Uncertainties in the Wind-Heel Analysis of Traditional ...vm2330.sgvps.net/~syrftest/images/library/20150805143533.pdf · Greek symbols heel angle air density ... traditional sailing

1

Uncertainties in the Wind-Heel Analysis of Traditional Sailing Vessels: The Challenges it Presents for Forensic Analysis of Sailing Vessel Incidents,

Bruce Johnson, Honorary Member, Co-chair, SNAME Small Working Vessels Operations and Safety Panel

William Lasher, Member, Editor, Journal of Sailboat Technology and Professor of Mechanical Engineering, Penn State Erie, The Behrend College

Matt Erdman, Student, Penn State Erie, The Behrend College

Jan Miles, Captain, Pride of Baltimore II

Bill Curry, Captain, SV Concordia

ABSTRACT: There are many uncertainties in the interpretation of full-scale sailing vessel data taken under dynamic conditions, and even more uncertainties when forensic analysis is attempted based only on survivor’s recollections. Frequently, the analysis is based on static equilibrium assumptions, sometimes modified to steady-state motions of the wind and heeling response of the vessel. Dynamic conditions are generally non-deterministic and statistical methods must be used. Even more complicated is the non-stationary random process nature of most accidents.

In the wind-heel research carried out on Pride II, it has been shown that wave action frequently adds uncertainty to the correct attribution of contributions to establishing the cause of the resulting heeling action. The best data are found in steady 10 to 20 knot wind strengths in minimum waves found in the lee of a shoreline. This criteria can be interpreted as minimizing the uncertainties in characterizing the wind-heel performance of a given sail combination at normal angles of heel.

Examples of quasi steady-state response are presented in the paper as characterized by the Wind Heel Stiffness Ratio (WHSR), which is equal to the square of the apparent wind velocity in knots divided by the resulting heel angle in degrees. WHSR is not non-dimensional but is independent of the system of

units, (SI vs. EG). The WHSR for each sail combination is most easily established by a maneuver the crew of Pride II has deemed “The Crazy Ivan.” However, it is uncertain whether this concept can make useful predictions at heel angles higher than those beyond GZmax in the absence of any good data taken during these conditions. CFD studies of various sail combinations provide very good agreement between the recorded wind-heel responses of the vessel up to deck edge submergence. The corresponding CFD predictions provide a method of predicting the normal wind heel responses of a traditional sailing vessel during the design process.

The paper discusses operational guidance uncertainties that appear as a “fork in the road” decision, with bearing away as one path and heading up as the other. The paper examines the tradeoffs in the decision making process relative to the type of vessel involved and the observable wind and sea conditions at the time. Recent attempts to re-analyze the dismasting of Pride II in 2005 and the sinking of the SV Concordia off Brazil in 2010 are also included.

Lastly, the possible downward lift force involving square sails at high angles of heel needs to be investigated in wind tunnels since full scale testing of this concept is virtually impossible.

THE 21st CHESAPEAKE SAILING YACHT SYMPOSIUM

ANNAPOLIS, MARYLAND, MARCH 2013

2

NOTATIONA sail area

AW apparent wind

AWA apparent wind angle corrected for mast motion

AWV apparent wind velocity corrected for mast motion

BLC boundary layer correction factor to wind velocity profile

BPCorr bare poles correction for CAMF

CAMF# Combined Area-Moment-BLC Factor for sail combination number #

CAMFBP Combined Area-Moment-BLC Factor for bare poles

CD drag coefficient, D/(0.5 V2A)

CL lift coefficient, L/(0.5 V2A)

CS side force coefficient (sideways projection of CLand CD)

CSF side force coefficient to match average HM and RM from full-scale data

CFD computational fluid dynamics

CLR center of lateral resistance of hull

CSYS Chesapeake Sailing Yacht Symposium

D aerodynamic drag

GM Metacentric Height

GZ Righting Arm

HA heel angle

HM heeling moment

IMS International Measurement System

KG vessel center of gravity above keel

KG/T dimensionless measure of CG relative to waterline

L aerodynamic lift (a force perpendicular to the free stream wind

MCA Maritime and Coastguard Agency, UK

MWA measured apparent wind angle

MWV measured apparent wind velocity

RANS Reynolds-Averaged Navier Stokes Equations

RM righting moment

SA Sail area

T vessel draft

TW true wind

TWA true wind angle corrected for mast motion

TWV true wind velocity corrected for mast motion

VS vessel (ship) velocity

VCP vertical location of the center of pressure above the CLR

WHSR Wind-heel stiffness ratio

Subscripts i index for each sail

# reference to sail combination number

Greek symbols heel angle

air density

Figure 1 � Shorthand notation used to identify Pride II�ssails

M = mainsail F = foresail S = fore staysail J = jib 4L = 4 Lowers = M+F+S+JJT = jib topsail SQT = square fore topsail MGT = main gaff topsail FTG =fore topgallant SS = studdingsail (windward stun’sail) RT = ringtail TRYSL = storm trysail STMJ = storm jib

3

INTRODUCTION Various efforts to provide a basis for operator guidance for traditional sailing vessels have been made since the pioneering work of Beebe-Center and Brooks (1966-67), research which led to many of the current pass/fail rules and guidelines. Typical Stability Letters, however, provide no operator guidance and only discuss the stability limitations of a vessel from a pass/fail point of view. The analysis of the Pride of Baltimore I casualty by Chatterton and Maxham (1989) and discussed further in Parrott (2004) examined capsize and sinking events caused by extreme wind conditions.

Wills (1991) attempted an analytical investigation of gust performance in waves but concluded that mathematical modeling of the dynamics of wind�heel responses left much to be desired. Fossati et al. (2011) used wind tunnel tests performed in dynamic conditions to measure dynamic effects on sail plan aerodynamics caused by yacht motion induced by sea waves. Such studies may eventually lead to

operator guidance in terms of avoiding worst-case conditions involving both wind and sea conditions.

Operator guidance for traditional sailing vessels is described in the classic book Seamanship in the Age of Sail (Harland 1984). In chapters 15 and 16, Harland discusses the conventional wisdom of what to do in preparation for storms and squalls, including shortening sail and deciding whether to luff up or bear away. While these choices are described anecdotally by Harland, they had not been modeled and analyzed until SNAME began supporting full-scale measurements aboard Pride of Baltimore II in 2004. Progress made in the SNAME-sponsored wind-heel project has been reported at CSYS (Miles et al. 2007; Lasher et al. 2009) and at the 2009 SNAME annual meeting (Johnson et al. 2009). The current status of these investigations is reported at the 2013 CSYS in this paper and in Franzen (2013).

The MCA Large Commercial Yacht Code: Curves of Maximum Steady Heel AngleDeakin (1990, 1991, 2009) performed some systematic wind-heel studies in the UK and recommended a methodology for establishing design and operational sailing stability requirements. Based on these studies, the British adopted Deakin’s approach to sailing stability in the Large Commercial Yacht code, using the GZ value at the lesser of the downflooding angle or 60 degrees as the governing parameter. The intersection of the wind heeling lever curve with the GZ curve at this critical angle is converted to an upright Wind Lever at zero heel (WL0) using a division factor of Cos1.3. A “derived wind heeling lever” is then calculated that represents half the wind force of the Wind Lever arm at any angle of heel, and its intersection with the GZ curve gives the maximum heel angle at which a vessel could resist a steady 41.4% increase in wind speed (doubling the wind pressure) without heeling beyond the 60 degree (or downflooding) limit. Note that at the angles of heel up to deck edge submergence, the wind heel moment rolls off more rapidly than the GZ curve from a straight line, so the vessel effectively stiffens up slightly with heel angle in this range, as shown in Franzen (2013). The Cos1.3 factor replacing the Cos2 factor used in the Sarchin-Goldberg wind-heel criteria has been used in all the SNAME-sponsored tall ship research prior to this paper.

The MCA rules address two different types of operator guidance based on the current heel angle of the vessel: Gust curves and Squall curves. The Gust curve is for non-squall conditions while the Squall curve is for threatening weather. The wind speed in a Squall may be several times greater than the preceding mean wind velocity, and hence, there can be a multiple increase in wind pressure. The squall curves are based on comparing the righting arm at the downflood point or 60 degrees of heel, whichever is

smaller, and the righting arm at the average heel angle preceding a squall. The formulas and procedures then determine the value of the squall wind speed that is expected to heel the vessel to the downflooding point on the back side of the righting arm curve. Examples are generally given for 30, 45 or 60-knot squalls, (Figure 2) which will heel the vessel to the downflooding angle. The corresponding heel angle can be found for the vessel with the same sails set in the lesser wind speed preceding the squall. A set of squall curves based on the righting arm characteristics of the SV Concordia was the only operator guidance (other than the captain’s experience) on board that vessel when it capsized and sank in a sudden squall in 2010.

MCA Squall and Gust Curve Inputs and Assumptions(Deakin 1990, 1991 and discussed in Miles et. al. 2007, Lasher et al. 2009 and Johnson et al. 2009)1. Input regarding the current sail plan is not required. 2. Input regarding the apparent wind (AW) direction and

point of sail is not required. 3. Required inputs are steady heel angle ( °) and mean

apparent wind speed before the squall. 4. A reliable anemometer and inclinometer are required. 5. Sail plan and apparent wind angle will remain constant

during the gust, and the wind gust speed will not be more than 1.4 times the preceding mean wind speed which implies a doubling of the wind pressure.

6. As a sailing vessel heels, the (horizontal) wind heeling moment decreases and at any heel angle ( )° between 0 (upright) and 90 degrees, it is related to the upright value by the function:

HM = HM0° x Cos1.3

4

The rationale of the MCA approach is that knowledge of a vessel’s squall curves will better enable a master to judge the margin of stability-related safety when deciding how much sail to carry in conditions when squalls are possible.

Since the squall curves were provided to the SV Concordiaas operator guidance, they will be discussed in more detail later in this paper.

Figure 2 Generic Squall Curves from MCA

OVERVIEW OF WIND-HEEL RESEARCH ON PRIDE OF BALTIMORE IIFor the past eight years, SNAME has supported the full-scale measurement of wind speed and direction and corresponding heel angles on board the Pride of Baltimore II. Results from this project were first presented at the 2007 CSYS (Miles et al. 2007). That paper presented a first cut at providing operator guidance for appropriate sail selection in various wind speeds. The full-scale measurements also contributed to an understanding of the far-from-equilibrium dynamic stability situations experienced by the Pride II during the squalls that led up to her dismasting on September 5, 2005. These squalls were particularly interesting as the first (wet) squall gave adequate warning to prepare the vessel, while the second (dry) squall gave no warning of its occurrence or its ultimate extreme magnitude, which, while not threatening to the vessel’s stability, led to a rigging hardware failure and the ship’s subsequent dismasting.

Several uncertainties were present in this original data as a result of the data collection system being used. The 2005 data acquisition system did not have an integrated mast motion correction for apparent wind speed and direction. The first set of uncertainties stemmed from unknown time lags in the anemometer sensor and the heel/pitch angle sensor which was used to calculate the roll and pitch rates needed to make a mast motion correction to the measured apparent wind values (MWV and MWA). The heel and pitch angle measurements were central-differenced to get

an estimate of the roll/pitch rates and these were applied to the anemometer values to estimate AWV and AWA. Thus the mast motion correction with its unknown time shift was being applied to a base signal which was also time-shifted by some unknown amount, resulting in an estimate of the true apparent wind speed and direction that could not be used with any certainty.

The other uncertainty was in the method used to correct the apparent wind speed and direction measured by the Raymarine anemometer sensor for the effect of the surrounding sails, masts, and rigging. The vessel’s original wind sensor used for the 2005 data set was mounted on the forward side of the mainmast at the cross trees between the masts. To correct the anemometer sensor’s data for this location, an assumed basic “slot effect correction” was applied, the value of which was based on previous literature (this problem will be discussed later in this paper). The assumed correction was later proven to be invalid by a CFD analysis of the flow field velocities on the windward side of the sails. The CFD results showed that at the location of the anemometer, there are significant variations in both wind speed and direction, both of which are dependent on the vessel’s heading (Figures 7a and 7b). Recent attempts to calibrate the mid-level mainmast with the masthead foremast and mainmast data have also been inconclusive.

5

Figure 3a Pride II with the foremast B&G anemometer and the mid-mainmast Raymarine anemometer. A more detailed CFD flow field analysis discovered that the sail combination of 4 Lowers plus the Square Topsail (4L+SQT) used by Pride II in many wind conditions acts as a wind dam, slowing down the apparent wind and changing its local direction well to windward of the masts and beyond the location of the new wind sensors (Figure 3 does not include the mainmast B&G anemometer). In addition, the wind dam extends even further to windward and significantly slows the wind speed if the sails are over-trimmed while beam reaching (Figure 3b), a situation which could result from a large lifting wind shift that the vessel is unable to follow.

In the Johnson et al. 2009 paper, a section on the selection of sail combinations by the master of a traditional sailing vessel was presented. Results from the CFD modeling, spreadsheet modeling and full-scale testing were provided, and validation of the models was discussed. A method and format for presenting guidance on sail selection to operators of traditional sailing vessels was introduced. In the original project (Miles et al. 2007), a color-coded risk approach based on wind-heel relationships was suggested. The original color scheme has been modified to two shades of green for low-risk operating conditions, yellow for moderate risk operating conditions and red for high risk operating conditions

A more efficient model for predicting the sail forces and moments for various sail combinations was subsequently developed (Johnson et al. 2009). The new method was based on significant heel angle analysis of AWA2/HA1/3scatter diagrams from extensive full-scale tests using a number of sail combinations. The results of these verification data sets were then used to predict the response of untested sail combinations using ratios of combined area moment factors (CAMF) unique to each sail combination in a standardized spreadsheet analysis. The

Figure 3b Flow streamlines near the center of pressure of the lower sails with AWA = 90 degrees and the sails over trimmed use of a spreadsheet model that incorporated CAMF ratio scaling based on a variety of wind conditions in modest wave action appears to give conservative worst-case guidance. Note that the 2009 method does not require the use of mast motion corrections to the anemometer data for MWV2/HA1/3 because the highest 1/3 heel angle statistics showed nearly identical average values with larger standard deviations for the uncorrected anemometer data. All that is needed to record full-scale data on traditional sailing vessels consists of the vessel’s anemometer in a good location, a heel angle sensor with a similar time constant as the anemometer, GPS data for vessel speed and heading and a data logging system.

Figure 4 from Johnson et al. (2009) gives the operator a way of considering which sail combination to select for expected wind conditions. The chart represents many, but not all, of the sail combinations (gears) normally used in sailing Pride II, and it demonstrates how little is gained by small reductions in sail area on a multiple-masted traditional sailing vessel. The authors are interested in whether this form of operator guidance is more or less useful than Squall Curves for traditional sailing vessel captains and crews.

There are two primary factors to consider when choosing a sail combination aboard any sailing vessel, modern or traditional: the strength of the wind and the desired steering balance in terms of the amount of rudder set and weather or lee helm. (For traditionally rigged vessels, the inherent rig strength limitations also influence the master’s choice of sails for any given wind strength.) There is one generality that most all sailing vessel masters use in sail selection: the desire to have “a bit” of weather helm so the vessel is “steering stable.” In this condition, the vessel will tend to round up towards the wind in a gust. The latter is discussed extensively in the 2013 CSYS paper written by Iver Franzen (Franzen, 2013).

6

Figure 4 Operator guidance chart for CSF = 1.54: downflooding risk (water on deck) at a KG=12.45 feet and displacement = 186 tons. Note: Figure 10 below expands the individual combinations on a single set of

curves on a wind velocity vs. heel angle chart

THE “CRAZY IVAN” MANOEUVER A new method of characterizing the wind-heel response of various traditional sailing vessel sail combinations (called the Crazy Ivan manoeuver by the Pride II crew) has been developed and was presented at the 2012 Tall Ships America (Johnson and Miles 2012).

On Tuesday, 26 October 2010, the first “Crazy Ivan” manoeuver was executed onboard Pride II in the Chesapeake Bay near Annapolis with all sails sheeted flat for sailing as close to the wind as possible (see Appendix A for an explanation of the procedure). The manoeuvers were done with the 4L+SQT sail combination in a true

wind of between 12 and 14 knots. This experiment of bearing away with sails sheeted flat (the USCG criteria for sailing vessels) during a good breeze on the relatively flat waters of the Chesapeake had not been attempted previously. For this sail combination, the minimum (worst- case) ratio of AWV2/HA was predicted to be 28 in the 2009 SNAME transactions paper. These conditions were achieved three times, just before and after 10:46 am and just before 10:52 am during the sail (see Figure 5).

7

Figure 5 The original Crazy Ivan manoeuver recorded using a motion compensated masthead anemometer mounted on Pride II�s mainmast on October 2010.

For the 2009 SNAME paper, the time-based comparison had been abandoned too soon. We overlooked the fact that for a given true wind speed, apparent wind speed increases as the wind goes forward, and thus the heel angle increases as well (illustrated in time from 10:46 to 10:49 above). As the wind goes forward (90 to 60 degrees AW, 115-80 degrees TW), maximum heel occurs at 60 degrees apparent (true between 75-90 degrees). This shows that for both CFD analysis and wind tunnel testing, the effect of boat speed and heading on apparent wind speed must be considered; otherwise one would incorrectly infer that the worst case conditions occur with the wind on the beam

rather than forward of the beam. The maximum heel angles agree quite well with the Pride II operator guidance downflooding risk chart (Figure 4) if one interprets them as being based on true wind at, or just forward of, the beam. In actuality, with the wind forward of the beam (close reaching), there is a lift component to the heeling force as was suspected when the project was started. This combined contribution of lift and drag to the heeling moment (see Figures 6a, b, and c) helps to explain why the heeling force coefficient for the SNAME sponsored study was 1.54 rather than a lower drag-only coefficient of 1.0-1.2

.

Figures 6a, b, and c. Side force as a function of apparent wind for a Cal 40 sheeted for max drive, (courtesy of Iver Franzen)

8

During the week of September 19-26, 2012, Captain Jan Miles was able to record data sets that enabled the comparison of close-hauled sheeting angles and beam reaching sheeting angles for two different sail combinations. The Crazy Ivan results below are for Pride II sail combination 4L+JT+SQT+MGT, which has a

“worst-case” Wind Heel Stiffness Ratio (WHSR= 23). Coincidentally, this combination on Pride II has the same WHSR value as the estimated WHSR for the “reduced sail plan” that was set on the SV Concordia ” (Figure 15) just before an intense squall capsized her off Brazil. (Note that WHSR=AWV2/Heel Angle = (23 knots)2/23 deg.)

Figure 7a Mainmast data taken 19 September 2012. Note: The data are uncorrected for mast motion

Figure 7b Mainmast uncorrected data taken 19 September 2012Figures 7a and 7b present data taken by the mainmast anemometer on Sept 19, 2012 that replicate data recorded

in October 2010 (except for WTP2 mast motion corrections which are now only available to the foremast

9

anemometer). The earlier data demonstrated that the mainmast anemometers were giving apparent wind angles about 10 degrees too low (note the difference in the angle made by masthead pennants in Figure 3a), and subsequently led to additional funding from SNAME T&R for a second anemometer in order to take simultaneous measurements at both mastheads.

Note that the highest heel angles occur when the hollow green markers are very near the WHSR = 24 line. This corresponds to a TWA of around 80 degrees and an AWA

of around 50-60 degrees. Note also that the mainmast anemometer indicates that the sails are fully powered up to an apparent wind angle of 36 degrees, not likely for traditional sailing vessels. This data indicates that the sail plan generates distortions in the apparent wind field as a result of the circulation associated with the lifting sails. The scale on the left axis of Figures 7a, 8a and 9a is for heel angle, apparent wind speed and true wind speed. The remaining variables indicated by hollow markers use the scale on the right hand axis.

Figure 8a Foremast data taken on Sept 19th 2012 with the CFD results overlaid.Figures 8a and 8b present the same data using the B&G WTP2 mast-motion-compensated system on the foremast. The maximum heel angle (14+ degrees) occurs near 11:56 at an AWA of 54 degrees along with the maximum AWV of about 20 knots (TWV of 16+ kts). At that instant, the WHSR is 24. With the wind well aft (11:59-12:00) the WHSR drops to 18, but this could result from wave action during the period of low heel angle. The stable condition from 12:01-12:02+ yields a nearly constant WHSR of about 24 for TWA’s between 114 and 78 degrees corresponding to AWA’s between 78 and 60 degrees. Note that there is a difference in the calculated true wind angle depending on whether the vessel was bearing off or heading back up. This shows up as a “hysteresis loop” in the data plotted against apparent wind angle. This and other current data acquisition problems that add to the uncertainty will�we hope�be resolved during the next season.

The agreement between the experimental data and the CFD results is excellent considering that the sail shape is estimated and the sheeting angle is held constant, so the effect of the stalled flow on the sail forces appears to be realistic.

This also means that CFD-generated WHSR vs. Apparent Wind Angle plots can be used with confidence during the design phase of a vessel to insure that its stability is adequate, and also to produce safety guidance for the operator before the vessel is even launched! The simulation of the Crazy Ivan manoeuver in five-degree increments of apparent wind angle and a fixed sheeting position is very efficient since only one input must be changed for each computational run.

10

Figure 8b Foremast data taken on Sept 19th 2012 with the CFD analysis results overlaid.Superimposed on the WHSR data in Figure 8b are the CFD results for the same sail plan and apparent wind angles. Note that CFD_FS refers to the far-field apparent wind angle used as boundary conditions in the calculations and CFD_FM refers to the apparent wind angle at the top of the foremast from the CFD results. This difference will be further researched in the future as it is a source of uncertainty in interpreting CFD data based on the remote wind.

The Crazy Ivan results and the CFD results confirm that lift (forces perpendicular to the direction of flow) contributes a significant part of the heeling moment although this effect is not included in rules based on sails sheeted flat with the apparent wind on the beam. Sailing vessel stability regulations should probably be changed to reflect that true wind ahead of or near the beam causes the worst case heeling moment.The actual worst-case heeling moment is greater than the heeling moment due to the projected area drag forces alone, because of lift forces generated by the sails as the vessel moves forward under both steady state and dynamic conditions. With the sails drawing well and not luffing, there appears to be a wide range of apparent wind angles where the heeling coefficient is relatively stable, even in the lower apparent wind speeds as the wind goes aft. The effect on the combined lift and drag as the apparent wind goes aft and likely stalls the leeward side of the sail seems to adhere to a relatively stable value of WHSR. The data scatter with the wind aft of the beam is probably caused by small

wave-induce rolling motions superimposed on small wind induced heeling values as was observed in Pride II�s race from Bermuda to Charleston SC in 2009. Figures 15a and 15b in Johnson et al. 2009 show that the statistical 1/3 highest heeling values were nearly independent of apparent wind velocity in this downwind race.

To investigate the condition of sails not sheeted flat, a Crazy Ivan manoeuver was done with the same sail combination sheeted for beam reaching on Sept 26th. Figures 9a and 9b show the analysis of this data taken at lower TWV’s between 8-10 knots. These conditions appear to be insufficient to separate wind driven heel angle responses from heel angles driven by wave action. We will attempt to obtain better Crazy Ivan data for the beam reach sheeting conditions during the next sailing season.

Interestingly, for TWA’s on the beam (AWA’s between 70-80 degrees in Figure 9a and 9b), the WHSR is less than the assumed worst case of 23, adding to the uncertainty and requiring that we obtain additional full-scale data at much higher heel angles while reaching and running during the 2013 season. It could turn out that the worst case WHSR for sails sheeted flat is higher than the worst case WHSR for beam reach sheeting, but the minimum occurs well off the wind where the apparent wind is lower for a given true wind. Note in Figure 9a that the highest heel angle still occurs at an apparent wind angle of 60 degrees and the WHSR is where the prediction says it should be.

11

Figure 9a Foremast data taken on Sept 26th 2012 using beam reach sheeting.

Figure 9b Foremast data taken on Sept 26th 2012 using beam reach sheeting.Nevertheless, Figure 9b shows the advantage of luffing up in these conditions (in this case to less than 60 degrees AWA and 8-10 knots TWV), which effectively stiffens the vessel to the wind (much higher WHSR=AWV2 /HA) and reduces the heel angle from 5+ degrees to 2+ degrees in higher apparent wind speeds. Note that the advantage of heading up in a squall is quite apparent.

A note on WHSR: Wind-heel stiffness is not the same as vessel stiffness which at low angles is measured by GM. WHSR could be interpreted as the system stiffness to the wind as the sail configuration is reduced (which increases WHSR) or increased (which lowers WHSR). For a given righting arm curve it estimates how much apparent wind a powered-up sail plan can take for a specified heeling angle without luffing up to depower the sails or falling off to reduce the apparent wind velocity.

12

As one might expect, it appears that the WHSR value is dependent on the trim of the sails. Beam reach sheeting creates additional sail camber, resulting in greater lift force components. This fact could explain the lower values of the “Worst Case” WSHR in Figure 9b, adding uncertainty to this methodology. We are now at a stage where the “Worst Case - Close Hauled Sheeting” (Figure 4) developed for Johnson et al.( 2009) can be summarized in a new format ( Figure 10), which may or may not give the worst case for beam reach sheeting (as shown in Figure 9b). Resolving this uncertainty will require more research, including the validity of assuming that the WHSR remains

constant or at least increases beyond deck edge immersion all the way to GZmax .

Figure 10 expands the color coding risk assessment (on the x-axis ) in Figure 4 to predict the worst case heel angle for a given sailplan at a given apparent wind velocity. For examples of its use, apparent winds above 28 knots can drive Pride II to its GZmax with the sail combination used in Figures 7, 8 and 9. An apparent wind of 75 knots can drive Pride II to GZ max with only a trysail and storm jib up. To survive 98 knots with the wind near the beam one had better have bare poles only.

Figure 10. Worst case WHSR curves for significantly different sail combinations aboard Pride II.Note that the MCA squall curves discussed in the SV Concordia section of the paper also assume a constant value of AWV2/HA. They are obviously just parabolas with the horizontal and vertical axes in the WHSR curves in Figure 10 reversed. The Squall Curves lack any color- based risk- assessment bands.

We now have an operational dual anemometer data acquisition system aboard Pride II for further tall ship stability studies. What is needed is a more automated system to record strong wind events at much higher heel angles. We are preparing another proposal for T&R funding to relieve the ship captain of the distraction of starting to record data while simultaneously dealing with a stronger wind situation. At present we know little about

wind heel response during dynamic conditions such as those found in squall gusting conditions (Fossatti and Muggiasca, 2011). The sparse data taken at high angles of heel aboard Pride II in 2005 cannot be trusted because of the anemometer location and the damped heel angle sensor.

To extend the WHSR approach beyond our two-masted schooner database, Professor Susan Swithenbank of the Coast Guard Academy has volunteered to encourage student research projects on the wind-heel characteristics of USCG Eagle. We have initiated transferring the wind-heel data acquisition system currently mounted on the Brig Niagara to the USCG Eagle.

13

REEVALUATION OF THE UNCERTAINTIES IN THE FORENSIC ANALYSIS OF THE PRIDE II DISMASTING IN 2005Using the method discussed in Johnson et al. (2009) for characterizing the wind-heel relationship, it was decided to reanalyze the 2005 measured apparent wind data of the squalls leading up to the dismasting of the Pride II. The analysis was limited to the duration of each of the two squalls and the data were taken by the mid-mast Raymarine anemometer, which was lost during the event (See Figure 3a).

For the wet squall, Captain Miles considered risking sailing into it because he noted the rain under the isolated cloud was falling vertically, meaning no angular direction and there was no tell-tail wind signature on the surface near the rain line. As the apparently windless rain line approached he feathered the vessel up as a precaution for any actual increased wind inside the rain curtain. In the dry squall, though, Captain Miles was not able see the wind gusts coming and hence did not feather the vessel up or ease the sails in time to avoid the measured 40 to nearly 50 knot wind gusts. These gusts stressed the rig (seen twice in the recorded data, which is not corrected for mast motion),

rupturing the stress-fractured iron fitting holding the bowsprit down to the stem and dismasting the rig (Miles et al. 2007). Note that the dry squall experienced by the Concordia (discussed later) likewise gave little warning before the vessel started heeling to high angles from which recovery was unlikely.

This reanalysis of the 2005 data raised the question of what can be reliably determined from anemometer readings lacking mast-motion corrections at headings other than in beam and broad reaching situations. Johnson et al. (2009) demonstrated that both measured (MWV2/HA) and resulting (AWV2/HA) scatter plots visually give the same significant heel angle statistics regardless of the apparent wind angle. However, it should be noted that if the purpose of the measurements is to calculate the stress on the rig caused by wind gusts, the use of an IMU unit to enable a mast motion correction is necessary to get the corrected apparent wind. Without mast motion correction, determining whether or not the squall contained a lifting wind shift is nearly impossible.

Figure 11a & b Wet and Dry squall records using the uncorrected anemometer recordings (Slot factor 1)

14

Figures 12a and 12b Raw (measured apparent wind) data scatter diagrams for the first (wet) squall and the second (dry) squall on Sept 5, 2005 using uncorrected anemometer data (from Johnson et al. 2009).

Since the vessel was being sailed with the maximum prudent sail area for the wind and sea conditions, it is unreasonable to assume that the wind speeds recorded by the old mid-mast level anemometer (Figure 3) were correct since the WHSR for that sail combination is 28. By adjusting the so-called slot factor to 0.82, one can make the analysis show that AWV2/HA touches 28 as a worst-case scenario. This assumes that for the sail plan and sheeting being used on Sept 5th, a masthead anemometer

would have recorded an apparent wind velocity that was only 82% of that recorded in the slot between the foresail and the mast. A recent attempt to simultaneously record the B&G foremast anemometer and the replacement mid-mast Raymarine anemometer with the same sail settings suggested the inverse of the correction applied in Figure 13, further adding to the uncertainties in a forensic analysis of the dismasting

.

Figures 13a and 13b Slot effect 0.82 corrected records of the Wet and Dry Squall on Sept 5 2005

15

In the top chart (Figure 13a), Pride II began heading up at 1:31 towards an apparent wind angle of 20 degrees, unloading the sails and gradually increasing the WHSR from around 30 to 90. When it turned out there was down-drafting wind inside the rain that peaked to 35 knots between 1:34 and 1:35 (peak gust not correctly recorded as the mast is rapidly moving away from the gust), the vessel still heeled briefly to beyond deck edge immersion. Note that the mismatch between the anemometer response and the heel angle indicator response results in considerable uncertainty in calculating the instantaneous WHSR by dividing the square of the apparent wind velocity by a recorded heel angle with a different time constant.

In the second chart (Figure 13b), Pride II took what appear to be two sudden squall gusts without heading up ahead of time. The peak recorded heel angle of 33 degrees probably

would not have been greater because the bowsprit iron did not fail until later, at time stamp 2:27, although it is likely the bowsprit iron had been over-stressed, and hence weakened, prior to failure near 24 degrees of heel. The foremast did not start coming down until after time stamp 2:28, as can be seen by the drop in apparent wind angle at the same time as wind speed fell to around 20 knots and heel angle reduced to less than 5 degrees. The mainmast with the anemometer fell after 2:28:30. As mentioned previously, the lack of mast motion correction gives great uncertainty to any interpretation of the nature of this dry squall. The heel angle trace (dark blue time line) is steady between 10-15 degrees before the wind gust hit, at which time the vessel heeled to 33 degrees and then came back to 15 degrees, before heeling over again to 25 degrees, followed by a slow decent to zero.

UNCERTAINTIES IN THE FORENSIC ANALYSIS OF THE SV CONCORDIA SINKING IN 2010

Figure 14 Sail Training vessel Concordia under full sailOn February 17, 2010, the Barbadian-flagged sail-training vessel Concordia was knocked down and sank as a result of downflooding through open hatches, vents and doors. Because the vessel was chartered to a Canadian corporation, the Transportation Safety Board (TSB) of Canada conducted an investigation of the incident and subsequently issued a Marine Investigation Report

(M10F0003) that included findings as to the possible causes of the knockdown. Finding #4 as to Causes and Contributing Factors of the TSB report states that “an increase in wind speed, probably including a vertical component caused the vessel to heel to angles sufficient to immerse the leeside doors and ventilators.”

16

The report issued by NSOD/MSC-NCRI/Environment Canada (2010) contains the following conclusion:�Although the occurrence of a strong microburst can neither be confirmed nor denied from this meteorological analysis (due to a lack of information), a weaker downburst occurrence is much more likely in this case, because the intensity of the convection was weaker than many other cases have shown, when strong microbursts were evident.” (It should be noted here that the distinctions between microbursts, downbursts and downdrafts as a function of intensity are occasionally misunderstood.)

At 1420, the second officer observed that the heel angle of the vessel was approximately 23° and that the apparent wind speed was approximately 23 knots (WHSR = 23), most likely indicating that the sails were full and powered up in a squall with the apparent wind at 120-140 degrees. According to the TSB report “As the apparent wind speed increased with the onset of the squall, the vessel’s heel angle reached roughly 23° for approximately 2 to 3 minutes without mitigating action being taken. In response to a further, modest increase in wind speed, probably including a vertical component, the vessel began to heel beyond 23°. At this point, the action taken to steer

downwind was too late to prevent the vessel from heeling to angles sufficient to immerse the lee-side doors and ventilators.”. “If the vessel were affected by winds inclined from the horizontal (such as in the downdraft from a squall) the initial horizontal wind speed could have been reduced to between 23 and 31 knots.” (The range is a result of using two different heeling coefficients in estimating the heeling arm curves in the original version of Figure 23.) “The apparent wind continued to increase, and at 1423 the winds were estimated to be 26 knots climbing to 30. According to the TSB report, at this point the sailing vessel passed its critical point of GZmax and theoretically heeled continuously from 38° to 68-70°. Flooding through unsecured openings in the deckhouse doors quickly occurred and the knockdown was completed. Captain Curry believes that there was no hesitation in the heeling motion at 68-70 degrees and that the Concordia was knocked down in a continuous motion (this is strongly supported by the following CFD analysis). Thus the vessel completed the heeling motion to sails in the water before significant progressive flooding began. These discrepancies add to the uncertainty in a forensic analysis of the accident.

Uncertainty due to the possibility of inclined windsBecause of the sail configuration at the time of the incident (Figure 15), and the CFD modeling of the incident, it has been suggested that the roll-off model used may not be adequate in light of vessel wake shielding of the sails as demonstrated in Figures 18-21. The MCA roll-off model assumes that the heeling moment decreases by a factor of cos1.3( ), where is the heeling angle. It had been hypothesized that the large square topsails set high above the deck may be able to create a significant downward

component of force due to lift with the foot as the leading edge of the wing like topsail when heeled over. Downward lift produced in this condition could cause the heeling moment to be much larger than suggested by the MCA model, which like the USCG model, only considers profile drag. Downward lift could help explain that section of the heeling moment curve between heel angles of 45-65 degrees (see Figure 16a).

Figure 15a and 15b Sail Combination on SV Concordia during its capsize and sketch showing Downward Lift Force generated by heeled topsails. Note the topsails were braced two points ahead of the beam, as

shown in the CFD model.

Downward Lift Force

17

A preliminary CFD analysis at 23.5 knots apparent wind speed, as was observed by the watch, shows that the downward component of force on the two topsails does in fact contribute more to the heeling moment than the side force at high angles of heel. The heeling moment for the topsails along with the round-up moment is shown in the graphs below. Note that the maximum roundup moment around 45 degrees of heel and the downward lift component in the heeling moment exceed the profile drag component beyond 57 degrees. Both fall off sharply as the sails enter the stalled area downwind of the hull. Also note that the 41-degree heel photograph (Figure 29b)

corresponds to the maximum roundup moment attributed to the square sails, thus making falling off nearly impossible. (Also see Franzen 2013 for diagrams of this situation.) The bracing of the sails leaves the upper half of the squares fully loaded even at 70 degrees of heel (Figures 20 and 21), with the downward lift peaking between 50 and 60 degrees. The roundup moment (picture the thrust vector acting at the CE in Figure 15b) is a key factor that must be weighed when deciding whether to bear away or head up when hit by a squall.

Figures 16a and 16b - Heeling and Roundup moments generated by the two topsails on ConcordiaCFD analysis of the entire sail plan shown in Figure 15, however, suggests that the total heeling moment decreases at a rate of cos2.2( ) for this low aspect ratio sail plan�

much higher than the rate given by the MCA model. A comparison of these two roll-off rates is shown in Figure 17.

Figure 17 - Comparison of the heeling moment vs. heeling angle calculated by CFD compared to existing criteria assuming only horizontal winds.

The reason for the increased roll-off predicted by the CFD analysis is not entirely clear, but it may be due to the blanketing of the sails by the hull as it heels over, as illustrated in Figures 18 - 21. The sequence of images shows that as the vessel heels over from 30° to 70° degrees, the majority of the sail plan becomes encompassed by the low-speed separated flow region behind the hull. It seems like an unreasonable

simplification to assume that the heeling moment depends only on profile drag and that it rolls off with a single power of the cosine function. If the increased roll-off rate is true, it strongly suggests that some vertical (downward) component of wind may have been necessary to break through the wake of the vessel and provide the necessary force to complete the knockdown. Further research into this question is ongoing.

18

Figure 18 - CFD Velocity vectors at 30° of heel

Figure 19 - CFD Velocity vectors at 50° of heel

Figure20 - CFD Velocity vectors at 70° of heel

Figure 21 - CFD analysis at 70° of heel showing effect of hull form on wind path lines

Furthermore, note that only steady-state conditions are illustrated in the CFD analysis. The difficulty of analyzing dynamic effects including roll overshoot caused by a sudden gust is currently beyond our capabilities.

Because of this and other uncertainties, the knockdown sequence beyond 70 degrees must assume that a modest downburst of 35-49 knots could have been the source of the vertical component of the wind velocity which caused the knockdown and ultimately flooded and sank the vessel.

The following is quoted from a private communication between Captain Curry and Kenneth Pryor, of The Center for Satellite Applications and Research:

“The TSB report uses a threshold for microburst occurrence as a measured wind speed of 50 knots. This is not consistent with published literature that sets the minimum peak wind speed for microburst occurrence at 10 m/s (~20 knots). Ted Fujita and Roger Wakimoto (microburst experts), in their analysis of microbursts during the Joint Airport Weather Studies (JAWS) project, used the 20 knot threshold for microburst occurrence.

“In my study of downbursts that occurred over the Mid Atlantic coastal region during the summers of 2010 and 2011, only 9 out of 44 (~20%) events were associated with peak winds greater than 50 knots. I have found that the majority of measured downburst wind speeds were between 35 and 49 knots with an average speed of 46 knots, well below the 50 knot threshold. In U.S. coastal waters, convective storm winds of 34 knots or greater meet the threshold for a ‘Special Marine Warning’.”

Captain Curry supports Kenneth Pryor’s significant observations: 1) regardless of terminology, the consensus is that downburst events did occur at the time and place of

19

the Concordia knockdown and were most likely microbursts because of their scale, and 2) they were typical

microbursts, involving winds well under 50 knots and were not extreme events generated from a large storm.”

Figure 22 From Pryor 2010. Schematic of a vertical cross-section through a mature bow echo. Location “X” marks a likely position of SV Concordia. (Courtesy of COMET (1999))

A possible�but difficult to validate�interpretation of the effect of an inclined wind is that it would shift the heeling arm curve over as illustrated in Figure 23. Note that to clear the righting arm bump at 78 degrees, a horizontal

wind of at least 60 knots would be required, with no reduction in the heeling arm due to the sails entering the wake of the vessel. This horizontal wind scenario is therefore very unlikely.

.

Figure 23 Effect of inclined winds on the assumed wind heeling arm in 31 knots neglecting hull form shielding (Figure 6 based on the TSB Stability Report and also used in Captain Curry’s Riga

presentation).

Uncertainty caused by the time-dependent nature of downflooding Although Figure 24 already accounts for flooding through the open doors in the forward deckhouse, once progressive flooding began into the lower deck through the sanitary exhaust vent and possibly the forward mushroom ventilators, the intact righting arm curve becomes time-dependent and thus a static analysis is no longer appropriate. Captain Curry estimates that it took 15 to18

minutes to flood the buoyant volume to a bow-down trim angle to where the water could flow into the lower deck through various open forward hatches and then eventually down first the stairwell in the forward deckhouse and soon after the stairwell in the after deckhouse, which then flooded quite rapidly. Both stairwells were offset just to starboard of the centerline.

20

Figure 24 � Concordia RA curve showing a downflooding angle of 56 degrees. The extended increases in RA’s are only temporary unless all watertight doors, hatches and vents were closed to prevent progressive flooding. The assumed horizontal wind speed for the heeling arm curve was 27 to 37 knots, depending on

the heeling coefficients used. (TSB 2010) The rise in the TSB-generated RA curve beyond 70 degrees is thought to be caused by the aft deckhouse. The aft deckhouse was assumed to be watertight until the entry points into the space were significantly immersed (about 88 degrees). The radio room window and wheelhouse door (which is raised above the main deck level) immersed first, at 68.6 and 74.2 degrees, respectively, and were assumed to be temporarily watertight, which would give a small

initial trim by the bow, lowering the forward downflooding possibilities as shown in Figure 25. However, Captain Curry did not observe this trim by the bow and survivors did not observe any hesitation in heeling past 70 degrees, so the calculated increase in static RA between 70 and 90 degrees may not be applicable in resisting a dynamic knockdown.

Figure 25 Possible downflooding points: Port side mushroom vent and centerline hatch to lower deck (circled in yellow)

The photo of the foredeck clearly shows the mushroom ventilators servicing the compartment forward of the collision bulkhead and the compartment immediately abaft that, which was Concordia�s deck stores and bosun's locker. Also visible are the two centerline hatches for these compartments, which were regularly accessed by the crew. Captain Curry doesn’t know at what angle of heel these

hatches would have flooded as that data is not listed in the tables (page 25) in the TSB report. However, from what he knows of centerline hatch flooding angles he suspects is was probably over 80 degrees. Although the forward hatch was only accessed when getting paint stores, and both hatches were kept closed when not in use, Captain Curry assumed they were open at the time of the knockdown.

21

Captain Curry was told by the TSB that they did not calculate any rates of downflooding at any angle of heel, and specifically, the critical arc in question from approximately 68 to 90 plus degrees. Concordia passed

through this arc in seconds, so there is much uncertainty whether or not downflooding to the hull during this period would have been significant enough to explain a full knockdown from horizontal winds.

Figure 26 Photo of the sanitary vent (middle right side of photograph,) which was connected via ducting to compartments below the weather deck. TSB has calculated its downflooding angle to be 65 degrees (See

Table 1). The open hatches to this forward deckhouse flooded at 56.5-58 degrees, so this space is not counted as a watertight volume in Figure 24 above.

Table 1 Downflooding angles from TSB Canada report. Note that rotating the vessel about the upright center of flotation yields a galley door immersion of 39 degrees. Because of the shape of the hull form, the

vessel lifts up as it heels over, giving an angle of galley door immersion of 56.5 degrees Note that cabin deck in the general arrangements plan (Figure 27) has watertight bulkheads near the waterline below the accommodation deck. There are also several watertight bulkheads in the accommodation deck with the amidships watertight door open. Progressive flooding while in a knockdown state is slowed by gradual

penetration of water from forward vent entry points through the cabins, which act as baffles to the flooding water at 90+ degrees of heel. Note that a possible reason the vessel eventually capsized to keel-up was because of most of the reserve buoyancy was in the remaining watertight portions of the lower deck.

22

Figure 27 General Arrangements of the Concordia Captain Curry does not have any photos of the stern deck with the hatch to the steering flat, but that hatch was closed and dogged at all times Concordia was underway and was only accessed by the engineers. There were also two quite small mushroom vents (approximately 4 inch diameter). As far as we know, the principal downflooding points (the forward companionway just to starboard of the centerline, the engine room fiddley skylight and the after companionway just to starboard of the centerline) may not have initially downflooded in the 68-90 arc.

Captain Curry has suggested an interesting solution to the deckhouse downflooding problem (Figures 28a and b) that would provide protection in accessible deckhouses against progressive downflooding of lower decks through the internal centerline companionways.

Figure 28a Possible Deckhouse Modification

23

Figures 28 b Deckhouse modifications to add reserve buoyancy following a knockdown to sailing vessels with access needs underway

This and other solutions will be considered during a follow-on project to examine design alternatives for

preventing progressive flooding after a knockdown since being in the wrong place at the wrong time has a low but non-zero probability. The authors are seeking examples of traditional sailing vessels that actually recovered from a knockdown to learn how this was accomplished.

In a previous section, Figure 9b shows that beam reach sheeting can result in lower values of WHSR for a given sailplan but this may be true only for small heel angles. Photographs taken by crew members and contained in the Concordia report show the booms eased well out as (see Figure 29). But note the vertical-oriented rigging tackle between the midpoint of the boom and the deck edge. This tackle would prevent boom lift hence preventing the gaff from swinging further outboard when encountering an increasing wind strength. Such lack of outward gaff swing could contribute to drive aloft, increasing the heeling moment rather than reducing it had the gaff swung outboard freely.

Figure 29a and 29b Photographs taken aboard Concordia just before the knockdown and capsize. TSB Canada identified the left photo as 29 degrees of heel and the right photo as 41 degrees of heel

(Courtesy of Erica Trimble)

Figure 30 GZ comparison between Pride II and Concordia (without heel offset) showing why the Figure 15 sail plan produces the same WHSR response as Pride II with everything but the foretop gallant up.

Note that Concordia�s KG was well above the waterline (KG/T = 1.27)

24

A modified Figure 10 graph could look like Figure 31, but only if the shielding of the hull is ignored. As pointed out previously, the WHSR of Concordia just prior to the capsize (23) is similar to that of Pride of Baltimore II (a much stiffer vessel) when Pride II is flying all sails except for the topgallant and studdingsail. So even the conservative value of 23 (23 degrees of heel at 23 knots) demonstrates that the 41 degrees of heel only requires 30 knots (without shielding effects) with a considerable uncertainty since GZmax occurs at 35 degrees as in the case of Pride II. It is possible that there was a gradual transition from WHSR=23 to around WHSR=40 as the sails were shielded by the heeled hullform. This increased wind-heel stiffness would require 50 knots at 60 degrees of heel unless there was a vertical component of wind which decreased the shielding effect and maintained the WHSR

at 23 to a much higher heel angle. This uncertainty is difficult to interpret and needs further confirmation.

With the sails eased, the Pride II experience with beam reach sheeting (at low angles of heel) shows no improvement in WHSR and a possible lower wind-heel stiffness because of increasing lift component contributing to heeling moments. The SNAME wind-heel team regards any heel angle beyond that for GZmax to indicate a high risk (black) situation because any significant vertical wind component can add enough heeling moment to finish the knockdown process even with hull form shielding of the greatest heel angles. Note that Concordia at WHSR of 23 enters the black zone just below 30 knots and even if the WHSR increased to 40 because of hullform shielding at high angles, the 37 knot wind still puts Concordia in the black zone (a situation that should be avoided).

Figure 31 WHSR Curves for Concordia based on beam reach sheeting as shown in the photographs Because of the high deckedgesubmergence height, the yellow warning zone is gone.

It is highly unlikely that any of these predictions of wind heel response are valid beyond GZmax . That is one reason to color the rest of the curve in black and that applies to the MCA squall curves as well. The question is, is the parabola assumption is conservative at high angles of heel? And does this assumption have limited applicability to winds with vertical components? If one is heeled over to 30 degrees does the method outlined in Figure 23 really work? There is much uncertainty to attempt to resolve in

the next few years. Wind tunnel tests at high angles of heel, instrumented so downward lift force components can be measured, are needed. Also needed is CFD modeling of winds with vertical components striking the water surface in the vicinity of the vessel, a very complex challenge.

25

Uncertainties in the use of MCA squall curves aboard Concordia for operator guidance.

Figure 32 The Squall Curses on board Concordia for operator guidance (from the TSB Canada report) Note that the Max Steady heel angle is less than deck edge immersion

These squall curves look suspiciously like parabolas except at the lowest values so we should be able to match these curves with WHSR curves, which are also parabolas.

Figure 33 shows an attempt to match the MCA Squall Curves for Concordia using the appropriate equations for the match.

Figure 33 Concordia Squall Curves developed as WHSR curves as explained in the example calculation.

2

2

30

2

45

=

Where Wind-Heel Stiffness Ratio for the MCA criteria is found from

30 /1.414 450 18.75. 24

45 /1.414 1012.5. 24

kts

kts

AWVWHSRResulting Heel Angle

WHSRMax Steady Heel Angle

WHSRMax Steady Heel Angle

2

60

42.2

60 /1.414 1800 75.0. 24ktsWHSR

Max Steady Heel Angle

26

COMPARISON BETWEEN MCA SQUALL CURVES AND AUTHORS’OPERATOR GUIDANCE METHODOLOGYThe SNAME supported operator guidance curves were developed for Pride II as described in Johnson et al. (2009) and summarized in Figures 4 and 10. They are all based on establishing a Wind Heel Stiffness Ratio for each sail plan based on the Combined Area Moment Factor outlined in the 2009 paper. In the latest version (Figure 10) they give an estimate of the risk of exceeding the heel angle for the maximum righting arm, a conservative guidance parameter since it may account for vertical wind components that reduce the shielding effect of the heeled hullform which in turn may act to increase WHSR. WHSR is based on the measured apparent wind velocity, not the reduced velocity in the wake of the hull. This is still being explored.

Limitations of MCA Squall Curves 1. The information cannot be used as a definitive

prediction of heel angle in a squall since squall speeds and corresponding wind shifts cannot be reliably predicted.

2. There is no specific guidance provided on how to reduce the sail plan to achieve lower heel angles. This is left to the Master. Sail plan reduction greatly

depends on the size and complexity of the vessel and the time it takes to reduce the sail area

3. The curves assume that AW direction does not change; only the heading steered by the vessel can be changed. This implies that the WHSR for a given sail combination remains constant with heel angle up to the downflooding angle, a dubious assumption beyond GZmax .

4. The curves assume that heeling moments are computable beyond GZmax, by using a single heeling arm correction based on cosx for all vessels, another dubious assumption.

5. Since the downflooding heel angle is the principal benchmark, this raises the question as to whether this criterion provides a suitable margin of safety. In a sudden squall it is possible that the dynamic motion of the ship could cause the heel angle to increase beyond the downflooding angle once GZmax is exceeded.

6. Finally, since it is based on the assumption of horizontal winds and criteria based on RA characteristics beyond GZmax, it provides no warning of the dangers of vertical wind components. WHSR-based criteria may include this risk assessment but this has yet to be proven.

Fork-in-the-Road: Operator guidance/seamanship during a significant wind increase

In the report on the loss of Concordia, published by the Transportation Safety Board of Canada, it is stated that the vessel was sailing with the wind just abaft the beam, and that a rain-filled squall to windward and on an intersecting course was being tracked visually and on radar. The watch officer did not bear away in front of the rain and continued on a convergent course with the squall because he did not perceive any threat from the cell since it lacked signs of increased wind. When increased wind associated with the cell materialized, the vessel increased her heel from approximately 10 degrees to 23 degrees (as indicated by the video) and sailed at this increased angle for about two minutes. Captain Curry believes that during this period the vessel could have been quickly and effectively run off before the wind, as he had specifically balanced the vessel to facilitate a turn to leeward when he had chosen her sail plan hours earlier. The mate, being unaware that the increase in heel angle portended a problem, maintained his course. Two minutes into the squall, the wind suddenly increased and pushed Concordia to an extreme angle of heel near her maximum righting arm. At this point, the watch leader tried to turn away downwind. The Concordiacontinued her rapid heeling action until the sails entered the water. She never recovered, eventually flooding, capsizing, and sinking. That Concordia did not respond to the effort to bear away after being struck by the squall raises a question of what

might have happened if the helm had been shifted to turn toward the wind instead. Sailing masters of square-rigged vessels today describe how their vessels do not turn quickly, particularly at significant heel angles. Hull form, full length keels, and the low aspect ratio and longitudinal distribution of sail plan are contributing factors to this lack of turning agility. Instinctive perception of the influence of the sail plans, and in the specific case of Concordia�s pre-squall adjusted sail plan, suggests that a forward concentration of sail area would facilitate bearing away. Normal operating heel angles have proven this out. Various accounts of experiences at extreme angles of heel point to an inability to bear away under those conditions. Steering control in those accounts was regained only after the angle of heel decreased, either through a reduction of sail area or wind strength.

Operator guidance exploration of the above raises the question of the reliability of both assessments of the potential wind strength associated with visible approaching concentrated falling rain, and the wisdom of bearing away before a sudden increase of wind. Review of the Pride IIdata during the wet squall shows that there is a sudden increase of heel to 30 degrees at 1:34:30 (Figure 13a), yet there is no corresponding wind increase measured. Meanwhile, there is little change to the AWA to indicate significant change in conditions or vessel heading. Is it possible there was a change in the angularity of the wind

27

inside the wet squall? Interpretation of data from the two squalls (wet & dry) encountered by Pride II and accounts of the loss of the Concordia together seem to point to significant uncertainty in reliably assessing the possibility and magnitude of increases in wind strength in squally conditions.

By waiting until after the onset of a sudden increase in wind that could cause a vessel to heel to an extreme heel angle, one risks losing the opportunity to safely bear away in front of an approaching squall. Bearing away before the impact of increased wind, however, is likely to be a successful strategy to prevent heeling to extreme angles.

Another question related to the provision of operator guidance might be termed a “fork in the road” for dealing with sudden increases of wind. Specifically, in a situation in which bearing away is not working, would reversing the strategy and heading up have beneficial (or even safe) results? Some hull forms and rig types are known to be

able to round up more easily than others, and to benefit more if they do, if bearing away is not successful (Franzen 2013; see diagrams depicting round up torque from offset center of effort). As described above, the hull forms and rigs of square-rigged vessels do not respond quickly to their helm. We do not know of any contemporary descriptions describing a successful effort to bear up in order to reduce angle of heel with a square rigger (as opposed to a topsail schooner) in a sudden increase of wind that created extreme heel. While the physics of the influence of the location of the center of effort and center of resistance apply to all sailing vessels, there are clearly different degrees of influence when accounting for different hull forms, rigs and sizes of vessel. This difference is an area of interest for operator guidance in the instance of preparing for significant wind increase and for alternative operator guidance when trying to reduce extreme heel that has already occurred. More research is needed on this topic.

CONCLUSIONS 1. A method of presenting operator guidance concerning wind heel relationships for each sail combination has been developed for traditional sailing vessel operators. The original color scheme used in Johnson et al (2009) has been modified to two shades of green for low risk operating conditions, yellow for moderate risk operating conditions and red for high risk operating conditions. Black is used for operation beyond GZmax which is a realistic upper limit given the uncertainty of actual heeling moments beyond that angle. The uncertainty arises because the heeling moment rolls off in an unpredictable manner due to flow shielding by the heeled hull and the resulting variation in lift and drag-induced heeling forces. The original bar graph listing for several sail combinations is supplemented by curves of steady-state heel angle versus apparent wind velocity, which are parabolic curves based on constant values of the heeling coefficient. All of these methods attempt to use risk assessment as the basis for operator guidance rather than the more typical pass-fail rules used by government and international agencies including the MCA Squall Curves.

2. A more efficient and useful method for predicting the sail forces and moments for various sail combinations has been developed using data obtained from a manoeuver called the “Crazy Ivan”. The new method is based on significant heel angle analysis of Wind Heel Stiffness Ratio (WHSR=AWV2/HA) diagrams from full-scale tests using a number of sail combinations. The results of these verification data sets are then used to predict the response of untested sail combinations using ratios of combined area moment factors (CAMF, Johnson et al. 2009)) unique to each sail combination in a standardized spreadsheet, or alternatively using CFD analysis. The use of the spreadsheet model incorporating CAMF ratio scaling

based on a variety of wind conditions in modest wave action appears to give conservative worst case guidance.

3. The question of under what circumstances a vessel can be considered to be in quasi-static equilibrium has been discussed but remains unresolved. In Johnson et al. (2009) it was shown that reasonably stable statistical values of the parameter AWV2/HA indicate that the vessel may be considered to be in the near-equilibrium steady state condition. However, in the wind-heel research carried out on Pride II, the authors have learned that wave action frequently adds uncertainty in establishing the cause of the resulting heeling action. The best data are found in steady 10 to 20 knot wind strengths in minimum waves found in the lee of a shoreline. This criteria can be interpreted as minimizing the uncertainties in characterizing the wind-heel performance of a given sail combination at normal angles of heel.

4. The general agreement between the current authors’ proposed operator guidance based on full-scale measurement and CFD analysis for a variety of sail combinations is sufficiently good to expand to other vessels which can help generate a limited set of wind speed vs. heel angle data. This will be done next season on USCG Eagle, and then on other traditional sailing vessels after the methodology is fully developed. Note from Johnson et al. (2009) that the new method does not require using mast motion corrections to the anemometer data for MWV2/HA. All that is needed is the vessel’s anemometer in a good location, a heel angle sensor and a data logging system.

5. The real advantage of using WHSR-based operator guidance is that it appears to be conservative in over-estimating the risk for criteria based on a horizontal wind

28

assumption because it may ignore the reduction in heeling moment resulting from hull form wake beyond GZmax. Thus the worst case WHSR-based risk assessment may be valid for down bursts of moderate strength but verification will take considerable additional research.

6 The possible downward lift force involving square sails at high angles of heel needs to be investigated in wind tunnels since full scale testing of this concept is virtually impossible.

ACKNOWLEDGMENTS This work was supported by generous instrumentation grants from the SNAME Technical and Research (T&R) Steering Committee. The authors wish to acknowledge the invaluable assistance of Mike Jones of P-Yacht in supporting the needed instrumentation and data acquisition packages. Iver Franzen provided the new sail balance and heel force diagrams as a function of apparent wind angle, Ken Pryor provided the references on vertical wind effects. Paulo Ekkebus and Abigail Fyfe provided a TSB Canada perspective and Fabio Fossati provided feedback with his reviews. Special thanks to Stephen Duff for his extensive editorial review of the semi-final draft which greatly improved the readability.

REFERENCES Beebe-Center, J. G. and Brooks, R.B., 1967, “On the Stability of Sailing Vessels”, SNAME Chesapeake Section, Washington, D. C., 1966, revised 1967.

Chatterton, H. A. and Maxham, J. C.,1989, “Sailing Vessel Stability-with Particular Reference to the Pride of Baltimore Casualty, Marine Technology, Vol. 26, No. 2, April 1989, pp 87-104.

Deakin, B., “Model Test Techniques Developed to Investigate the Wind Heeling Characteristics of Sailing Vessels and their Response to Gusts�,1991, Proceedings of the Tenth Chesapeake Sailing Yacht Symposium, Annapolis, MD, February, 1991, pp. 83-93.

Deakin, Barry, 1990. The Development of Stability Standards for UK Sailing Vessels, Transactions of RINA, 1991.

Deakin, Barry, 2009. Stability Regulations of Very Large Sailing Yachts, 10th International Conference on Stability of Ships and Ocean Vehicles (STAB09), St. Petersburg, RU.

Fossati, F and Muggiasca, S.,2011, “Experimental Investigation of Sail Aerodynamic Behavior in Dynamic Conditions”, Journal of Sailboat Technology, Article 2011-02,SNAME, 2011.

Franzen, I.,2013,�A Refinement of the Method Used to Determine the Balance of a Sailing Vessel During the Design Phase, with Application to Sail Design and Subsequent Sail Selection and Sailing Operations,”Proceedings of the 21st Chesapeake Sailing Yacht Symposium, Annapolis, MD, March 12-13, 2013.

Harland J. 1984 Seamanship in the Age of Sail, Naval Institute Press,

Johnson, B. and Miles, J. 2012 “Sailing Vessel Stability, Part 3: SNAME-Pride II Approach,” Tall Ships America Safety Under Sail Forum. Accessed 25 November 2012. <http://tallships.wordpress.com/2012/01/18/safety-under-sail-forum-stability/>

Johnson, B., Miles, J., Lasher, W. C. and Womack, J., 2009 “Operator Guidance Based on Assessing the Wind-Heel Angle Relationship of Traditionally-Rigged Sailing Vessels”, Transactions of the SNAME, 2009.

Lasher, W.C., Musho, T.D., McKee, K.C and Rybka, W.,2007, “An Aerodynamic Analysis of the U.S. Brig Niagara”, Proceedings of the Eighteenth Chesapeake Sailing Yacht Symposium, Annapolis, MD, March 2-3, 2007, pp. 185-197.

Lasher, W. C., Tinlin, D.R., Johnson, B, Womack, J, Miles, J. C., Rybka, W. and Heerssen, W. 2009 “Assessing the Wind-Heel Angle Relationship of Traditionally-Rigged Sailing Vessels”, Proceedings of the Nineteenth Chesapeake Sailing Yacht Symposium, Annapolis, MD, March 20-22, 2009.

MCA, The Large Commercial Yacht Code (LY2), Section 11.2.2 of MSN 1792 (M), Maritime and Coastguard Agency, U. K., 2007.

Miles, J.C., Johnson, B., Womack, J. and Franzen, I, 2007, “SNAME’s Stability Letter Improvement Project (SLIP) for Passenger Sailing Vessels”, Proceedings of the Eighteenth Chesapeake Sailing Yacht Symposium, Annapolis, MD, March 2-3, 2007, pp. 165-184.

NSOD/MCS-NRC/Environment Canada (2010) Brazil Coast Incident 17 February 2010 at 1722Z Meteorological Analysis

Parrott, D.S.,2004 Tall Ships Down, The Last Voyages of the Pamir, Albatross, Marques, Pride of Baltimore, and Maria Asumpta, McGraw Hill, 2004.

Pryor, K. L., 2010: Recent developments in microburst nowcasting using GOES. Preprints, 17th Conference on Satellite Meteorology and Oceanography, Annapolis, MD, Amer. Meteor. Soc. Pryor, K. L., 2010: Microburst applications of brightness temperature difference between GOES Imager channels 3 and 4. arXiv:1004.3506v3 [physics.ao-ph]

29

Transportation Safety Board of Canada Marine Reports � 2010-M10F0003 Knockdown and Capsizing-Sail Training Yacht Concordia 300 miles SSE off Rio De Janeiro, Brazil

Wills, J.A.B., “Recent Research on Wind Loading”1991, Phil. Trans. R. Soc. Lond. A (1991) 334, 229-240.

Appendix A: An Overview of the Crazy Ivan Procedure 1. This procedure is best accomplished at a minimum of

8 degrees of heel, ideally between 10-20 degrees of heel during the maneuver under minimum wave action and roll.

2. Set appropriate sails for a given sail plan to obtain the desired heel angle. Aim for 3-5 different sail plans, with a square sail set if the vessel has one

3. Start off at close-hauled with the sails trimmed appropriately for close-hauled. Begin the manoeuver with the sails just luffing, then slowly turn away from the wind but with sail trim held constant for each Crazy Ivan trial

4. Continue to fall off until on a broad reach (145-165 degrees). The whole manoeuver should take about 10

minutes so that there is sufficient data to give a good plot and that the steady state assumption is valid.

5. Slowly head back up to the point of luffing. 6. Note start and stop time in a data logger file. Note the

sail plan. 7. Repeat the process using different sail plans and sail

trims (i.e. beam reaching trim).

Necessary Data Required for Analysis:

1. Save data in a .csv or excel file. 2. Apparent wind speed and apparent wind angle data. 3. Heel angle data 4. Time to the second. 5. Boat speed if available from GPS or through hull.

Appendix B Overview of the CFD model (also see Lasher et al. 2007)The hull of the ship was created in DesignModeler. The hull was scaled from the hull diagram provided in Appendix B of the TSB report. This model included the two deckhouses located on deck. The sails themselves were modeled as surfaces using the dimensions scaled from Figure 1 of the TSB report. The areas of these sails were made to match the areas of the sails at time of capsize. These areas can be found in the appendix of the Stability report for the Concordia. The sails were then placed in their corresponding regions and trimmed based on the information provided in the TSB report. Exact sail shape and trim angle were estimated based on both provided information and knowledge of authors. The hull and the sails were place in a volume that was 300 meters wide, 300 meters long and 100 meters high in order to allow room for the flow to fully develop. The model was then run at numerous heeling angles with a uniform inlet velocity of 23 knots using the Realizable k- turbulence

model. Heeling moments at different wind speeds were scaled based on the ratio of the square of the velocities.

For the spars and rigging, the heeling moment was calculated using a simple drag model with a drag coefficient of 1.13. This heeling moment was reduced by cos( ) in order to account for the inclined angle of the wind, and by the cosine of the apparent wind angle, and the result was added to the CFD heeling moment.

According to the TSB report, at the time of 1420, the apparent wind speed recorded by the anemometer was 23.5 knots at an apparent wind angle of 88.9°. The measured heel angle at the time was 23°. The required wind speed at the anemometer location in the CFD simulation would have to be 23.4 knots to produce the heeling moment necessary to heel the ship to 23 degrees, thus validating the model.