3rd High Performance Yacht Design Conference Auckland, 2-4 December, 2008

Measurement and Simulation of Pressure Distribution on Full Size Sails
William Graves 1 , wsgraves@mit.edu Todd Barbera 2 , t.barbera@comcast.net J.B. Braun 3 , jbb@prg.northsails.com Len Imas 4 , limas@stevens.edu

Abstract. Technological advances in wireless networking equipment and atmospheric pressure sensors have enabled the measurement of the pressure distribution across full size sails in normal operation. The pressure measurements may be generated separately on the windward and leeward sides, and on other parts of the rig. Pressure maps of the entire sail can be produced in real-time to provide a detailed time-dependent picture of the pressure distribution, showing the fluctuations and deviations from nominal values due to any changes of the rig, boat motion, or wind. This information is useful in the design process, where sail designers and research aerodynamicists can quantify real-world, full-scale measured differences against computer simulations or alternative sail shapes. It also can be used by sail trimmers to set the best sail shape rapidly and in changing conditions. The pressure sensors are small and lightweight, allowing them to be applied in high density across the entire sail. Laptop software collects and analyses the data, displaying the pressure values and logging the data for post-analysis. This paper provides a description of the technologies used to implement the wireless pressure measurement system, presents examples of the data obtained, and compares the measured results with CFD simulations. NOMENCLATURE AWA AWS CFD FSI kbps m/s MEMS Pa Pa/m TWA TWS VMG ΔP ΔCp ρ Apparent wind angle Apparent wind speed Computational fluid dynamics Fluid-structure interaction Kilobits per second Meters per second Micro-electro-mechanical systems Pascal Pascal per meter True wind angle True wind speed Velocity made good Pressure difference Coefficient of pressure difference Air density sensors now enable the measurement of the pressure distribution across full size sails in normal operation. Sensor systems using this advanced technology are now commercially available [1]. The information generated is invaluable for understanding the accuracy and limits of the computer models in use, for quantifying the actual forces at work, and for developing sails with better performance. Furthermore it may be used directly by the sailors to adjust the sail shapes to achieve the best lift and drag. In addition to actual sail pressure measurements, pressure measurements can be extended to include other geometry interacting with the sail, such as mast spreaders and boom. For quantitative comparison of measured and simulated results several factors contribute to the accuracy and validity of the solutions. These include the physical description of the environment and the flow, turbulence modelling, aerolasticity and fluid-structure interactions, and computational mesh type and quality. There has been little validation of the accuracy of these tools through direct measurement of the dynamic pressure on different parts of the rig and sails of full size yachts. In the past the most common method of measuring sail performance has been to gauge the sail force and efficiency through changes in the yacht’s speed and attitude, limited by roll moment. The present work has moved beyond this to measure the actual pressure distribution on different parts of the rig. This paper first discusses the value of measuring the pressure directly, followed by the challenges the

1. INTRODUCTION Simulation of aerodynamic flows around sailing yachts or yacht components has been undertaken in various studies utilizing Navier-Stokes based CFD solvers with quantitative success. Advances in solver technology, and structural dynamic modeling of membranes, have enabled sail designers and research aerodynamicists to implement high fidelity aerodynamic and fluidstructure interaction (FSI) simulation models of yacht and sail aerodynamics. However there is little validation of these tools and how they relate to actual sail aerodynamics. Technological advances in wireless networking equipment and atmospheric pressure
1 2 3 4

Principal Research Scientist, Massachusetts Institute of Technology President & CEO, Cima Potencia, Ltd Designer, Performance Racing Group, North Sails and Aerodynamics Coordinator, BMW-Oracle Racing Associate Professor, Department of Civil, Ocean, and Environmental Engineering, Stevens Institute of Technology / Aerodynamics CFD engineer at BMW-Oracle Racing at time the presented work was performed

measurements must overcome, and then the technical details of the sensor units, followed finally by a presentation of the results and discussion of future directions. 2. WHY MEASURE SAIL PRESSURE? CFD programs have long been available to predict pressure distributions and the optimum sail shapes to produce the desired pressures. But these programs are able to model only a limited subset of the conditions that exist in the real world. The sail shape is but one factor in generating the best pressure distribution. Even with the same mean apparent wind angle and speed, the same sail shape will not necessarily result in the same driving force and drag due to the differences in air turbulence, shear, and sea state at different times and different locations. Gaining the ability to measure changes in lift and drag over time and under different conditions quantifies an important performance metric that today depends on the skills of individual sailors. Crews are experienced in generating sail-shapes that give good performance, but they do not in fact have direct knowledge of the pressure distribution itself, the optimization of which is the goal of sail shaping. One important result of the ability to measure the changes in lift and drag is to quantify the changes in sail shape needed. The ideal situation is to measure directly the pressure distribution in order to optimize it, rather than the sail shape which is secondary. The shape that a sail-design code recommends cannot account for turbulence due to weather or nearby structures, or for variations in wind shear, or transient or periodically changing motion such as wave and wake encounters, or sudden acceleration and deceleration due to race maneuvers and tactics or nearby competition. All of these factors have a significant impact on boat performance, which can be quantified and optimized through measurement of the sails’ pressure distribution. Measurement of the pressure distribution on full-scale sails has been limited to dedicated tests [2] with equipment that is often cumbersome and prone to damage [3]. Measurement of the pressure distribution on sails and rigs during racing conditions and with realtime feed back of the data to the sailors has been impractical with these systems. This is in spite of the fact that the instantaneous pressure distribution determines the lift and drag of the sail and rig and thus is the primary determinant of boat speed. 3. FULL-SCALE MEASUREMENT ISSUES There are many challenges to measuring the dynamic sailing pressure accurately. The sailing pressure varies only slightly from the nominal atmospheric pressure, which at sea level is approximately 105 Pa. Atmospheric pressure varies with altitude at the rate of ~13 Pa/m, so that the absolute pressure variation from the mast base to its top is 100 – 200 Pa, which is

Figure 1: CFD generated pressure map showing ΔCp. similar in magnitude to the dynamic sailing pressure. The atmospheric pressure is always changing due to weather, varying by as much as 3000 Pa. The dynamic pressure generated by the sails is just a small fraction of that. Figure 1 shows the delta Cp across the jib and main for a sloop. The pressure difference from windward to leeward is ΔP = ½ ΔCp×ρ×AWS2, so that the maximum pressure difference is 150 Pa for an AWS of 10 m/s. Most of the sail is experiencing lower pressures so that the sensors should be quite sensitive, with a resolution of ~1 Pa or better, or about ten parts per million compared to the atmospheric pressure, a difficult specification to meet under the challenging environmental conditions of a boat at sea. Further important attributes for the sensing units include low static pressure offsets and low noise fluctuations so that the pressure values from different sensors can be directly compared with one another for quantitative study. A height change of just 10 cm corresponds to a pressure change of over 1 Pa, so that motion of the boat can significantly alter the pressures. Similarly weather system pressure variations cause the reference pressure to continually change at the level of a few Pascal over timescales of seconds. One method of overcoming these sensitive variations is to simultaneously compare pressures at different points to each other. For instance, sensors can simultaneously measure the pressure on the windward and leeward sides of the sail and compare their values so that only small differences are apparent. Any changes due to common motion or changing weather are subtracted, leaving only the differential pressure across the sail.

Battery charger connector

Central processor unit Wireless unit Data acquisition unit 14 bit analog-todigital converter

On/Off switch

Rechargeable lithium-ion battery 12+ hours between charges Height = 7 mm

Pressure sensor

Figure 2: 2nd generation miniature pressure device containing pressure sensor, ADC, CPU, battery, and wireless unit. The sampling frequency of the measurements should be at least 1 Hz to provide prompt feedback during changing conditions, and may be considerably higher if the user is interested in studying rapid fluctuations or turbulence. While maintaining a high rate, the ADC must accurately convert the pressure readings. The sensor ADCs must have at least 17 bits of accuracy to achieve sub-Pascal resolution over the full range of atmospheric pressure. Most small, low power pressure sensors are based on MEMS technology using silicon substrates. These devices are sensitive not only to pressure but also to changes in applied voltage or ambient light level (photovoltaic effects). Thus measurements at sub-Pa resolution require care with the mechanical mounting and electrical circuit design. The power consumption of the units must allow at least a full day of sailing without recharge, and the sensor units should be compact so as not to interfere with the sail shape or flow. Furthermore the sensor units should be robust, able to withstand mechanical shocks due to flogging or contact with the rig, and must be weatherproof and splash proof. The sensors should of course be lightweight and small so that they do not interfere with the sail, and they must easily attach and detach for measurements on different sails. In addition to the above requirements on the individual sensors, the entire measurement system must be capable of operating reliably in a difficult environment. The wireless communications must have adequate range, even in the presence of obstructions, the different units must all be synchronized to within tens of milliseconds so that the data represents simultaneous measurements, and the networking protocol should be robust to loss or replacement of sensors, or intermittent signals. 4. HARDWARE & SOFTWARE DESCRIPTION A system of miniature, wireless, digital atmospheric pressure sensors has been developed that address the issues above. Their performance is a significant

advance in the ability to measure the actual pressure distribution on aerodynamic structures such as sails, rigs, or other industrial devices. Figure 2 shows photographs of the sensor unit without its case. The unit size is approximately 100 mm long X 50 mm width X 7 mm height plus a 2 mm thick case. The pictured unit represents the second generation of sensors, which has undergone extensive testing during which we came to understand the many requirements on the units and the system. The results of those tests have been integrated into the design of the next (3rd) generation units. The 2nd generation achieved a resolution of 6 Pa using a 14 bit ADC. It used a single pressure sensor per wireless unit with a 1 Hz sampling frequency. The 3rd generation, just beginning operation now, incorporate a number of improvements. The new sensor element with integrated ADC measures with 19-bit resolution, equivalent to 0.2 Pa. These sensors are very small (5 mm diameter), allowing multiple pressure sensors to be integrated onto a single board. The multiple sensor layout has a number of advantages. The sensors can be ganged so that the effective sampling rate is 40 Hz, which is useful for data processing with digital filter techniques, or different sensors can sample different areas of the sail. The housing is designed so that the sensors are isolated in their own pressure chambers. The pressure port can be routed for example through the sail or to another area of interest. With appropriate design it may be useful for boundary layer investigations, but the current design and its placement are targeted for measuring variations in the dynamics of the sail shape and the corresponding static pressure distribution. It is not yet meant to serve as a tool for measurements of turbulence pressure fluctuations inside the boundary layer of the sail. The new units also use less power resulting in longer battery life, and have a higher communications rate of up to 250 kbps. After some trial and error, mounting techniques and housing design have been developed that result in good isolation from light and water, and enable operation in the difficult environment of a racing sailboat. The sensor unit housings are resistant to shock, water intrusion, and dust, meeting IEC standard IP67. The sensors may be left installed for extended periods of time with a battery life that is currently around 16 hours of continuous operation. The sensors may be remotely switched into or out of a low power state to preserve battery power allowing testing over multiple days without need to recharge. The software consists of two packages, one performing low-level hardware tasks and the other for high-level user interaction. The low-level package performs data acquisition, power management, and network communication among sensors and with the laptop base-station. The high-level package presents several user screens for control and display of data. It also performs numerical analysis on the data, and logs it to



10 Wind Speed (knots) Boat Speed (knots) 20 9.5 TWS AWS Boat Speed

operating simultaneously. For very large systems, more sensors can be accommodated by using multiple base-stations. Network software dynamically determines the best route to the base-station, allowing data to hop from sensor to sensor to reach the basestation. This permits robust communications over even very large areas (hundreds of square meters).




5. MEASURED BOAT PERFORMANCE A number of experimental studies have been carried out to characterize the pressure sensor measurements and their correlation with other boat performance measurements. Figure 3 displays several minutes of measurement data obtained on board a 25 m sloop crewed and helmed by professional sailors. The boat instruments were well calibrated. The data acquired and displayed in the plots includes TWS, AWS, and boat speed in the top plot, dynamic pressure on windward and leeward sides of the jib and main in the middle plot, and TWA and AWA in the bottom plot. The vertical dashed lines identify times of particular interest that highlight aspects of the forces present as a result of changes in wind or boat handling by the skipper and crew. At all times the boat is sailing closehauled. However there are subtle differences in the sailing mode at different times. From the start of data at 14:27:00 until time T2 the boat is sailed in highmode, where the sails are trimmed relatively flat and the object is to point as close as possible to the wind while maintaining good speed. At T2 the bow is pointed down a few degrees, the sails are retrimmed for maximum power, and the boat sails at higher speed in low-mode while still maintaining good pointing to windward. At time T3 the boat changes to an intermediate mode between low-mode and high-mode, and sails that way until the very end of the data when it luffs into the wind. At 14:27:00 the boat is close-hauled and sailing in high mode in about 12 kts of wind. The mean TWA is ~27 degrees with an oscillation of +/- 2 degrees. The oscillation has a period of about 1 minute and is due to changes in both TWS and the helmsperson alternately bearing off slightly to build speed and heading closer to the wind to achieve best windward VMG. This is a common technique for maintaining best performance. The method’s impact on boat performance is quantified by this data, which can be used to directly compare the relative performance of 2 boats with different sail configurations or to evaluate trimming techniques and driving strategies. At 14:27:00 the boat speed and the TWA are both at local maxima of their oscillations indicating that speed has built as the boat bore off the wind slightly. The pressure data shown in the middle plot is also at a maximum at that time. The sailing pressure on the main and jib at the sensor locations is found by subtracting the (negative) leeward pressure from the (positive) windward pressure. The sailing

10 14:27:00 14:27:43 14:28:26 14:29:10 14:29:53 14:30:36 14:31:19 Time




Main Windward Pressure (Pa)

Jib Windward Main Leeward Jib Leeward




14:27:00 14:27:43 14:28:26 14:29:10 14:29:53 14:30:36 14:31:19 Time


35 Wind Angle (degrees)




5 14:27:00 14:27:43 14:28:26 14:29:10 14:29:53 14:30:36 14:31:19 Time




Figure 3: Top plot is TWS and AWS vs time. Middle plot shows windward and leeward pressure on jib and main. Bottom plot shows TWA and AWA. The vertical dashed lines identify times T1, T2, T3. See discussion in section 5. disk for later use. Both the low-level and high-level packages can be modified to accommodate new features and functionality. The analysis and display account for ambient atmospheric pressure and variations due to change in height, showing only the pressure due to wind force. The logged data is available to other software applications in real-time via a standard TCP/IP port. The system is scalable and fault-tolerant, automatically adding or dropping sensors as they are added or removed. The sensors self-form their own local radio network allowing signals to be transmitted robustly from the masthead to deck. A single base-station can handle up to 256 sensors

Table 1: Reconstructed sail shape parameters from images during sailing.
Stripe Camber Draft Front Back Entry Exit Twist (%) (%) (%) (%) (deg) (deg) (deg)

4 3 2 1

4.34 40.7 76.14 6 42.6 77.61 7.66 40.48 73.89 5.79 47.23 70.49

59.29 10.07 4.47 10.17 72.74 11.12 8.18 9.25 67.19 19.29 8.23 7.49 60.32 7.77 5.58 3.41

Figure 4: Photo showing placement of pressure sensors on sail at different vertical locations. pressure on the mainsail at 14:27:00 is 100 Pa + 18 Pa = 118 Pa, and on the jib is 113 Pa + 98 Pa = 211 Pa. Note that the windward pressure on both main and jib are very similar but that the jib has substantially higher leeward pressure, in agreement with CFD models. The mean absolute pressure that determines whether the sailing pressure is positive to negative is set by the pressure value during tacks when it is equal on both sides of the sail. Common-mode variations in pressure due to changes of sensor altitude from boat heel, pitch or wave action are subtracted. At time T1 the helmsperson has completed one cycle of his steering oscillation. Halfway to T1, the TWA has reached a minimum indicating that the boat is headed relatively high upwind while at the same time the speed has reduced and the pressures have reduced to 87 Pa + 16 Pa = 103 Pa on the main and 91 Pa + 61 Pa = 152 Pa on the jib. The 28% reduction of pressure on the jib is more than twice the 13% reduction on the mainsail. It is the reduction in driving force on the sails that slows the boat, and this is apparent in the data, where the change in sail pressure is consistently several seconds ahead of changes in boat speed. With practice, crew can use the pressure data to know in advance that the boat is about to slow or accelerate. At time T1, the first oscillation is complete with the boat speed, TWA, and pressures resuming their maximum values (for highmode). Two more oscillations occur from T1 to T2. At T2, the helmsperson cracks off to a TWA of ~35 degrees +/- 2 degrees with a resulting increase in boat speed and pressure. The TWS goes through a cycle, returning to mean strength again at T3. At T3 the pressure is at a maximum and the boat speed is building toward its highest point. The main pressure is 114 Pa + 19 Pa = 133 Pa and the jib pressure is 108 Pa + 112 Pa = 220 Pa. These values are 13% and 4% higher than the maximum values in the high mode for the main and jib respectively. The oscillations near T3 are again due to TWS and helmsperson changes.

It is evident that pressure changes in the sail correspond closely to changes in the rig tune, boat handling, and wind. The pressure changes foretell change in boat speed and can be used to optimize boat performance, particularly in changing conditions where constant adjustment is needed. 6. COMPARISON TO COMPUTER MODELS In addition to comparing the pressure sensor data to performance measurements on board, it is very useful as a design tool to validate CFD simulation. A test has been carried out to measure the vertical pressure distribution on the mainsail, simultaneously record the sail shape in detail, and then to reconstruct the sail shape and experimental wind conditions in a CFD simulation to compare its solution to the measured data. Sensors were placed at mid chord as indicated in Figure 4. For comparison to the steady-state CFD solution data from the sensors was averaged over small oscillations during steady-state sailing to windward. The condition simulated was selected from a time of consistent TWS, TWA, and boat speed. During the testing period images of the sail shape were recorded and then analyzed to document the flying sails shape. Stripe 4 is near the top portion of the sail at the 7/8th height, stripe 3, 2, and 1 are ¾, ½, ¼ height respectively. The geometry is described in Table 1. The flying sails shape was regenerated using the North Sails sail design software suite [4] with the flying sail shape exported from Membrain. Membrain is a program package for structural analyses of sails, mast and rigging. The core program attempts to calculate the deformed sail and mast geometry resulting from the aerodynamic pressure acting on the sails plus gravity and any other load applied on the mast through the rigging. The package includes supporting programs that prepare data for execution, and it post processes the results for data visualization. The deformed geometry is iteratively calculated by balancing external loads and internal stresses until a static equilibrium geometry is reached. The aerodynamic pressure load can either be defined once for all data in the preparation phase or the pressure can be updated at regular intervals as the iteration progresses. The latter method requires that Membrain is run in tandem with an external flow solver, and that these two programs exchange data during the iterations. Membrain was used to adjust the computer simulation until the flying sail shape matched the images from the




Height (m)





Measured CFD

0 0 20 40 60 Pressure (Pa) 80 100

Figure 6: The color variation represents changes in ΔCp as generated by a CFD model (only upper part of rig shown). In addition the measured data at 4 sensor locations is shown as color-coded spots at the level just above the forestay attachment.

Figure 5: Vertical pressure distribution along mainsail for measured and reproduced flying shape. measured tests. The converged sail shape was analyzed in North Sails Flow to develop a pressure map of the entire sail. North Sails Flow is a low order source/doublet potential flow model for thin surface sails and full bodies (hull and rotating mast). Both data sets were converted to delta pressure across the sail at the pressure sensor locations. The results of the measurements and CFD analysis are shown in Figure 5, showing good agreement in the reconstructed pressure at the sensor locations. In the context of CFD validation, namely validation of RANS tools, the use of the proposed pressure sensor measurement systems is self-evident. Measured pressures obtained over sailing conditions categorized in terms of local Reynolds number and aerodynamic angle of attack may be used to validate input parameters related to assumptions about the unsteady/steady nature of the flow in the near-field of the sail, turbulence physics and corresponding model selection, mesh and resolution both inside and outside the sail boundary layer. CFD models based on such calibration and validation information would not only lead to an improved quantitative prediction of full-scale sail aerodynamics but may also better explain differences observed between model-scale and fullscale aerodynamic flows over sails. Figure 6 shows a CFD map of the ΔCp distribution that reproduces one of our sailing tests. It also shows corresponding measured data at 4 points, again showing good

Sensors on sail and mast

Figure 7: The sensors are shown arrayed on the sail and rig. At high density entire pressure maps can be produced as in Figure 8.

agreement between measured and modelled pressure distributions. 7. CONCLUSIONS AND FUTURE DIRECTIONS The wireless digital pressure sensors have been shown to accurately measure the dynamic sailing pressure on full-scale sails in normal use. This has been demonstrated both through correlation of changes in boat performance with changes in pressure readings, and through careful comparison of measured pressures with the results of CFD simulations that reconstruct the sail shapes and wind conditions encountered in the tests. The sensors have been developed through several generations to achieve the current accuracy and reliability, which must be maintained while operating in the harsh marine environment, and without adversely affecting the boat’s operation or performance. The data gained and the resulting insights into performance can be used in at least two ways: first to validate the computational tools used for sail design and performance prediction, and second as a sailing instrument that can provide immediate display of the effects of changing wind conditions, or turbulence due other boats or structures, or changes in rig trim or helming. The sail pressure changes are a direct measure of changes in the driving force in the sails in contrast to secondary measures such as the boat speed. Although the data shown in this paper was generated with the 2nd generation sensor units, the 3rd generation sensors are now under test. They demonstrate significantly improved properties including better

resolution (<1 Pa), higher sampling frequency, and smaller size and power consumption. Future work includes generation of pressure maps over the entire rig. It is possible to generate a complete map of the pressure distribution, similar to the CFD map of Figure 6, over the entire rig by increasing the number of sensors as shown in Figure 7. The number of sampling points is then sufficient to interpolate an accurate pressure map across the entire rig. This idea has been tested by generating a CFD pressure map, sampling it only at the sensor locations shown in Figure 7, and recreating the entire map as shown in Figure 8. That display includes RMS noise as measured in the sensors and so is an accurate representation of the experimental pressures. The maps can update constantly so that the sailors observe the changing pressure distribution as it occurs. It is saved to disk for later use to analyse the sail performance and compare with design codes, as well as review the crew performance. Acknowledgements The authors wish to thank members of the BMWOracle Racing team for discussion and support.



Figure 8: Simulated full pressure maps on port and starboard sides recreated from samples at discrete points with realistic statistical variations to model sensor data. Such a map can be generated in realtime updating at 1 Hz to produce a continuous display of sail pressure maps while underway.

References 1. Pressure sensor systems from BB&G Technology Inc. See http://bbgtech.net Puddu, P., Erriu, N., Nurzia, F., Pistidda, A., Mura, A., “Full scale investigation of one-design class catamaran sails”, Proceedings of High Performance Yacht Design Conference, Auckland, 2006, 131 – 136. Flay, R.G.J. and Millar, S., “Experimental considerations concerning pressure measurements on sails: wind tunnel and full-scale”, Proceedings of High Performance Yacht Design Conference, Auckland, 2006, 123 – 130. See the design software description at site http://na.northsails.com/NorthTechnology/SailDesi gn/DesignSoftware/tabid/8522/Default.aspx