Академический Документы
Профессиональный Документы
Культура Документы
Naval Architect, Naval Surface Warfare Center, Carderock Division, Seakeeping Department, West Bethesda, Maryland
Formerly Group Director (now deceased), TRANSAS, Hydrodynamics and Research Department, Washington, DC
Engineer, Naval Surface Warfare Center, Carderock Division, Seakeeping Department, West Bethesda, Maryland
The Beck, Newman and Tuck (BNT) numerical predictions are used in the Coastal
and Hydraulics Laboratory (CHL) Channel Analysis and Design Evaluation Tool
(CADET) model for predicting underkeel clearance (UKC) resulting from ship motions
and squat. The Ankudinov empirical squat prediction formula has been used in the
CHL ship simulator and was recently updated. The World Association for Waterborne
Transport Infrastructure (formerly The Permanent International Association of Navi-
gation Congresses, PIANC) has recommended several empirical and physics-based
formulas for the prediction of ship squat. Some of the most widely used formulas
include those of Barrass, Eryuzlu, Huuska, ICORELS, Romisch, Tuck, and
Yoshimura. The purpose of this article is to compare BNT, Ankudinov, and PIANC
predictions with measured DGPS squat data from the Panama Canal for four ships.
These comparisons demonstrate that the BNT, Ankudinov, and PIANC predictions fall
within the range of squat measurements and can be used with confidence in deep
draft channel design.
477
(2006) has recommended examining squat predictions with more underkeel clearance (UKC). Compared with other empirical for-
than one formula and comparing the results based on the type of mulas, it is somewhat more complicated because it includes addi-
ship, channel, and formula constraints. Because there is no one tional input parameters to account for the effects of the ships
formula that is universally accepted, an average, range, or maxi- propeller, bulbous bow, stern transom, initial trim, and the chan-
mum value might be considered in channel design. nels depth, blockage, and cross-section. Modifications have
The PIANC MarCom Working Group 49 (WG49) is updating recently been incorporated to improve its accuracy.
the 1997 PIANC guidance (1997) and expects to publish their Most mariners and pilots are more concerned with the static
report in 2013. The WG49 has reduced the number of squat pre- UKC and not just ship squat. Static UKC is the safety margin or
diction formulas to seven of the most popular formulas. These what is left after subtracting static draft and trim from the
include the updated versions by Barrass, Eryuzlu, Huuska, channel depth. Ship squat and an allowance for vertical ship
ICORELS, Romisch, Tuck, and Yoshimura. motions (for ships exposed to waves) must be subtracted to get
The Channel Analysis and Design Evaluation Tool (CADET) the net UKC. A brief mention of the static UKC is provided for
is a computer program originally developed by the US Navy to each ship, but the main focus of the article is comparisons of
determine the optimum dredge depth for entrance channels (2005) ship squat.
used by military vessels. It uses an external program, BNT (based The purpose of this article is to compare the BNT, Ankudinov,
on the work of Beck, Newman & Tuck 1975), to predict ship and PIANC predictions for four ships with measured ship squat
squat. BNT is a potential flow program that predicts sinkage and data from the Panama Canal. This article is an update of prelim-
trim from vertical force and pitching moment resulting from the inary comparisons of Ankudinov and PIANC predictions (Briggs
dynamic pressure on the hull. Briggs et al. (2010a) found reason- & Daggett 2009). These comparisons of almost 3000 squat mea-
able agreement between BNT and PIANC predictions for several surements demonstrate that the BNT, Ankudinov, and PIANC
ships including an aircraft carrier, two containerships, a tanker, predictions fall within the range of the squat measurements and
and a bulk carrier for two different channels. can be used with confidence in deep draft channel design. The
The Ankudinov ship squat formula is an empirical formula that second section in this article describes the ship and channel
predicts maximum squat resulting from midpoint sinkage and parameters in the Panama Canal study. Section 3 describes the
vessel trim (Ankudinov et al. 2000). It has been used in the CADET/BNT ship squat program. The Ankudinov ship squat
Coastal and Hydraulics Laboratory (CHL) Ship/Tow Simulator to formulas are presented in the next section. Section 5 describes
account for ship squat in the determination of instantaneous the five PIANC empirical squat formulas used in this study.
Nomenclature
Ac wetted cross-section area of canal, m2 KS channel width correction factor, Huuska SMax maximum ship squat at bow or stern,
As ships underwater amidships KTr trim coefficient, Ankudinov BNT, ft
cross-section, m2 KTT r stern transom factor, Ankudinov Ss stern squat, m
B ships beam, m KTT 1 initial trim effect factor, Ankudinov SM,j measured ship squat at location j, m
BS bias or difference between predicted KU unrestricted channel correction factor, SP,j predicted ship squat at location j for
and measured squat, m Romisch BNT, Ankudinov, or PIANC, m
BTr stern transom width, Ankudinov KDT squat at critical speed, Romisch S M average ship squat for measured
CB ships block coefficient Lpp ship length between perpendiculars, m DGPS, m
CF correction factor for ship MAE mean absolute error between predicted S P average ship squat for predicted BNT,
shape, Romisch and measured squat, m Ankudinov, or PIANC, m
CV correction factor for ship n inverse bank slope s1 corrected blockage factor, Huuska
speed, Romisch nTr trim exponent, Ankudinov T ship draft, m
Fnh Froude depth number PCh1 channel effect parameter, Ankudinov Tap draft at aft perpendicular
g gravitational acceleration (9.81 m/s2) PCh2 channel effect, trim correction Tfp draft at forward perpendicular
h water depth, m parameter, Ankudinov Tr ship trim, m/m, Ankudinov and BNT
hm mean water depth, Romisch, m PFnh ship forward speed parameter, Vcr critical ship speed, m/s
hmT restricted channel water depth, Ankudinov Ve enhanced ship speed, Yoshimura, m/s
Romisch, m PhT propeller effect in shallow water on trim Vk ship speed, knots
hOut height outside underwater trench, parameter, Ankudinov VS ship speed, m/s
similar to hT, ft Ph/T water depth parameter, Ankudinov W channel width, measured at bottom, m
hT height of dredged underwater trench, m PHu ship hull parameter for shallow water, WEff effective width of waterway, m
j location along Panama Canal Ankudinov WTop channel width, measured at top, m
K channel coefficient, Barrass R Pearson correlation coefficient sM standard deviation of ship squat for
Kb correction factor for channel RMSE root mean square error, m measured DGPS, m
width, Eryuzlu RS ratio between predicted and sP standard deviation of ship squat
KbT bulbous bow factor, Ankudinov measured squat for predicted BNT, Ankudinov,
KC canal channel correction factor, S blockage factor As/Ac or PIANC, m
Romisch Sb bow squat, m r ship volume of displacement, m3
KPS propeller sinkage factor, Ankudinov Sh channel depth factor for R and
KPT propeller trim factor, Ankudinov C channels, Ankudinov
KR restricted channel correction factor, Sm midpoint ship sinkage, m/m,
Romisch Ankudinov and BNT
478 Comparison of Measured Ship Squat with Numerical and Empirical Methods
Results are presented and discussed in Section 6 for the four
ships. Finally, the last section provides some conclusions.
Comparison of Measured Ship Squat with Numerical and Empirical Methods 479
Table 1 Ship parameters for Panama Canal study
Ship ID: E Elbe, G Global Challenger, M Majestic Maersk, O OOCL Fair; Type: T tanker, B bulk carrier, C containership; Location: B
bow, S stern; Date: 1997 December, 1998 April.
In his early work, Tuck (1966) calculated the dynamic pressure between the forward and aft perpendiculars (Fig. 5). Typically,
of slender ships in finite-water depth and infinite and finite-water generic ship lines from a ship database are used and adapted for
width by modeling the underwater area of the hull. This under- a particular ship because ship lines are proprietary and not readily
water area was defined by the 21 equally spaced stations along the available for newer ships.
ships length. Therefore, the ships geometry file, draft, speeds, The dynamic pressure is obtained for each depth Froude num-
and water depths are used in the BNT squat calculations. Within ber, Fnh, by differentiating the velocity potential along the length
this analysis, the fluid is assumed to be inviscid and irrotational of the hull. The Fnh is defined as
and the hull long and slender. Input hull definition is provided in
Vs
terms of the waterline beam and sectional area at these 21 stations Fnh p 1
gh
where g is gravitational acceleration and VS is ship speed in m/s.
The sinkage and trim predictions are obtained from the dynamic
pressure by calculating the vertical force and pitching moment,
which are translated to vertical sinkage and trim angle. Channel
depths should be the same order as the draft of the ship to satisfy
the shallow-water approximations assumed in Tuck (1966).
The BNT program numerically calculates midship sinkage, Sm,
and trim, Tr, as a function of Fnh. Because English units are used
in CADET, sinkage is measured in feet positive for downward
movement. Trim in feet is the difference between sinkage at the
bow and stern positive for bow down. The equivalent bow Sb and
stern Ss squat are given by
Sb Sm 0:5Tr
2
Ss Sm 0:5Tr
480 Comparison of Measured Ship Squat with Numerical and Empirical Methods
The ship hull parameter for shallow water, PHu, was recently
modified by Ankudinov and Briggs (2009) as
!
BT
PHu 1:7CB 2 0:004C2B 6
Lpp
Ph=T 1 e 13
Fnh
4.1. Midpoint sinkage Sm
The trim coefficient, KTr, is a function of many factors and is
The Sm is defined as
given by
Sm 1 KPS PHu PFnh Ph=T PCh1 4 T
KTr CnTr
B 0:15KP KP KB KTr KT1
S T T T
14
The ship, water depth, and channel parameters in this midpoint The first factor in this equation CnTr
B is the block coefficient CB
sinkage equation are described subsequently. The propeller param- raised to the nTr power. The exponent is defined as
eter KPS is defined as
PCh1
0:15 single propeller nTr 2:0 0:8 15
KPS 5 CB
0:13 twin propellers where Pch1 was previously defined in equation 9.
Comparison of Measured Ship Squat with Numerical and Empirical Methods 481
The next two factors define the propeller effect on the vessel form ships with CB > 0.7 tend to squat by the bow and fine-form
trim. The first factor KPS is the same as the propeller parameter for ships with CB < 0.7 tend to squat by the stern. The CB 0.7 is an
the midpoint sinkage and the second factor is the propeller trim even keel situation with maximum squat the same at both bow
parameter KPT and stern. Of course, for channel design, one is mainly interested in
the maximum squat and not necessarily whether it is at the bow or
0:15 single propeller
KP
T
16 stern. With these constraints in mind, the average PIANC values
0:20 twin propellers consisted of predictions for all five PIANC formulas for bow squat
but only Barrass and Romisch for stern squat.
The last group of three factors define the effects of the bulbous
bow KbT, stern transom KTr
T T
, and initial trim KT1 on the vessel trim.
T 5.1. Barrass
The Kb is given by
0:1 bulbous bow The Barrass ship squat formula has evolved and been revised at
Kb
T
17 least four times. The one in this article (Barrass 2002, 2009) is
0:0 no bulbous bow
considered the third version for both Sb and Ss. It is a function of
T
The KTr is defined by CB, ship speed, Vk, in knots, and channel blockage coefficient, K,
8 and is defined as
< BTr
0:1 0:04 stern transom
KTr
T
B 18 KCB Vk2 Sb CB > 0:7
: 21
0:0 no stern transom 100 Ss CB 0:7
where BTr is the stern transom width and is typically equal to 0.4B, If CB > 0.7, maximum squat occurs at the bow Sb. If CB 0.7,
although values as high as 0.7B have sometimes been used. it occurs at the stern and is equal to the stern squat, Ss. Barrass
T
The KT1 is given by channel coefficient, K, is based on analysis of over 600 laboratory
and prototype measurements for all three channel types. It is
Tap Tfp
T
KT1 19 defined as
Tap Tfp
K 5:74S0:76 1K2 22
where Tap is the static draft at the stern or aft perpendicular and
Tfp is the static draft at the bow or forward perpendicular. The limits on K are designed so that K 2 for C (also for R)
Finally, the channel effect trim correction parameter PCh2 for channels. The blockage factor S was previously defined. If S >
both R and C channels is defined as 0.25 for C (also for R) channels, the value of K is set to 2 to ensure
PCh2 1:0 5Sh 20 the limits required for K.
482 Comparison of Measured Ship Squat with Numerical and Empirical Methods
The channel width correction factor Ks is defined as 6. Validation results
7:45s1 0:76 s1 > 0:03
Ks 26 As mentioned previously, only five of the PIANC formulas were
1:0 s1 0:03 included in this study. Although all five can predict bow squat, only
where the corrected blockage factor s1 S for C channels. Huuska the Barrass and Romisch formulas were appropriate for stern squat
defined other values for s1 for the other two channel types, but predictions for canal channels. Therefore, only these two predic-
they are not presented here because we are only concerned with tions were used to calculate the PIANC average for stern squat.
C channels in this article. Depending on the value of CB, only bow or stern squat predictions
were calculated as dictated by the constraints of each formulation.
Again, the PIANC values were used to calculate an average bow
5.4. Romisch (five averages) or stern (two averages) squat prediction at each
Romisch (1989) developed formulas for both Sb and Ss from location for each ship to compare with the measured DGPS values.
physical model experiments for a C channel. The Romisch squat is
defined as 6.1. Uncertainty analysis
Sb ; Ss CV CF KDT T 27 Each of the parameters in the prediction of ship squat has inher-
ent uncertainties. The channel depth, h, is assumed to have no bias
The factors in this equation are correction factors for ship or variability because it is a deterministic parameter. Uncertainty in
speed, CV, ship shape, CF, and squat at critical speed, KDT, defined as the static and dynamic drafts, T, comes from the estimation of the
2 " 4 # draft at the pier, from the draft marks, and the sinkage and trim
Vs Vs
CV 8 0:5 0:0625 28 estimate, S. According to Kopp and Silver (2004), the error band in
Vcr Vcr the static draft is assumed to be known within a range of 1%. The
critical points of the bow, stern, and bilge are assumed to have an
8 error band within 4.5% of the actual value at the bow and stern and
< 10BCB 2
CF Bow 29 1.5% of the actual value at the bilge. The sinkage estimate is based
: Lpp on an analytical or empirical method from experimental results.
1:0 Stern The uncertainty in the sinkage comes from the scatter of the data
p in model tests and how well the calculated results fit the model test
KDT 0:155 h=T 30 results. This gives a variability of the sinkage parameter of approx-
imately 1% with no bias. Of course, the ship squat empirical esti-
Critical ship speed, Vcr, for a canal is a function of wave celer- mates are based on many different parameters that are not included
ity, C, and a channel shape correction factor, KC, defined as in every formula. Even if included, they are weighted differently
in each formula.
p p Arc cos1 S 1:5
Vcr CKC ghm 2 cos 31
3 3
6.2. Error metrics
where the mean water depth, hm, is a function of the projected
width at the top of the channel, WTop, defined as Statistical and error metrics used to quantify uncertainty in the
predictions are described in this section. The first two metrics are
AC AC the means and the standard deviations for the measured and pre-
hm 32
WTop W 2nh dicted squat data.
Two simplistic goodness-of-fit measures to characterize the agree-
However, because the hm was provided for the Panama Canal ment between the model squat predictions and measured DGPS
data, that value was used in the Romisch formula. data are the ratio RS between predicted and measured data and the
bias or difference between the two data sets. The RS is defined as
5.5. Yoshimura SP; j
RS 35
SM; j
The last squat formula was developed by Yoshimura (1986) and
included by the Overseas Coastal Area Development Institute of The bias BS is given by
Japan (2009) as part of Japans Design Standard for Fairways in
Japan. It was enhanced by Ohtsu et al. (2006) to include predic- Bs SP; j SM; j 36
tions for C (also for R) channels. It is defined as where SP,j is the predicted squat at the bow or stern at location
" #
1:5T BCB 15T BCB 3 Ve 2 j and SM,j is the measured squat at location j.
Sb 0:7 33 Three better error metrics include the root mean square error
h Lpp h Lpp g (RMSE), Pearson correlation coefficient (R), and the mean abso-
lute error (MAE). They are defined as
where the enhanced ship speed term, Ve, is a function of ship
v
speed, Vs, in m/s given by uN
u
u ( SP; j SM; j 2
Vs tj1
Ve 34 RMSE 37
1 S N
Comparison of Measured Ship Squat with Numerical and Empirical Methods 483
differences between the two data sets. Although unbounded,
CovSP ; SM
R p smaller values indicate better agreement. The R error statistic
Var SP Var SM provides an indication of the correlation that may exist between
N
data sets. It is bounded between 1 (perfect negative correlation)
(
j1
SP; j SP SM; j SM and 1 (perfect positive correlation). It is an indication of linearity
ss 38 between the data sets because an uncorrelated value of R 0 can
N
N
( SP; j SP 2 ( SM; j SM 2
indicate a nonlinear or random scatter relationship between data
j1 j1 sets. Therefore, small values of R indicate that the data sets are
uncorrelated, but this does not mean they are unrelated, just not
linearly related.
N
( j SP; j SM; j j
j1 6.3. Squat comparisons
MAE 39
N Figures 6 to 9 show time series comparisons of the BNT,
where SP is the mean of the predicted squat, SM is the mean of the Ankudinov, and PIANC squat predictions to the measured DGPS
measured squat, and N is the number of measurement points. values for the four ships along the Gaillard Cut. The top plot
Values of RS > 1.0 indicate overprediction, RS < 1.0 under- shows the RS for these three predictors at each location. The
prediction, and RS 1.0 perfect match. Similarly, values of BS > bottom plot shows the measured and predicted bow or stern squat
0.0 indicate overprediction, BS < 0.0 underprediction, and BS for each ship. A spline smoothing was applied to each of the three
0.0 a perfect match. The RMSE and MAE measure the actual predictors to illustrate the general trend of the data. The degree of
smoothness was automatically selected using cross validation of
the data. The entire data set was used in all the statistical and
Table 2 Mean and standard deviations for measured and metric calculations, however.
predicted squat data In lieu of error bars on the plots, the statistical and error
metrics are listed in Tables 2 to 5 to improve readability of the
Ship ID No. Statistic DGPS BNT Ankudinov PIANC plots. Table 2 compares the squat means and standard deviations
for the measured DGPS and predicted BNT, Ankudinov, and
E 1015 S 0.42 0.27 0.54 0.35
s 0.083 0.052 0.073 0.055 PIANC data. Table 3 lists the minimum, average, and maximum
G 633 S 1.11 0.76 1.33 1.00 values of RS for BNT, Ankudinov, and PIANC squat predictions
s 0.132 0.045 0.057 0.048 for each ship. Table 4 lists minimum, average, and maximum BS
M 649 S 0.84 0.45 0.74 0.74 between measured and predicted bow and stern squat for each
s 0.307 0.172 0.201 0.318 ship. Finally, Table 5 lists the RMSE, R, and MAE statistics for
O 681 S 0.58 0.45 0.59 0.55 each of the four ships.
s 0.099 0.077 0.058 0.053
Total 2978
6.3.1. Elbe tanker. Figure 6 shows the bow squat for the Elbe
S Sample mean or average; s sample standard deviation. Panamax tanker. The average water depth ranged between 13.0
Ship No. of Squat Minimum Average Maximum Minimum Average Maximum Minimum Average Maximum
ID Measurements Ratio Ratio Ratio Ratio Ratio Ratio Ratio Ratio Ratio
E 1015 0.45 0.64 0.84 0.96 1.29 1.78 0.60 0.83 1.11
G 633 0.53 0.69 0.90 0.96 1.21 1.61 0.72 0.92 1.21
M 649 0.30 0.54 0.89 0.46 0.91 1.70 0.45 0.89 1.62
O 681 0.54 0.78 1.16 0.78 1.04 1.48 0.70 0.96 1.33
Ratio RS predicted squat divided by measured squat: 1.0 is perfect match; < 1.0 is underprediction; > 1.0 is overprediction.
E 1015 0.28 0.16 0.04 0.02 0.11 0.22 0.21 0.08 0.03
G 633 0.65 0.35 0.08 0.05 0.22 0.48 0.38 0.10 0.17
M 649 1.62 0.39 0.05 1.28 0.1 0.29 1.01 0.10 0.41
O 681 0.35 0.13 0.07 0.17 0.01 0.19 0.21 0.04 0.14
Bias BS predicted squatmeasured squat: 0.0 is perfect match; < 0.0 is underprediction; > 0.0 is overprediction.
484 Comparison of Measured Ship Squat with Numerical and Empirical Methods
Table 5 Error statistics between measured and predicted squat 0.6 to 1.1 with an average underprediction of 0.8. The BS ranged
from an underprediction of 21 cm to an overprediction of 3 cm
Ship ID No. Statistic Units BNT Ankudinov PIANC with an average underprediction of 8 cm. The RMSEPIANC
0.09 m and MAEPIANC 0.08 m were very small, again showing
E 1015 RMSE m 0.17 0.13 0.09
a strong correlation of RPIANC 0.76.
R 0.75 0.77 0.76
MAE m 0.16 0.12 0.08
In general, the SP and the BS values tended to confirm each
G 633 RMSE m 0.37 0.25 0.16 other. The BNT predictions were smaller than measured values
R 0.41 0.37 0.40 but followed the same trends and were closer to the PIANC
MAE m 0.35 0.22 0.13 predictions. The Ankudinov formula overpredicted by 30% and
M 649 RMSE m 0.44 0.21 0.21 12 cm, whereas the PIANC underpredicted by 20% and 9 cm. The
R 0.84 0.82 0.83 PIANC had the best error metrics followed by Ankudinov and
MAE m 0.39 0.15 0.16 BNT. The Ankudinov and BNT predictions were approximately
O 681 RMSE m 0.15 0.07 0.08 50% and 100% larger than the PIANC values, although relatively
R 0.70 0.75 0.74 small maximum differences of 8 cm.
MAE m 0.13 0.06 0.06
Total 2978
6.3.2. Global Challenger bulk carrier. Figure 7 shows the bow
RMSE root mean square error; R Pearson correlation coefficient; squat for the Global Challenger Panamax bulk carrier. This ship
MAE mean absolute error. was trimmed 12 cm by the stern (i.e., deeper draft at the stern).
The water depth ranged between 12.9 and 13.2 m with an average
and 13.2 m with an average static UKC of 1.8 m. Table 2 shows static UKC of 1.4 m. Table 2 shows the mean SM 1:11 m for
the mean SM 0:42 m for the DGPS data with a standard the DGPS data (largest of the ships) with a standard deviation
deviation sM 0.083 m. Comparing the BNT predictions, the sM 0.132 m. For the BNT predictions, the SP 0:76 m, an
SP 0:27 m, an average underprediction of 15 cm, with a average underprediction of 35 cm with a sP 0.045 m. The RS
sP 0.052 m. The RS ranged from 0.5 to 0.8 with an average ranged from 0.5 to 0.9 with an average underprediction of 0.7
underprediction of 0.6 times the measured bow squat. The BS times the measured bow squat. The BS ranged from an
ranged from underpredictions of 28 to 4 cm with an average underprediction of 65 cm to an overprediction of 8 cm with an
underprediction of 16 cm. The RMSEBNT 0.17 m and MAEBNT average underprediction of 35 cm. The RMSEBNT 0.37 m and
0.16 m were relatively small, showing a strong correlation of MAEBNT 0.35 m were twice as large as the Elbe values, show-
RBNT 0.75. For the Ankudinov predictions, the SP 0:54 m, ing a weak linear correlation of only RBNT 0.41. For the
an average overprediction of 12 cm, with a sP 0.073 m. The RS Ankudinov predictions, the SP 1:33 m, an average overpre-
ranged from 1.0 to overpredictions up to 1.8 with an average of diction of 22 cm with a sP 0.057 m. The RS varied from 1.0
1.3 times the measured squat. The BS ranged from underprediction to 1.6 times the measured squat with an average overprediction
of 2 cm to overpredictions of 22 cm with an average overpredic- of 1.2. The BS ranged from underpredictions of 5 cm to
tion of 11 cm. Again, the RMSEAnk 0.13 m and MAEAnk 0.12 m overpredictions of 48 cm with an average overprediction of
were relatively small, showing a strong correlation of RAnk 0.77. 22 cm. The RMSEAnk 0.25 m and MAEAnk 0.22 m were
Finally, for the PIANC predictions, the SP 0:35 m, an average twice as large as the Elbe values, showing the weakest correlation
underprediction of 7 cm with a sP 0.055 m. The RS ranged from of only RAnk 0.37 for all ships. Finally, for the PIANC
Fig. 6 Panamax Elbe tanker values for (a) ratio RS and (b) measured and predicted bow squat Sb; ship northbound, sailing from right to left;
DGPS black open circle; BNT solid line; Ankudinov dash line; PIANC dotdash; spline fit for predictors
Comparison of Measured Ship Squat with Numerical and Empirical Methods 485
Fig. 7 Panamax Global Challenger bulk carrier values for (a) ratio RS and (b) measured and predicted bow squat Sb; ship northbound, sailing from
right to left; DGPS black open circle; BNT solid line; Ankudinov dash line; PIANC dotdash; spline fit for predictors
predictions, the SP 1:00 m, an average underprediction of 10% and 9 cm underprediction. For design purposes, overpredic-
11 cm with a sP 0.048 m. The RS were closer to the measured tion is more conservative and potentially safer. The error metrics
bow squat, especially above the location at station 1850. They for the Global Challenge were twice as large as the Elbe with the
ranged from 0.7 to 1.2 times the measured squat with an average worst correlation of all the ships. However, the worst RMSEBNT
underprediction of 0.9. The BS ranged from an underprediction of 0.37 m for the BNT predictions was not a huge overprediction.
38 cm to an overprediction of 17 cm with an average under-
prediction of 10 cm. The RMSEPIANC 0.16 m and MAEPIANC 6.3.3. Majestic Maersk containership. As mentioned previously
0.13 m were relatively small, again showing a very weak correla- in Section 6, according to Barrass (2009), maximum squat will
tion of only RPIANC 0.40. occur at the bow for a ship with a CB > 0.7 and at the stern for one
Again, the S P and the BS values tended to confirm each other. with a CB < 0.7. Therefore, bow squat was reported for the first
The BNT underpredicted the measured squat but followed the two ships previously discussed. The next two ships will illustrate
same trends as the two other predictors. In general, the Ankudinov squat by the stern because they have a CB < 0.7. Thus, the average
formula overpredicted and BNT and PIANC underpredicted bow PIANC value used in the plots is based on the average of only the
squat. The PIANC formulas overpredicted the measured squat for Barrass and Romisch predictions.
a short section from location 1920 to 1970. Thus, the Ankudinov Figure 8 illustrates the stern squat for the Majestic Maersk
formula averaged 20% and 21 cm overprediction and the PIANC Panamax containership. The water depth ranged between 12.9
Fig. 8 Panamax Majestic Maersk containership values for (a) ratio RS and (b) measured and predicted stern squat Ss; ship northbound, sailing from
right to left; DGPS black open circle; BNT solid line; Ankudinov dash line; PIANC dotdash; spline fit for predictors
486 Comparison of Measured Ship Squat with Numerical and Empirical Methods
Fig. 9 OOCL Fair containership values for (a) ratio RS and (b) measured and predicted stern squat Ss; ship northbound, sailing from right to left;
DGPS black open circle; BNT solid line; Ankudinov dash line; PIANC dotdash; spline fit for predictors
and 13.2 m with an average static UKC of 1.3 m. Table 2 shows stern squat by 10% with minimum and maximum predictions
that the mean SM 0:84 m for the DGPS data (largest of the about the same.
ships) with a standard deviation sM 0.307 m (largest of the
four ships). For the BNT predictions, the SP 0:45 m, an 6.3.4. OOCL Fair containership. Figure 9 shows the stern squat
average underprediction of 39 cm, with a sP 0.172 m. The RS for the OOCL Fair containership because CB < 0.7. This ship had
ranged from 0.3 to 0.9 with an average underprediction slightly the most trim with a value of 0.8 m by the stern. The water depth
larger than half of the measured squat. The BS ranged from an ranged between 13.0 and 13.2 m with an average static UKC of
underprediction of 5 cm to 1.6 m (station 1940.73) with an 2.9 m. Table 2 shows that the mean SM 0:58 m for the DGPS
average underprediction of 39 cm. The extreme underprediction data (largest of the ships) with a standard deviation sM 0.099 m.
at station 1940.73 is probably an error in the measured data For the BNT predictions, the SP 0:45 m, an average
because they all appear to look unusually large in this section underprediction of 13 cm, with a sP 0.077 m. The RS ranged
of the canal from 1940 to 1950. The RMSEBNT 0.44 m and from 0.5 to 1.2 with an average underprediction of 0.8 times
MAEBNT 0.39 m were largest of the four ships, but had the measured squat. The BS ranged from underpredictions of
the strongest correlation of RBNT 0.84. For the Ankudinov 35 to 7 cm with an average underprediction of 13 cm. The
predictions, the SP 0:74 m, an average overprediction of RMSEBNT 0.15 m and MAEBNT 0.13 m were the smallest of
10 cm, with a sP 0.201 m. The RS ranged from 0.5 to 1.7 times the four ships with a strong linear correlation of RBNT 0.70. For
the measured stern squat with an average underprediction of the Ankudinov predictions, the SP 0:59 m, a nearly identical
0.9. The BS ranged from a worst underprediction of 1.3 m (sta- value that overpredicts by 1 cm with a sP 0.058 m. The RS
tion location 1940) to an overprediction of 29 cm with an aver- ranged from 0.8 to 1.5 times the measured stern squat with an
age underprediction of 10 cm. The RMSEAnk 0.21 m and average ratio of 1.0 (near exact match with measured data). The
MAEAnk 0.15 m were relatively small and second only to BS ranged from underprediction of 17 cm and overpredictions
the Elbe with the strongest correlation of RAnk 0.82. Finally, of 19 cm with an average of 1 cm (near exact match). The
for the PIANC predictions, the SP 0:74 m, an average RMSEAnk 0.07 m and MAEAnk 0.06 m were the smallest of
underprediction of 10 cm, with a sP 0.318 m. The RS ranged the four ships with a strong linear correlation of RAnk 0.75. The
from 0.4 to 1.6 with an average underprediction of 0.9 times Ankudinov predictions with the smallest RMSE and MAE errors
the measured squat. The BS ranged from an underprediction were the best fit of the three predictors for the OOCL Fair.
of 1.0 m to overprediction of 41 cm with an average under- Finally, for the PIANC predictions, the SP 0:55 m, an average
prediction of 10 cm. The RMSEPIANC 0.21 m and MAEPIANC underprediction of 3 cm, with a sP 0.053 m. The RS varied from
0.16 m were similar to the Ankudinov predictions with the stron- underpredictions of 0.7 to overpredictions of 1.3 with an average
gest correlation of RPIANC 0.83. of 1.0 (near exact match) times the measured stern squat. The BS
The SP and the BS values confirmed one another because ranged from an underprediction of 21 cm to an overprediction of
they were identical. Although the BNT model underpredicted 14 cm with an average underprediction of 4 cm. The RMSEPIANC
the measured values, it showed the same trends as the mea- 0.08 m and MAEPIANC 0.06 m were similar to the Ankudinov
sured data and the other predictors. From station 1670 to 1850, predictions with a strong linear correlation of RPIANC 0.74.
both Ankudinov and PIANC tended to underpredict stern squat Again, the SP and the BS values were nearly identical. The BNT
with Ankudinov predictions slightly better. Around station underpredicted the measured data, but not by as much as some of
1880 to 1960, the PIANC formula overpredicted stern squat. the other ships. The Ankudinov predictions were about the same
In general, both Ankudinov and PIANC tended to underpredict as the PIANC predictions in this case and both excellent.
Comparison of Measured Ship Squat with Numerical and Empirical Methods 487
6.4. Discussion stern squat by factors of 0.67 and 0.66, respectively. The
Ankudinov formulas overpredicted measured bow squat by a
On average for the ships in this Panama Canal data set, the factor of 1.25 and underpredicted stern squat by a factor of
BNT underpredicts squat by 30% to 35% for the bow and 20% 0.98. PIANC underpredicted bow and stern squat by 0.93 and
to 45% for stern squat. The Ankudinov formula overpredicts 0.88, respectively. Thus, the BNT predictions were generally
squat by 25% for the bow and underpredicts by 5% for the smaller than the measurements but showed the same trends as
stern. The PIANC underpredicts squat by 15% for both bow the other predictors. The Ankudinov predictions are slightly
and stern. All of the 2978 data points were used in the compar- larger than the PIANC predictions, although the Ankudinov pre-
isons although there are a lot of turns or bends in this section of dictions match canal channel types like the Panama Canal better
canal. Ships experience acceleration, deceleration, and roll than PIANC. Thus, all three predictors appear to give reasonable
while turning, which affect squat but is not accounted for in predictions of ship squat and can be used with confidence in
these squat predictors. deep draft channel design.
Portions of the Majestic Maersk measurements look to be
inordinately large from station 1940 to 1950. According to
Dr. Daggett, who participated in field measurements of the Majes-
tic Maersk, they conducted an acceleration test in this range
Acknowledgments
starting from a dead in the water condition to full ahead to We acknowledge the Headquarters, US Army Corps of Engi-
see is if such an acceleration would cause extreme squat at the neers, Institute for Water Resources (IWR), the Naval Surface
stern as the ships wheel dug into the water. They observed a Warfare Center, Carderock Division, and TRANSAS for authoriz-
steadily increasing heel to starboard, which affected both ship ing publication of this article. It was prepared as part of the
squat and UKC. The ship heeled over three to five times and rolled Improved Ship Simulation work unit in the Navigation Systems
up to a 3 heel. He also noted that there was quite a variation in Research Program (CHL) and the IWR NETS program. This arti-
ship speed for this ship throughout the measurements. These cle is dedicated to Dr. Vladimir Ankudinov, who died March
unusual ship responses would explain some of the observed dis- 2012. We acknowledge the assistance of the CHL Prototype
crepancies between measurements and predictions. Measurements Branch and Larry Daggett, Waterway Simulation
These comparisons indicate that the Ankudinov formulas are Technology, for providing channel, ship, and DGPS data. We also
conservative in most instances because they tend to overpredict thank the reviewers for their useful suggestions.
ship squat. The BNT predictions are generally lower than the
measured values. Possible reasons for the smaller BNT predic-
tions might be that the actual ship lines were not used as a result
of proprietary issues. The use of generic ship lines, although References
appropriate, can misrepresent the water line beam and sectional
ANKUDINOV, V., DAGGETT, L., HUVAL, C., AND HEWLETT, C. 1996 Squat
area curves of the ships during transit. A sensitivity study by Kopp predictions for maneuvering applications, Proceedings, International Con-
(2011) showed that there can be a 3% to 8% variation in predicted ference on Marine Simulation and Ship Maneuverability (MARSIM 96),
squat as a result of 10% variations in fore and aft sectional areas. September 913, Copenhagen, Denmark, pp. 467 495.
BNT has shown better agreement with US Navy projects when ANKUDINOV, V., DAGGETT, L. L., HEWLETT, J. C., AND JAKOBSEN, B. K. 2000 Pro-
totype measurement of ship sinkage in confined water, Proceedings, Inter-
actual ship lines and measured model-scale data are available for national Conference on Marine Simulation and Ship Maneuverability
comparison. As previously mentioned, Briggs et al. (2010a) found (MARSIM 00), May 8 12, Orlando, FL.
good agreement between BNT and PIANC predictions for a range ANKUDINOV, V., AND BRIGGS, M. J. 2009 Email and telecoms with updates
of ship and channel types. Additionally, the BNT results have to equations 2, 4, 5, 10, and 18 in CHETN-IX-19, April 22 and 29.
been found to be comparable to those produced by more expen- ANKUDINOV, V. K., AND JAKOBSEN, B. K. 1996 Squat predictions at an early
stage of design, Proceedings, Workshop on Ship Squat in Restricted Waters,
sive higher-order computational fluid dynamics predictions. October 4, Washington, DC, pp. 48 69.
Therefore, variations in the generic ship lines that are most repre- BARRASS, C. B. 2002 Ship squata guide for masters, private report.
sentative of the ships in this study could have a significant effect Available at: www.ship-squat.com/research_papers.htm. Accessed July 3, 2007.
on the predicted squat. It should be noted that CADET is not BARRASS, C. B. 2009 Ship Squat and Interaction. Witherby Seamanship
International Ltd., Bell & Bain Ltd., Glasgow, UK.
restricted to using the BNT model results because the user can BECK, R. F., NEWMAN, J. N., AND TUCK, E. O. 1975 Hydrodynamic forces on
always import other specific or model test squat data. The PIANC ships in dredged channels, JOURNAL OF SHIP RESEARCH, 19, 3, 166 171. Sept.
predictions are based on averages of all five of the PIANC formu- BRIGGS, M. J. 2006 Ship squat predictions for ship/tow simulator. Coastal
las for bow squat but only two (Barrass and Romisch) for stern and Hydraulics Engineering Technical Note CHETN-I-72, US Army Engi-
squat. In some instances, one or more of the PIANC formulas neer Research and Development Center, Vicksburg, MS.
BRIGGS, M. J. 2009 Ankudinov ship squat predictionsPart I: theory and
might match measured data much better than the averages. FORTRAN programs. Coastal and Hydraulics Engineering Technical Note
ERDC/CHL CHETN-IX-19, US Army Engineer Research and Develop-
ment Center, Vicksburg, MS.
BRIGGS, M. J., AND DAGGETT, L. 2009 Ankudinov ship squat predictions
7. Conclusion Part II: laboratory and field comparisons and validations. Coastal and
Hydraulics Engineering Technical Note ERDC/CHL CHETN-IX-20, US
This article has compared BNT, Ankudinov, and PIANC ship Army Engineer Research and Development Center, Vicksburg, MS.
squat predictions with DGPS measurements of four ships in the BRIGGS, M. J., KOPP, P. J., SILVER, A. L., AND MATHIS, I. 2010a Comparison
of PIANC and CADET ship squat predictions, Proceedings, PIANC 125th
Gaillard Cut section of the Panama Canal. These ships included a Congress, May 1114, Liverpool, UK.
Panamax containership, Panamax bulk carrier, Panamax tanker, BRIGGS, M. J., VANTORRE, M., ULICZKA, K., AND DEBAILLON, P. 2010b Chap-
and a containership. In general, the BNT underpredicted bow and ter 26: Prediction of squat for underkeel clearance. In: Kim, Y. C., ed.,
488 Comparison of Measured Ship Squat with Numerical and Empirical Methods
Handbook of Coastal and Ocean Engineering. World Scientific Publishers, David Taylor Model Basin, Carderock Division, Naval Surface Warfare
Singapore, pp. 723774. Center, Potomac, MD.
CANADIAN COAST GUARD. 2001 Safe waterways (a users guide to the MILLWARD, A. 1992 A comparison of the theoretical and empirical predic-
design, maintenance and safe use of waterways), Part 1(a) guidelines for tion of squat in shallow water, International Shipbuilding Progress, 39, 417,
the safe design of commercial shipping channels. Software user manual 6978.
version 3.0, Waterways Development Division, Fisheries and Oceans NORRBIN, N. H. 1986 Fairway design with respect to ship dynamics and
Canada, Ottawa, Ontario, Canada. operational requirements. SSPA Research Report No. 102, SSPA Maritime
DAGGETT, L., AND HEWLETT, C. 1998a Study of ship squat in the panama Consulting, Gothenburg, Sweden.
canal. Waterway Simulation Technology, Inc., Columbia, SC. OHTSU, K., YOSHIMURA, Y., HIRANO, M., TSUGANE, M., AND TAKAHASHI, H.
DAGGETT, L., AND HEWLETT, C. 1998b Study of ship sinkage in the Panama 2006 Design standard for fairway in next generation, Proceedings, Asia
Canal, Phase IIlow water conditions. Waterway Simulation Technology, Navigation Conference, October 2021, Jeju, South Korea.
Inc., Columbia, SC. OVERSEAS COASTAL AREA DEVELOPMENT INSTITUTE OF JAPAN 2009 Technical
ERYUZLU, N. E., CAO, Y. L., AND DAGNOLO, F. 1994 Underkeel require- standards and commentaries for port and harbour facilities in Japan. Over-
ments for large vessels in shallow waterways, Proceedings, 28th Interna- seas Coastal Area Development Institute of Japan, Tokyo, Japan, 998 pp.
tional Navigation Congress, PIANC, Paper S II-2, May 2227, Sevilla, PIANC. 1997 Approach channels: a guide for design. Final Report of the
Spain, pp. 1725. Joint PIANC-IAPH Working Group II-30 in cooperation with IMPA and
GULIEV, U. M. 1971 On squat calculations for vessels going in shallow IALA, Supplement to Bulletin No. 95, PIANC, Brussels, Belgium.
water and through channels, PIANC Bulletin, 1, 7, 1720. ROMISCH, K. 1989 Empfehlungen zur Bemessung von Hafeneinfahrten,
HOOFT, J. P. 1974 The behavior of a ship in head waves at restricted water Wasserbauliche Mitteilungen der Technischen Universitat Dresden, 1, 3963.
depth, International Shipbuilding Progress, 21, 244, 367378. STOCKS, D. T., DAGGETT, L. L., AND PAGE, Y. 2002 Maximization of ship
HUUSKA, O. 1976 On the evaluation of underkeel clearances in Finnish draft in the St. Lawrence Seaway, Volume I: squat study. Prepared
waterways. Report No. 9, Helsinki University of Technology, Ship Hydro- for Transportation Development Centre, Transport Canada, Ottawa,
dynamics Laboratory, Otaniemi, Finland. Ontario, Canada.
ICORELS (INTERNATIONAL COMMISSION FOR THE RECEPTION OF LARGE SHIPS) TUCK, E. O. 1966 Shallow-water flows past slender bodies, JFM, 26,
1980 Report of Working Group IV, PIANC Bulletin, 35, suppl, 39 pp. Part 1, 8195.
KOPP, P. J. 2011 BNT sensitivity study for the Tanker Elbe (Unpublished TUCK, E. O. 1967 Sinkage and trim in shallow water of finite width,
Memorandum). David Taylor Model Basin, Carderock Division, Naval Schiffstechnik, 14, 73, 9294.
Surface Warfare Center, Potomac, MD. YOSHIMURA, Y. 1986 Mathematical model for the manoeuvring ship
KOPP, P. J., AND SILVER, A. L. 2004 Program documentation for the channel motion in shallow water, Journal of the Kansai Society of Naval Architects,
analysis and design evaluation tool (CADET). NSWCCD-50-TR-2005/004, Japan, 61, 200, 4151.
Comparison of Measured Ship Squat with Numerical and Empirical Methods 489