ARTICLE IN PRESS Journal of Wind Engineering and Industrial Aerodynamics 92 (2004) 663–685 Wind effects on emergency vehicles J.-P. Pinelli*, C. Subramanian, M. Plamondon Florida Institute of Technology, Melbourne, FL, USA Abstract The paper presents the results of a study to define the wind speed limits and conditions beyond which fire and rescue vehicles should not be operated during a hurricane. For that purpose, reduced scale models of a typical fire truck, ambulance, and sports utility vehicle (SUV) were tested in a wind tunnel and surface pressures and overall forces were measured. For the fire truck wind tunnel tests, the pressure measurements at low yaw angles are compared with full-scale measurements on a real truck and the results are also compared to those from some numerical simulation, and give some confidence in the wind tunnel tests technique. Based on the results of the tests and the analyses, safe wind speeds are defined for the operation of these fire and rescue vehicles to prevent course deviation and overturning. r 2004 Elsevier Ltd. All rights reserved. Keywords: Emergency vehicles; Hurricanes; Wind tunnel tests; Field tests; Coefficient of static friction; Critical wind speed; Vehicle aerodynamics 1. Introduction When a hurricane hits a municipality, the fire and rescue department plays a very active role in helping with the preparation, evacuation, and recovery tasks. Consequently, fire department, as well as law enforcement, personnel and vehicles might have to be on the road even when the wind speeds are so high that the safety of the vehicles might be compromised. It is therefore important to define the threshold wind speed beyond which vehicles should not be allowed to operate. Likewise, in the event of an evacuation it is important to define when the fire vehicle can safely evacuate. *Corresponding author. Tel.: +321-674-8085; fax: +321-674-7565. E-mail address: pinelli@fit.edu (J.-P. Pinelli). 0167-6105/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.jweia.2004.03.008 ARTICLE IN PRESS 664 J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 Some fire and rescue departments do not allow vehicles on the road with sustained winds of 15 m/s (35 mph) while others may use a wind speed of 20, 25, or 27 m/s (45, 55, or 60 mph). These wind speed limits vary from department to department because there are no undisputed scientific criteria on which to base the decision. So, each department must rely on its experience and subjective evaluation of the risk involved when defining the wind speed limits. For this reason the Florida Department of Community Affairs asked a group of researchers to develop a set of criteria to define unequivocally the upper limit wind speeds at which different fire and rescue vehicles can operate safely. This paper describes the results of this study for a fire truck, an ambulance, and a sports utility vehicle. Various related studies have been done on the effects of wind on road/railroad vehicles. Gawthorpe and Baker [1,2] discuss the problems of vehicles in cross winds, including increased vehicle drag, course deviations and/or overturning. Gawthorpe [1] also suggests requirements of model testing such as ground simulation, scale effects, wind flow simulation, a moving model approach, and CFD techniques. Schmidlin et al. [3] looked at the effects of tornados on vehicles. Coleman et al. [4] 1 describes a series of wind tunnel tests performed on 50 th scale tractor–trailer model mounted on a model bridge deck. This study shows that turbulence significantly lowers the pressures on the roof of the vehicle, resulting in a higher lift force. Baker et al. [5] measured the aerodynamic forces and moments on vehicles in gusting 1 crosswinds by propelling a 50 th scale articulated lorry model across a wind tunnel with simulated atmospheric turbulence. The results from this study show the importance of modeling both turbulence and the moving road condition. Baker [6] describes a method of quantifying accident risk for road vehicles in cross winds through a series of mathematical equations based upon the vehicles weight, tire forces, aerodynamic forces and moments and driver response. There are several problems associated with this simplified method. One is that driver behavior is almost impossible to quantify. Hurst et al. [7] compared the data of a full-scale tractor–trailer combination to wind tunnel model tests (both at a zero yaw wind angle) with different Reynolds numbers. In the testing a moving ground was simulated with a moving belt beneath the model. In this study the only appreciable differences between the full-scale and model tests were in the magnitude of suction behind the forward edges. Ryan et al. [8] studied the aerodynamic forces induced upon a generic hatchback model of a passenger vehicle as it passed through a perpendicular crosswind jet with a relative yaw angle of 30
. The project described in this report had a relatively modest budget, and did not benefit from sophisticated wind tunnel facilities. Within these constraints, and following on the steps of these previous researchers the strategy adopted for this research was to investigate the effect of storm winds on the emergency rescue vehicles though a series of wind tunnel tests, validated to a limited degree by full scale field tests and numerical simulation. The wind forces and moments acting on the vehicles for different wind speeds were estimated from the tests. The critical wind speed occurs when the wind induced forces and moments exceed the resisting effects. ARTICLE IN PRESS J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 665 2. Wind tunnel tests 2.1. Wind tunnel setup Testing was conducted in a low speed wind tunnel facility at Florida Institute of Technology. This wind tunnel has a square test chamber section of 54 cm  54 cm and an overall length of 1.52 m. The wind tunnel inlet has a 14:1 contraction ratio with less than 0.1% turbulence intensity (see Pinelli et al. for more wind tunnel details [9]). The maximum attainable test section speed is 22 m/s. The models tested 1 1 were a fire truck (25 th scale) an ambulance (21 st scale), and an exact Chevrolet 1 Suburban (SUV) replica (19 th scale). The exact dimensions of the fire truck model were 0.31 m length, 0.09 m cab height, and 0.11 m frontal cab width. The ambulance and SUV had the dimensions of 0.32 m  0.14 m  0.15 m and 0.27 m  0.07 m  0.09 m respectively. Fig. 1 shows the fire truck model in the wind tunnel. The models were supported on a wooden platform that served the purpose of dampening vibrations and mimicking the ground effect a vehicle would experience. The platform, essentially a rectangular plate, had dimensions of 0.30 m width by 0.50 m in length. The platform extended 0.08 m in front of the fire truck model, 0.10 m to either side, and 0.12 m to the rear. Large enough arguably so that the boundary layer formed by the model will be contained by the platform, whilst small enough in width with respect to the 0.54 m width wind tunnel test chamber so as to not interact with the boundary layer formed on the walls of the wind tunnel. Blockage correction was considered in preliminary tests with the fire truck but the correction was found to be small and thus was not considered in later tests. A pitotstatic tube was used to measure the total and static pressure readings at the same test chamber cross section area corresponding to the leading edge of the model. The Pitot static tube permitted the measurement of the total and static pressure, so it served the Fig. 1. Fire truck model setup in the wind tunnel for testing. ARTICLE IN PRESS 666 J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 dual purpose of allowing the experimenters to calculate free stream wind velocity (from the dynamic pressure) and to provide a reference total pressure for the local Cp measurements. Located on the front, sides, top, and rear of each model were 15 pressure taps. The number of pressure taps used was limited by the size of the model and by the scanner capability to measure only 16 pressure channels at one time. Therefore the taps were located such that regions of rapid pressure changes, as in the corners, were measured. Since the investigators were concerned that the limited number of pressure taps might not be sufficient to allow for reliable force measurements a companion numerical simulation was also carried out. On surfaces where fewer taps were located the pressures measured were compared with the numerical simulation results conducted by Deshpande [10]. As shown in the next section, the Cp variations on the model agree with the numerical simulation despite the limited number of pressure taps used. In addition, testing was also conducted using a 2-axis mechanical force balance to measure the drag and side force acting on the model in the x–y plane (see Fig. 12 for a definition of this plane). These forces were measured so that they could be compared to the forces predicted based upon the measured pressure distribution on the vehicle. Tygon tubing connected the pressure taps to a multi-port pneumatic pressure scanner model 9010 manufactured by Pressure Systems Inc with piezoresistive pressure sensors capable of measuring with an accuracy of 70.15% of the full scale. The scanner takes 20 measurements per channel per second and gives an average. For varying velocities up to 20 m/s, surface pressures at the 15 tap locations were recorded. The pressure data was recorded every 3 s for a total time of 8 min at each velocity, and then averaged. Another parameter tested was pressure distribution as a function of incoming wind angle. This was done by repeating the tests for all velocities at yaw angles of 0
, 745
, 790
, 7135
, and 7180
(for the suburban and ambulance additional angles of 725
and 765
degrees were tested because the forces and moments for a moving vehicle, considered in the analysis later, varied significantly between 0
and 90
yaw wind angle). Positive angles are in the clockwise direction and negative are counter clockwise. 0
is defined as the condition where the front of the vehicle faces the oncoming wind. Fig. 2 displays the locations of the taps and the numbering convention for the fire truck model. Although only 15 pressure taps were installed on the model, symmetry and testing at negative angles gave 30 pressure readings for angles between 0
and 180
. At this point it should be noted that no attempt was made to simulate atmospheric turbulence or shear. Whilst this is far from ideal, it is relatively commonplace for cross wind tests on vehicles and the experience of other authors suggests that the results will not be greatly in error. 2.2. Wind tunnel results The surface pressure values measured at various locations on the vehicles were averaged over the test period and then used to calculate, with the reference total pressure, the value for the coefficient of pressure (Cp ) at each tap location. Cp is ARTICLE IN PRESS J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 667 Fig. 2. Tap location and numbering for the fire truck model. calculated as Cp ¼ pt  ps : ptotal  ps ð1Þ In this equation Cp is the coefficient of pressure, pt is defined as the tap pressure (15 of them) on the model, ps is the free stream static pressure measured with the Pitotstatic tube, ptotal is the free stream total pressure measured with the Pitot-static tube. As an example, Fig. 3 shows a typical plot of Cp vs. Reynolds number for the pressure taps on the windward side of the fire truck model during a 90
testing (i.e., the side facing the flow). Here Reynolds number is defined as Re ¼ rN VN x : mN ð2Þ In this equation rN ; VN and mN ; refer to the free stream density, velocity and viscosity respectively. The term x in this study was used as the total length of the vehicle. The matching of Reynolds number is one of the simulation requirements mentioned by Gawthorpe [1] for effective wind tunnel modeling of a vehicle. For a full size fire truck the Reynolds would be above 4.5  106. In the wind tunnel testing values of only 4  105 were realized. For this reason testing was carried out on a full size vehicle to investigate the effects of this Reynolds number discrepancy. The measured Cp values have a standard deviation 710% after consideration of a 75% standard deviation observed as the variation in mean total pressure. This large standard deviation is attributed to the instrument noise when measuring very low ARTICLE IN PRESS 668 J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 Fig. 3. Typical Cp vs. Reynolds number distributions at windward tap locations for 90
yaw angle. pressures. A look at the mean Cp for all velocities at a single location exhibits a total standard deviation of only 75%. The pressure distributions resulting from the Cp values were used to calculate the forces acting on each model. The Cp distribution also gives an idea of potential boundary layer flow separation where the pressure gradients are large. For example close to the edges the Cp variations are very large. If the local Cp varies more than 40% for the range of tested velocity then separation is likely to occur according to aerodynamic theory. Such regions were only observed when the pressure tap was placed very close to the edge—less than 0.01 m for the model. The tap on the edge of the windshield of the fire truck (both the full size vehicle and the model) was one example. Another example was a tap on the ambulance, which acted similarly. The complete results for each pressure tap for the various angles tested are in Pinelli et al. [9]. Symmetry allows for the combination of the positive and negative orientations to obtain a picture of the pressure distributions about the model at orientations between 0
and 180
. Fig. 4 shows the results for the fire truck for a case of a 45
wind direction. Similar results were obtained for the ambulance and the SUV. As mentioned previously, the results were compared to a numerical simulation of the full scale fire truck, carried out by Deshpande [10]. It was done with the computational fluids dynamic software Fluent 6.0 [14]. Fig. 5 shows a sample comparison between the wind tunnel test results and the computational simulation results, for the taps on the side of the vehicle at zero yaw angle. The measured Cp values agree with the simulation results very well in the regions of strong pressure ARTICLE IN PRESS J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 669 Fig. 4. Fire Truck 45
coefficient of pressure test results. 0.1 0 Fluent R.C. Fluent S.C. Cp Wind Tunnel -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.9 Wind Direction 7 6 5 14 4 3 Tap Numbering and Location Fig. 5. Coefficient of pressure comparison between the fluent simulation and the wind tunnel model testing. gradients (taps 7 & 6). The figure also depicts how the numerical simulation can give varied results. The figure shows the simulation results for the same vehicle with square corners (Fluent S.C.) and with rounded edges (Fluent R.C.). When square cornering is used as opposed to round cornering the Cp values show a greater ARTICLE IN PRESS 670 J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 difference with respect to the experiments. Similar agreement was noticed on the front face. 3. Full-scale fire truck field test 3.1. Experimental setup In order to validate the wind tunnel tests, a zero degree test was conducted on a full size fire truck with the help of the Satellite Beach fire department. An Emergency One Cyclone Pumper was instrumented with a total of 15 pressure taps, manufactured for the purpose of the test. The 15 taps were connected via Tygon tubes to the same multi-port pneumatic pressure scanner that was used for the wind tunnel testing. The vehicle used has an overall length of approximately 9.1 m, a height of 2.7 m, and a cab width of 2.5 m. All the pressure taps were taped to the surface of the vehicle at different locations on its front, sides, top, and back. Fig. 6 shows the locations of the pressure taps on the truck. In addition, an anemometer was mounted on a mast above the cabin, together with a Pitot-static tube that was also connected to the pressure scanner. The Pitot-static tube provided a measure of both the static and total pressure of the free flow of air around the truck. The pressure scanner was connected to a data Fig. 6. Location of the pressure taps on the full size fire truck. ARTICLE IN PRESS J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 671 Fig. 7. Instrumented full size fire truck for the field test. acquisition program loaded on a laptop. The anemometer was not connected to the laptop, and readings were stored separately and later uploaded to the computer. Fig. 7 shows the instrumented fire truck. 3.2. Calculation of wind speeds The truck was driven on a highway for intervals of half an hour at speeds of 20 m/s (45 mph), 25 m/s (55 mph), and 30 m/s (67 mph), on a day of relatively calm to moderate winds. The return distance was covered at an average speed of 29 m/s (65 mph). The speedometer was used to obtain an idea of how fast the vehicle was traveling, but for analysis the wind speeds were computed from the Pitot-static tube readings while the anemometer readings were used for comparison. The data acquisition system collected and stored the data from the pressure scanner at intervals of 5 s. Similarly, the corresponding wind speed was measured with the anemometer every minute. A small crosswind on the day of testing explains why the two measuring devices did not predict exactly the same vehicular speed (regardless, the yaw angle was less than 75
-essentially zero). The average measured wind speeds from the anemometer were 24 m/s (51 mph), 29.5 m/s (66 mph), and 34.9 m/s (78 mph). The average computed wind speeds from the pitot static tube were 24 m/s (51 mph), 28.2 m/s (63 mph), and 33.1 m/s (74 mph). The anemometer displayed a standard deviation of 1 m/s (2 mph) from the mean wind speed. The standard deviation based upon the pitot static tube readings was 1.5 m/s (3 mph). 3.3. Calculation of pressure coefficient, Cp, values The dynamic pressure readings of all the taps over the surface of the truck were acquired every 5 s from the pneumatic pressure scanner, and significant scatter was ARTICLE IN PRESS 672 J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 Fig. 8. Time history of coefficient of pressure on the front face of the fire truck. seen in the plots of pressure with respect to time, mainly at low speeds and pressures. The scatter was attributed to the instrument unsteadiness when reading small differential pressures. Indeed, some gusts due to the natural winds and passing vehicles on the expressways were also observed. In order to correct for the scatter, the pressure at each location was averaged for each velocity to compute the coefficient of pressure at each location. The standard deviation was observed to be quite high for some locations, incidentally where the pressure readings were very small, as much as 30%, while at others it was observed to be as low as 5%. This difference is attributed to the reasons mentioned above. The surfaces, which had the largest standard deviation in pressure, were those on the sides of the fire truck. From these pressure values, the pressure coefficients Cp were computed at different tap locations using Eq. (1). For example, Fig. 8 shows the time history plots of the Cp values for the different tap locations located at the front of the truck. The Cp values appeared to be constant within the degree of precision for the different velocities tested and these results are summarized in Fig. 9. 3.4. Comparison of results from the field and wind tunnel tests Figs. 10 and 11 show the coefficients of pressure measured at different locations on the wind tunnel model and full size fire trucks. The figures are drawn to scale, and the indicators show the exact location where each coefficient of pressure value was measured. The wind tunnel model and the full size model are not identical. The ARTICLE IN PRESS J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 673 Fig. 9. 0
coefficient of pressure full scale fire truck field test results. Fig. 10. Front, rear, and top of vehicle Cp comparisons between field test and wind tunnel test. model tested in the wind tunnel resembled a Typhoon Pumper more closely than a Cyclone Pumper. The significant difference between the two models is that the Cyclone has a gap between the cab and the rear, whereas the Typhoon does not. In addition the locations of the pressure taps for the wind tunnel model and for the full size fire truck were not exactly the same, so only an approximate comparison can be done. Nevertheless, comparing the results we see a fair agreement for Cp on the front, sides and rear of the fire trucks between the scale model and the full size vehicle. This suggests that the difference in Reynolds number between the wind tunnel and the full scale tests has no significant effect, at least for the yaw angles ARTICLE IN PRESS 674 J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 Fig. 11. Side of vehicle Cp comparisons between field test and wind tunnel test. 1 tested. The locations where disagreement reaches more than a few 100 ths are on the leading or trailing edges. On the front face we can compare 0.8 from the field test to the average of 0.97 and 0.74 on the front of the wind tunnel model. On the top of the cab there is a slightly larger disagreement between the field test and the model. The field test has a much lower value of Cp than the model. The field test had a value of 0.06 in comparison to 0.28 or 0.2 for the wind tunnel model. The explanation is due in part to the differences in surface geometry on the top of the cab. The full size fire truck had two very large protrusions in front of the location where pressure was measured, while the model had virtually no protrusions windward of the pressure taps. 4. Forces on the vehicle 4.1. Determination of active forces acting on the vehicle When a vehicle (assumed symmetric about it’s centerline) is traveling down the expressway with no crosswind, normal forces acting on the vehicle will only include a pitching moment, a lift force, and a drag force. In a cross wind the vehicle will experience several additional actions. These actions can include a side force, a yawing moment, and a rolling moment. All the forces acting on a vehicle with a non-zero yaw angle (between vehicle axis and relative wind direction) are shown in Fig. 12. From the wind tunnel tests, coefficients of pressure at various locations on the truck and for various relative wind directions are known. This information is used to compute the forces acting on the vehicle as a function of wind speed. First, the ARTICLE IN PRESS J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 675 Fig. 12. Definition of forces and moments acting on a vehicle. surface of the vehicle is divided into sections or panels where Cp is assumed constant. Local forces on each vehicle panel are computed using the local panel area, Apanel ; multiplied by the local dynamic pressure estimated at each tap location:   1 2  r  Cp  Apanel ; Fpanel ¼ VN ð3Þ 2 where r is the density of air (assumed to be 1.225 kg/m3), VN is the free stream wind speed, and Cp is the coefficient of pressure measured at that tap. Summing the forces on all the panels over the fire truck for a certain wind angle and velocity gives the resulting force acting on the model. Both computer simulation and wind tunnel testing provided information on the location and extent of regions of flow separation. This information helped define the areas used for each panel. For the panels on the sides and top of the vehicles the edge of the area was chosen as half the distance between two neighboring pressure taps. For the front and rear of the vehicles the panels where chosen slightly differently, based largely on the observed assumption that steep pressure gradients occur only near the edges. Thus larger panel areas were chosen for pressure taps near the center of the face, while smaller areas were chosen for pressure taps on the edges. Knowing the forces acting on the vehicle allows for the calculation of dimensionless force coefficients for the vehicle. In this manner the side force coefficient is calculated through the following equation. Cs ¼ Side Force : 1 2 2 rAVN ð4Þ ARTICLE IN PRESS 676 J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 In this equation density (r) and velocity (VN ) are as previously defined. The symbol A represents a reference area. For our purpose we chose the frontal area of the vehicle at zero yaw angle as this reference area to be consistent with the aerodynamics practice. The same reference area is also used for consistency and convenience to calculate the drag force coefficient and the lift force coefficient by choosing the respective forces As mentioned earlier a strain gage force balance was used to measure drag and side force on the models at 0
, 45
, and 90
for different velocities. These forces were compared to the forces predicted from the pressure distribution through Eqs. (3) and (4) above. The side force coefficient comparison is shown in Fig. 13 for different Reynolds numbers at a yaw angle of 45
. A comparison is also shown for the drag force coefficient in Fig. 14 for different Reynolds numbers at a yaw angle of 0
. The complete results for the comparison between the force balance and pressure distribution are in Pinelli et al. [9]. For all the models at different yaw angles the comparisons were similar to those shown. The force balance gives the total side and drag force on the model, while the pressure measurements are made only at 30 discrete points. Local flow anomalies can alter the pressures at unmeasured locations. The panels are assumed to be flat, and the effects of small inclination of the panel to reference co-ordinates are neglected. Also, the pressure measurements Fig. 13. Fire truck force balance and pressure distribution side force coefficient comparison at 45
. ARTICLE IN PRESS J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 677 Fig. 14. Fire truck force balance and pressure distribution drag force coefficient comparison at 0
. do not include the viscous effects. If skin friction were considered for a turbulent boundary layer the pressure distribution results would be slightly higher. Fig. 15 shows the variation of the drag coefficient with respect to yaw angle for the highest Reynolds number tested. Fig. 16 shows the variation of side force coefficient with respect to yaw angle also for the highest Reynolds number tested. Also included in Fig. 16 is the side force coefficient predicted by Fluent simulations [10]. It is encouraging to note that the comparisons are fairly close for the tested yaw angles, and best for 45
(the angle of most interest in this study). These plots suggest that the 30 pressure tap measurements can be used to find a reasonable force estimate. 4.2. Resistive forces For each wind speed, based on simple equilibrium equations, the yawing moment and side force are decomposed into the equivalent side forces acting on each axle. In order for a course deviation to occur the active forces on either axle must be greater than the resistive force available on that axle. The resistive force on each axle depends upon the coefficient of static friction, m; between the tires and the road, and the weight acting on each axle. For this study, analyzes were conducted for a full fire truck, and a fire truck with an empty water ARTICLE IN PRESS 678 J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 Fig. 15. Variation of drag coefficient vs. yaw angle (Re=4.2  105). tank. The water tank is assumed to be symmetric above the rear axle. So for the empty analysis the weight acting upon the rear axle was reduced. Due to the pitching moment and lift, the weight per axle will differ with wind speed and yaw angle. The result is an effective weight per axle, E; which is computed based on simple equilibrium equation relating the effective weight per axle to the lift, pitching moment, and the vehicle weight per axle (supplied by the manufacturer). In terms of course deviation analysis the resistive force per axle, Rforce ; can then be defined as, Rforce ¼ Em: ð5Þ For galvanized rubber on different road surfaces the coefficient of static friction (m) is a value that is not very well documented because it depends upon numerous variables [11]. For automobiles, m will usually be somewhere between 0.1 and 0.9. For example, the National Highway Traffic Safety Administrations ‘‘Rating System for Rollover Resistance’’ [12] suggests that: ‘‘For a good dry paved surface, m may be in the neighborhood of 0.9. But for wet, icy roads, m is considerably less.’’ There are several reasons for this broad uncertainty. The first is that this value depends upon the chemical composition of the tire and the road. Another factor is the tire pattern and the tire pressure. And finally, it depends upon the conditions of the road. For a ARTICLE IN PRESS J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 679 Fig. 16. Variation of side force coefficient vs. yaw angle (Re=4.2  105). hot, dry day m will be highest. For a cold wet day m will be considerably lower. For the wet road analysis a coefficient of static friction of 0.2 was used. For the dry road analysis a coefficient of static friction of 0.8 was used. 5. Critical wind speeds 5.1. Types of vehicle accidents The studies conducted here investigated two possibilities of vehicle accidents as a result of strong winds (see [13]). The first possibility is a course deviation, which can lead to accidents between vehicles, or the vehicle being blown off the road. This event is very dependent upon the abilities of a driver to handle a vehicle in extreme driving conditions (abilities of multiple drivers if an accident involves more than one vehicle). Course deviation is also very dependent upon road conditions (i.e. how slippery the road is) and this is an uncontrollable variable. The second possibility is overturning of a vehicle due to strong winds. It will be shown that overturning due to high wind speeds is unlikely to occur before course deviation. Overturning which can not be corrected for by the driver or good driving conditions, may still occur, however, if a vehicle is not permitted to slide sideways ARTICLE IN PRESS 680 J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 (parked against a curb, or in a day when the roads are hot and dry- conditions leading to high coefficient of static friction). A possible way overturning might also result is through overcompensation by the driver to a gust of wind. It is also possible that in a turn, a strong crosswind may induce a rollover. Crosswinds in a turn were not studied for this report. If such a study was conducted as a minimum information such as road curvature, vehicle speed, and height of the vehicles center of gravity would be needed to predict the strength of a crosswind that would lead to a rollover while turning. For this study the critical wind speed for course deviation was defined as the speed at which the active forces on either axle exceed the resistive forces. The critical wind speed for overturning was defined as the speed at which the rolling moment due to the side force and the lift exceeds the restoring moment provided by the mass of the vehicle. 5.2. Fire truck critical wind speeds Table 1 shows the calculated wind speeds at which a still or parked fire truck will be displaced. These speeds do not change between an empty and a full fire truck. A course deviation occurs as a result of the front of the fire truck slipping sideways. The weight on the front axle does not change and the pressure distribution around the fire truck does not change when the water tank is full. Because of this the front of the fire truck will slide sideways at the same wind speed regardless of whether the tank above the rear axle is full or empty. When a vehicle is traveling down a highway, the vehicle speed adds up to the wind speed to determine the speed at which a course deviation will occur. Fig. 17 shows the resultant wind speed that for wind yaw angles between 0
and 90
will be higher than just the wind speed relative to the road. Taking vehicle speed into consideration, Table 2 shows vehicle speed and necessary cross wind speed to create a 45
wind resulting in a course deviation for the fire truck. This analysis was done for wet roads. For a range of vehicle speeds between 16 and 25 m/s (35–55 mph), the wind speed required to cause a course deviation for an empty Typhoon fire truck model decreases from 31 to 28 m/s (70– 64 mph). The lowest critical crosswind to cause overturning of an empty fire truck was found to be 72 m/s (160 mph) at a yaw angle of 59
, when the truck travels at 25 m/s. Table 1 Wind speed V to cause a static course deviation-Typhoon Wind yaw angle (degrees) 45 90 135 V m/s (mph)—Full truck V m/s (mph)—Empty truck Dry roads Wet roads Dry roads Wet roads 71 (159) 65 (145) 77 (173) 40 (90) 38 (85) 45 (100) 71 (159) 65 (145) 76 (170) 40 (90) 38 (85) 45 (100) ARTICLE IN PRESS J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 681 Fig. 17. The difference between relative wind and resultant wind. Table 2 Typhoon vehicle speed and necessary crosswind speed to cause a resultant wind speed of 40 m/s at 45
Vehicle speed Crosswind to cause a course deviation Yaw angle (deg.) 16 m/s (35 mph) 20 m/s (45 mph) 25 m/s (55 mph) 31 m/s (70 mph) 30 m/s (66 mph) 28 m/s (64 mph) 72 81 83 Therefore, overturning due to high wind speeds is unlikely to occur before course deviation occurs, but it is possible that the truck could be blown off course and hit an obstacle in such a way as to cause overturning. To summarize for the fire truck, the critical wind speed falls in the range of 28– 31 m/s. Yaw angles for this event range between 45
and 90
. But this is not to say wind angles outside this range are not dangerous. Wind yaw angles between 90
and 180
will contribute largely to a tailwind and not so much a crosswind and will make stopping difficult. 5.3. Ambulance critical wind speeds For the ambulance model, for a range of vehicle speeds between 12 and 25 m/s (27 and 55 mph), the wind speed required to cause a course deviation decreases from 21 to 15 m/s (48–34 mph). The ambulance is at risk of overturning for wind speeds above 40 m/s (90 mph). The possibility can be delayed to about 100 mph if the vehicle is driven slowly. As with the fire truck it is possible that overturning can occur at lower wind speeds during a violent course deviation. To summarize for the ambulance, the critical wind speeds falls in the range of 15– 20 m/s. The corresponding yaw angles are between 68
and 90
. One explanation for why this range of wind speeds is so much lower than for the fire truck or the SUV is because it has a lower density—or mass per unit volume. The fire truck is a large vehicle but it does not have much empty space. The SUV is not as large as the fire truck, but based upon the assumption that it is full of equipment it does not have ARTICLE IN PRESS 682 J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 much empty space either. The ambulance on the other hand because of the large patient module (which is mostly empty space) the vehicle is very susceptible to wind effects. 5.4. SUV critical wind speeds From the analysis for the SUV model, for a range of vehicle speeds between 12 and 25 m/s (27 and 55 mph), the wind speed required to cause a course deviation is around 34 m/s (77 mph), for yaw angles between 83
and 100
. The SUV is at risk of overturning for wind speeds above 61 m/s (138 mph). As with other vehicles, overturning can result at lower wind speeds during a violent course deviation. One should also note that for all wind angles except 45
the vehicle displays instability on its rear axle. The 45
case is unstable on the front axle with very little side-to-side resistance left on the rear axle because of the unique shape of this vehicle. At low yaw angles such as 25
there is positive pressure on the windshield and the hood just in front of the windshield. This contributes to a pitching moment downwards on the front axle. This dramatically increases the wind speed that will cause a course deviation because now the rear axle is the one that deviates first. This is very different than what was observed for the fire truck and the ambulance. To summarize for the SUV, the critical wind speeds occur when wind speed is close to 34 m/s (77 mph). As with the other vehicles lower wind speeds will likely lead to dangerous driving depending upon various conditions. 6. Conclusions As mentioned earlier the purpose of this study was to provide the sponsoring agency with a wind speed limit for operation of their fire and rescue vehicles. To this end the results and methods are explained in the previous sections. There is of course a margin of error inherent in any experimental study. Several variables not considered in this study are that turbulence could not be simulated in the wind tunnel and the forces beneath a vehicle that would contribute to lift and a pitching moment were not measured because the moving ground plane could not be simulated. Another is that conditions like centrifugal forces in a curve, or operation on open causeways were not taken into account. There is also some error involved in the method and variables considered in this study. For example the basis for the results is that the flat surface areas of the vehicle can be divided into 30 panels and that the pressure measured at one location is homogeneous on the entire panel. Another issue is the large standard deviation observed in the measured coefficient of pressure. Admittedly there is error involved in these simplifications. But, comparisons with force balance measurements showed that the procedure was accurate enough for the purpose of this study. The apex of uncertainty in determining safe wind speeds for a vehicle result from determination of the friction forces that act between the tire and the road. These are ARTICLE IN PRESS J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 683 the forces that will keep a vehicle from being blown off the road or into other vehicles. And to date these forces are not very well quantified due to the variables mentioned in this paper. Based upon experience it is known that for the same vehicle with the same tires the coefficient of static friction can easily change from 0.2 to 0.8 depending upon road conditions such as rain, temperature, etc. Because heavy rains usually accompany strong hurricane winds this is the assumed worst-case scenario for this study and hence the coefficient of static friction of 0.2 was used. It cannot be stressed enough that this is not an exact value. Due to these uncertainties a factor of safety was chosen as 1.25. Such a value of the factor of safety results in a definition of the lowest critical speed for emergency vehicles that agrees well with past experience in hurricanes. Consequently, a classification of wind speeds is provided in the Fig. 18–20 for each vehicle. The horizontal lines define a critical range for the wind speeds. The lower wind speed limit for the critical range for each vehicle is defined as the lowest critical wind speed for a moving vehicle marking the onset of course deviation, divided by a factor of safety of 1.25. The higher wind speed limit for the critical range for each vehicle is defined as the lowest critical wind speed for a non-moving vehicle marking the onset of sideways displacement divided by a factor of safety of 1.25. In each case, the critical range is bisected by a curve representing the variation of critical wind speed as a function of vehicle speed. The area of the critical range to the left of the curve represents combinations of wind and vehicle speeds that will make driving the vehicle hazardous. The right area of the critical range is for dangerous driving conditions beyond the onset of possible Fig. 18. Wind speed–vehicle speed recommendations for the Typhoon fire truck. ARTICLE IN PRESS 684 J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 Fig. 19. Wind speed–vehicle speed recommendations for a F-450 type I ambulance. Fig. 20. Wind speed–vehicle speed recommendations for a 4  434 ton suburban (full of equipment). course deviation, where the driver should slow down. Finally Seek Shelter indicates the wind speed for which this vehicle should not be driven, and parked indoors if possible. ARTICLE IN PRESS J.-P. Pinelli et al. / J. Wind Eng. Ind. Aerodyn. 92 (2004) 663–685 685 Acknowledgements This study was funded by Grant CFDA83.548 from the Florida Department of Community Affairs. The authors wish to acknowledge the support and advice of Fire Chief Daniel H. Rocque, Satellite Beach Fire Department, and Mr. David Halstead, Senior Management Analyst, Division Of Emergency Management, Florida Department Of Community Affairs, who acted as project managers for this study. The encouragement of Mr. Robert Lay, Emergency Manager, Brevard County is also gratefully appreciated. Finally, many thanks to the whole team of the Satellite Beach Fire Department, and in particular to Operations Commander Donald Hughes, and Captain James Keating. References [1] R.G. Gawthorpe, Wind effects on ground transportation, J.Wind Eng. Ind. Aerodyn. 52 (1994) 73–92. [2] C.J. Baker, The problems of road vehicles in cross winds, J. Highways Transport. 38 (5) (1991) 6–9. [3] T.W. Schmidlin, P.S. Kind, B.O. Hammer, Y. Ono, Behavior of Vehicles during tornado Winds, J Saf. Res. 29 (3) (1998) 181–186. [4] C.J. Baker, S.A. Coleman, High sided road vehicles in cross winds, J. Wind Eng. Ind. Aerodyn. 36 (1990) 1383–1392. [5] C.J. Baker, N.D. Humphreys, Forces on vehicles in cross winds from moving model tests, J. Wind Eng. Ind. Aerodyn. 41–44 (1992) 2673–2684. [6] C.J. Baker, The quantification of accident risk for road vehicles in cross winds, J. Wind Eng. Ind. Aerodyn. 52 (1994) 93–107. [7] D.W. Hurst, J.W. Allan, K. Burgin, Pressure measurements on a full-scale tractor–trailer combination and comparison with data from wind tunnel model tests. Int. J. Vehicle Des. (Special Publication SP3) UK (1983) 471–479. [8] A. Ryan, R.G. Dominy, Aerodynamic forces induced on a passenger vehicle in response to a transient cross-wind gust at a relative incidence of 30 degrees, Developments in Vehicle Aerodynamics, SAE Special Publication, Vol. 1318, February 1998, SAE, Warrendale, PA, USA, pp. 181–189. [9] J.P. Pinelli, C. Subramanian, M. Plamondon, V. Chakravarthi, Wind Effects on Emergency Vehicles—Final Report; Wind & Hurricane Impact Research Laboratory, Florida Institute of Technology, Melbourne, FL, 2003. [10] V. Deshpande, A numerical study of flow around fire and rescue vehicles, Masters of Science Thesis, Florida Institute of Technology, December 2003. [11] Jones, Childers, Contemporary College Physics, (Chapter 4.8) Force and Motion, McGraw-Hill College, New York, 2001. [12] National Highway traffic safety administrations ‘Rating system for rollover resistance’ Special Report 265, National Academy Press, Washington, DC, 2002. [13] R.H. Barnard, Road Vehicle Aerodynamic Design, 2nd Edition, MechAero Publishing, Hertfordshire, England, 2001. [14] Fluent, Fluent 6.0 UDF Manual, Fluent Inc, November 29, 2001.