught, and L/B values between 6 and
12 (i.e., more or less common inland ships) can be approximated well by means of a simple second-order polynomial, but
longer ships and ships with smaller and larger L/B ratios do
not fit these polynomials well. Therefore, two estimation methods
are developed. A simple method, with an applicability that is
limited to the mentioned more or less common vessels in the
form of second-order polynomials and a method featuring a more
complex formula covers the entire range of investigated designs.

Table 7 Complex estimationcoefficients for transversely

framed vessels
Unstandardized
Coefficients

c1
c2
c3
c4
c5
c6
c7

Standardized
Coefficients

Value

Standard
Error

Beta

Significance

2.597E01
2.320E-01
1.552E-03
4.444E-02
8.134E-07
1.024E00
7.691E02

11.305
9.047E-03
3.583E-04
2.659E-03
1.856E-08
0.132
399.793

0.339
0.079
0.226
0.533
0.072
0.009

2.297
25.600
4.332
16.715
43.823
7.784
1.924

0.022
0.000
0.000
0.000
0.000
0.000
0.055

7.1. Simple estimation method

The simple estimation method consists of a polynomial of the
form equation (1):
Wsteel c1  LBT 2 c2  LBT

The coefficients of this formula, however, differ per draught and

framing type. Table 6 displays coefficients c1 and c2 as well as
the R2 value for each draught and framing system.
7.2. Complex estimation method
The more complex, more generally applicable estimation method
uses a formula in the form of equation (2):
Wsteel c1 c2  LB c3  L2 T c4  LBT c5  L3:5 B
L1:3 T 0:7
1
c7  2 1:5
c6 
BT
B

Table 8 Complex estimationR 2 for transversely framed vessels

The coefficients to be used with each of the variables from equation (2) are presented in Table 7. When the significance of the
MAY 2015

variables is reviewed, it can be seen that all variables are significant with the possible exception of the last variable, which is a
correction for especially longitudinally framed vessels. From the b
value, it can be seen that its impact is very small. As a result, it
is not removed from the formula to keep the variables identical
for both framing systems.
When applying these parameters to the data set of transversely
framed dry bulk vessel designs, it is found that a very large part
of the variance of the data is explained by the formula, as is
shown by the R2 and adjusted R2 values in Table 8.
That the regression provides a good match with the original
data, i.e., the modeled ship designs, is apparent from Fig. 6, in
which the error distribution is shown: approximately 60% of all
original data points deviate less than 5% from the value predicted

R2

Adjusted R2

Standard Error of the Estimate

0.996

0.992

0.992

56.764

JOURNAL OF SHIP PRODUCTION AND DESIGN

85

Fig. 6 Complex estimationerror distribution for transversely framed vessels

Fig. 7 Complex estimationweight graphs for transversely framed vessels

by the estimation method, whereas only approximately 10% of

the weights of the designs deviate more than 10% from the value
resulting from the estimation method.
Finally, to demonstrate that the estimation method does not
lead to in unintended and undesirable instabilities in results as a
result of, e.g., overparameterization, in Fig. 7, weight surfaces,
based on calculated data points with a 5-m beam interval and a
10-m length interval, are shown for draughts of 2 and 4 m.
When equation (2) is used for longitudinally framed dry bulk
vessels, it can be seen from Table 9 that all variables with the
Table 9 Complex estimationcoefficients for longitudinally
framed vessels
Unstandardized
Coefficients

Standardized
Coefficients

Value

Standard
Error

Beta

Significance

c1
c2
c3
c4
c5
c6
c7

4.985E01
2.290E-01
1.234E-05
1.910E-02
9.584E-07
2.880E-01
1.066E03

7.976
6.383E-03
2.528E-04
1.876E-03
1.310E-08
0.093
282.060

0.324
0.001
0.094
0.608
0.020
0.013

6.250
35.899
0.049
10.181
73.185
3.099
3.781

0.000
0.000
0.961
0.000
0.000
0.002
0.000

86

MAY 2015

exception of the variable belonging to c3 (variable L2T) are

significant. However, it can also be seen from the b value that
its contribution to the final result is negligible. As a result, to still
be able to provide a single set of variables for the steel weight
of all European dry bulk vessels, the variable is not taken out of
the equation.
With these coefficients, good results are again achieved: R2
values are even higher than for transversely framed vessels and
the standard error of the estimate is even smaller, as is clear
from Table 10.
From the error distribution shown in Fig. 8, it is again clear
that the estimation method provides a good approximation for
the steel weight of longitudinally framed dry bulk ships: over
60% of the weights of the modeled ships deviate less than  5%
from the value that is predicted by the formula, whereas less
than 10% of the modeled ships deviate more than  10% from
the value resulting from the formula.
In Fig. 9, it is again demonstrated that the estimation method
does not lead to in unintended and undesirable instabilities in
Table 10 Complex estimationR 2 for longitudinally
framed vessels
R

R2

Adjusted R2

Standard Error of the Estimate

0.998

0.996

0.996

40.048

JOURNAL OF SHIP PRODUCTION AND DESIGN

Fig. 8 Complex estimationerror distribution for longitudinally framed vessels

Fig. 9 Complex estimationweight graphs for longitudinally framed vessels

results as a result of, e.g., overparameterization. Weight surfaces,

again based on calculated data points with a 5-m beam interval
and a 10-m length interval, are shown for draughts of 2 and 4 m.

8. Conclusions
In this article, the development of new, improved methods
for the estimation of the steel weight of inland dry bulk ships is
presented. It is shown that the design model that is used leads to
acceptable steel weight estimates. Furthermore, it is shown that the
estimation methods that are derived from the generated designs
lead to a good approximation of the weight of these generated
designs. As a result, it is concluded that the estimation methods
can be used for the estimation of the weight of European inland
dry bulk ships during the early stages of ship design, thereby
filling a major gap in the knowledge about inland ships.

References
C E N T R A L C O M M I S S I O N F O R T H E N A V I G A T I O N O F T H E R H I N E . 2010
Rijnvaartpolitiereglement (RPR) 1995, edition 2010, Central Commission
for the Navigation of the Rhine, http://www.ccr-zkr.org/13020500-nl.html.
EUROPEAN PARLIAMENT AND COUNCIL OF THE EUROPEAN UNION. 2006/87/EC
Directive of the European Parliament and of the Council of 12 December
2006 laying down technical requirements for inland waterway vessels and
MAY 2015

repealing Council Directive 82/714/EEC, European Parliament, Council

of the European Union, http://eur-lex.europa.eu/legal-content/EN/ALL/
?uri=CELEX:32006L0087.
GERMANISCHER LLOYD. 2006 Rules for Inland Navigation Vessels, Germanischer
Lloyd AG, Bureau Veritas, Inland Navigation Rules.
HEKKENBERG, R. G. 2013 Inland Ships for Efficient Transport Chains,
TU Delft Library, ISBN 978-94-6186-099-6, http://www.library.tudelft.nl/
dup/eng/.
HENGST, S. 1995 Binnenvaart in beeld, Delft University Press, Delft,
Netherlands, ISBN 90-407-1066-X, http://www.library.tudelft.nl/dup/eng/.
H EUSER , H. 1986 Anwendung Deim Entwurf von Binnenschiffen,
Schifftechnik, Band 33, Heft 1.
HEUSER, H. 1987 Neue Entwicklungen beim Entwurf von Flachwasserschiffe,
Teil 1, Colloquium Schiffstechnik, University of Duisburg-Essen,
Essen, Germany.
HOFMAN, M. 2006 Inland container vessel: Optimal characteristics for a
specified waterway, Proceedings, RINA (Royal Institution of Naval Architects) International Conference - Coastal Ships and Inland Waterways II,
March 1516, London, UK.
LLOYDS REGISTER. 2008 Rules and regulations for the classification of
inland waterways ships, http://www.lr.org/en/.
MICHALSKI, J. 2005 A Parametric Method for Preliminary Determining of
Mass Characteristics of Inland Navigation Ships, Polish Maritime Research,
3, 2005.
SCHELLENBERGER. 1978 Optimal Structure of self-propelled inland cargo ships
and push barges, Hansa, Schifffahrt, Schiffbau, Hafen, Volume 115, No. 22.
THILL, C. 2009 Effects of climate change on the operation of inland waterway vessels, Proceedings, Smart Rivers 2009 Conference, October 28,
Vienna, Austria.
WATSON, D. G. M. 1998 Practical Ship Design, Elsevier Science Ltd.,
Kidlington, Oxford, UK, ISBN 0080440541.
JOURNAL OF SHIP PRODUCTION AND DESIGN

87