On Drag Coefficients of Tractor Trailer Trucks Kevin Ihlein Kurt Toro Ryan B. White May 2, 2007 ME 241 Professor Roger Gans Abstract: The objective of this project was to reduce the drag force exerted on tractor trailers by modifying the rear of the carrier trailer. Three different design modifications were tested in order to observe and calculate the total drag force reduction. We concluded that scaled model testing can be inaccurate if Reynolds and Mach numbers are not adequate. Although this may be true, testing showed that a trapezoidal fin design on a blunt body model proved to reduce the drag coefficient by 0.168 %. The calculated drag coefficient was comparable to that of a real life model with Reynolds number similarities. 1 Introduction: A typical semi-truck driving on the highway will expend about 65% of its total energy trying to overcome the drag forces that oppose the truck’s motion1. The goal of this project was to modify the trailer of existing semi-trucks in order to refine truck aerodynamics and reduce the drag forces exerted on the trucks. Drag force, due to the combined effects of wall shear stress and pressure forces, is defined by: Fd = Cd 1 ρV2A 2 (1) Here, Cd is the drag coefficient, ρ is the density of air, V is the velocity, and A is the cross-sectional area2. One source says that for every 2% reduction in drag, one can expect a 1% reduction in fuel consumption3. A reduction in drag results in an increase in fuel economy, which would both save money and conserve natural resources. Figure 1 shows several key target areas on a semi-truck that contribute to the overall drag force, but we focused mainly on the rear of the trailer. Approximately 20-25% of the drag force on a semitruck comes from the turbulent air flows from the rear of the box trailer3. Reducing the wake caused by the turbulent air flow at the rear of the trailer will help reduce the overall drag force on the truck. Figure 1 shows the target areas of the aerodynamic pressure drag. Area of interest Figure 1: Target Areas of Aerodynamic Drag4 We simulated trailer drag forces by experimentally testing a 1:43 scale model truck inside a wind tunnel at the University of Rochester wind tunnel facility in Gavett Hall. The predicted drag force for our model represented as a blunt body at a wind speed of 24.6 m/s (55 mph) was 2.02 N. The idea behind the experiment was to try to 2 reduce the contributed drag from the trailer by attaching three different trailer design modifications to the rear, and evaluating the drag effects. Our hypothesis was that by adding an attachment to the trailer, the overall drag force felt by the truck would be reduced. The three attachments that we tested were a bubble shape, a trapezoidal fin shape, and a triangular delta shape. We also decided to test the drag effects from the gap between the tractor and trailer. Each attachment to the rear was tested with and without the gap closed off. Also, a blunt body model was tested with each of the designs and used for comparison. Nomenclature: A Cd Fd h Re V w ν ρ wf = = = = = = = = = = Tractor Trailer Frontal Area = w*h (m2) Drag Coefficient = Fd / (½ ρAV2) Drag (N) Trailer height = 0.0889 m (3.5 in) Reynolds Number = V*w/ν Air Velocity (m/s) Trailer width = .0635 m (2.5 in) Kinematic viscosity = 1.526*10-5 m2/s Air density = 1.184 kg/m3 Width of Full Scale Trailer = 2.44 m (8ft) Experimental Set Up & Procedure: We began the experiment by purchasing two 1:43 scale model Semi-trucks from online toy store5; one was a carrier box truck and the other was a fuel tanker truck. They came as a pair, but we only needed the box trailer to perform our proposed project. After we received the trucks in the mail, we were able to begin construction on the platform that would go in the wind tunnel. Next, we constructed the raised platform that would position the model in the center of the air stream as opposed to on the floor where the boundary layer effects would influence the air flow and the resulting drag forces. The platform was built from ½ x 6 inch pine lumber that we cut to the desired length. The boards that held the platform up were 0.14 m (5.5 in) tall and 0.14 m (5.5 in) wide. The board on which we placed the truck was 0.51 m (20 in) long; just 0.05 m (2inches) shorter than the wind tunnel floorboard (in the direction of airflow). A 0.038 m (1.5 in) hole was cut at a distance of 0.305 m (12 in) from the rear edge of the board so the pole connecting the model to the load cell could be attached. This same size hole was also cut in the 3 floorboard to allow passage of the rod. Then we sealed the open sides of the platform with 0.9525 cm (.375 in) plexi-glass to prevent any turbulent wind effects on the transducer rod. We inserted five screws, two on the sides and one on the top, of the plexi-glass so that it would tightly seal the area underneath the platform. With the help of Ken Adams, a machinist at the University of Rochester machine shop, we designed an aluminum bracket that would accept a threaded rod which connected it to the load cell. Figure 2 shows the schematic design drawing for the piece. This piece clamps around the truck’s chassis, just behind the cab, securing it firmly. After all the parts were constructed, the platform was attached by drilling two screws into each bottom end of the platform through the floorboard. Machine Screws Top Plate Truck Chassis Front View Machined Aluminum Threaded Rod Figure 2: Machined Connector Originally, the load cell was attached using four screws beneath the floorboard (outside of the wind tunnel). We were encouraged by Scott Russell to do this because the equipment is expensive and could have been damaged inside the tunnel. A long pole connected the load cell to the bottom of the truck, but after some testing we realized that the pole was bent and that its threads did not secure tightly. We needed a new method for connecting the truck to the load cell because this way was not going to work properly. 4 After speaking with Scott Russell about what was going wrong with the experiment, we modified the set up by placing the load cell inside the raised platform and securely attaching it the floorboard. We precisely cut a small hole on the bottom edge of one of the plexi-glass walls so the load cell cord could connect back to the computer module unharmed. After testing, it was confirmed that the shorter pole secured the model for firmly and extracted more accurate results than before. Figure 3 shows the entire setup. Connector Rod Aluminum bracket Load Cell Figure 3: Experimental Setup This photo shows the front cab of the truck with lots of masking tape covering it. Part of it was used to close the cab-trailer gap. Tape also covered the open windows in the doors and held them shut. We also taped over the front wheel wells so they would not affect the drag results. The custom devices that were built for the rear of the trailer consisted of paper index cards. We fitted and taped them securely, so that the high wind speeds would not cause them to detach and be destroyed in the wind tunnel. Using three index cards stacked on top of each other, we constructed our specified designs to be sturdy and withstand the high wind speeds. Three different attachments were created; a bubble shape, a trapezoidal fin shape, and a triangular delta shape. These were securely fastened to our model trailer using masking tape. We chose the bubble shape and trapezoidal fin contraption because we had seen devices like them being used on real trailers already. We thought the triangular modification piece was another possibility for reducing drag because it was a combination of 5 the bubble and fin shapes. Figures 4, 5, and 6 depict the actual index card shapes that we attached to the model trailer. Figure 4: Bubble Attachment Figure 5: Trapezoidal Fins Figure 6: Triangular Delta Design The data from our tests were gathered using the computer program LabView by National Instruments. We constructed a VI that recorded the voltage measured by the load cell in the direction parallel to the model. We programmed the LabView DAQ Assistant to gather the data from Transducer Module 3 and to display the voltage in realtime on the screen and to produce a waveform graph while running continuously. The input range was set to read from negative 5mV to positive 5mV from channels a0 and a1. The program allowed us to control the number of samples (100) and the rate (1000). The select signals box converted the numerical indicator to display the parallel axis instead of the perpendicular axis. Figure 7 shows a diagram of the VI. 6 Select Signals Input Output # Samples Control DAQ Data # Samples Rate Rate Control Numeric Indicator Waveform Graph Figure 7: LabView DAQ Assistant VI Before we could test the model in the wind tunnel, we first had to calibrate the load cell. We set the floorboard vertically on a workbench so that the pole could be loaded in the direction that we would be testing with the load cell. We started by hanging masses ranging from 50 grams up to 2 kilograms at the end of the pole. Then we recorded the measured voltages in MS Excel. Each time we inserted the model truck and floorboard setup into the wind tunnel, we started by recording a zero reading. We recorded the data by reading the measured voltage off the screen and writing it down on a notepad next to the corresponding fan speed percentage. We felt it would be easier to insert this data into MS Excel later rather than having LabView do it for us. After taking the initial reading, the wind tunnel was turned on and set at a speed of 30% or 18.8 m/s. Once the wind tunnel was running, we recorded each measurement between speeds of 30% to 90%, in intervals of 10%. It turned out that the voltage readings did not fluctuate enough to make it too complex to read from the LabView numerical indicator on the screen. Results: The results presented below detail the frontal force (drag) coefficient for the tractor-trailer combination and its components. This drag coefficient represents the force along the axis of the vehicle in the direction of travel. All coefficients were calculated based on the measurements recorded by the load cell and LabView. Without wall corrections from the wind tunnel, the computed coefficients will differ from those of the equivalent model in free-air. However, the differences in drag between the configurations should be representative of the effects of the associated geometric modifications. 7 The baseline geometry for this study represents a modern tractor trailer design with the standard aero package including a rooftop deflector. For the first tests we left the gap between the cab and trailer open, like on a real truck. For the second test, we closed this gap off to see how the gap affected the drag on the model. However, for the third test run, the truck was transformed into a blunt body object with dimensions 0.0635 m (2.5 in) wide by 0.0889 m (3.5 in) meters tall. We did this to exclusively look at the differences between the configurations attached to the rear of the trailer. Calibration Results: Table 1 displays the results of the calibration process explained above. Here we recorded the mass that was hung from the load cell and the voltage measured by LabView. Then we used Excel to translate the voltages into more manageable values for analysis by multiplying the voltage by -1000. We thought it would make the data easier to read without the extra zeros and negative sign. If all the voltages were transformed in this manner then there would be no discrepancies in the results. Then we subtracted the initial value from the subsequent values to find the scaled change in voltage called DeltaVolts. The last column converts the mass into Newtons by multiplying mass by the gravitational constant g=9.81 m/s. Then we plotted Load versus DeltaVolts seen in Figure 8, and found a linear trend line. This would later be used to translate measured voltages from the wind tunnel tests to drag forces in Newtons. Table 1: Calibration Data Mass (grams) Raw voltage (volts) Raw voltage * (-1000) DeltaVolts (voltage - v0) Load (Newtons) 0 -0.001664 1.664 0.000 0.000 20 -0.001671 1.671 0.007 0.196 50 -0.001682 1.682 0.018 0.491 100 -0.001700 1.700 0.036 0.981 200 -0.001736 1.736 0.072 1.962 400 -0.001808 1.808 0.144 3.924 500 -0.001844 1.844 0.180 4.905 1000 -0.002025 2.025 0.361 9.810 1500 -0.002206 2.206 0.542 14.715 2000 -0.002387 2.387 0.723 19.620 8 Calibration: Load vs DeltaVolts 25.0 Load (N) 20.0 15.0 10.0 y = 27.152x R2 = 1 5.0 0.0 0.00 0.20 0.40 0.60 0.80 DeltaVolts (-m V) Figure 8: Calibration Plot and Trend line Equation Using the Add Trend Line feature in MS Word, we found that the equation relating DeltaVolts to Load is y=27.152x, where y is the load in Newtons and x is the variable DeltaVolts in -mV. We knew that this equation was accurate and fits the data well because the R2 value was 1. This equation was used in all of the subsequent calculations to translate the LabView measurements into a load measurement. Configurations: As mentioned above, we tested each of the four trailer configurations: baseline, bubble, fins, and delta, three times. First we left the cab-trailer gap open, then we closed it up, and then we treated the model as a blunt body. We recorded the measured voltages from LabView, and then transformed the data into more manageable values in the same manner as mentioned in the Calibration paragraph. We used the equation y=27.152x to calculate the load, in Newtons, from the DeltaVolts column. We also converted the percent fan speed into airspeed by multiplying the percentage by the maximum velocity of the wind tunnel, which was 62.6 m/s (140mph). For example, 50% of the fan’s speed is 31.3 m/s. 9 Then, we plotted airspeed versus drag for each device attached to the trailer. This would later allow us to find a relationship between wind speed and drag. Figure 9 displays this plot for each of the four devices in the open gap configuration, while Figure 10 represents the data for the closed gap configuration, and Figure 11 for the blunt body. Drag vs Airspeed: Open Gap 4.00 3.50 3.00 Drag (N) 2.50 2.00 1.50 1.00 0.50 0.00 0.0 10.0 20.0 30.0 40.0 50.0 60.0 Airspeed (m/s) Baseline Bubble Fins Delta Figure 9: Drag vs. Airspeed: Open Gap Configuration Drag vs Airspeed: Closed Gap 4.00 3.50 3.00 Drag (N) 2.50 2.00 1.50 1.00 0.50 0.00 0.0 10.0 20.0 30.0 40.0 50.0 60.0 Airspeed (m/s) Baseline Bubble Fins Delta Figure 10: Drag vs. Airspeed: Closed Gap Configuration 10 Drag vs Airspeed: Blunt Body 10.00 9.00 8.00 Drag (N) 7.00 6.00 5.00 4.00 3.00 2.00 1.00 0.00 0.0 10.0 20.0 30.0 40.0 50.0 60.0 Airspeed (m/s) Baseline Bubble Fins Delta Figure 11: Drag vs. Airspeed: Blunt Body Configuration From these plots, we used the Add Trendline feature of MS Excel to show us the equations that related airspeed in meters per second to drag in Newtons. Table 2 shows the given trend line function and its corresponding R2 value for the open gap configuration including all data points. The closed gap configuration and blunt body results are in Table 3 and Table 4 respectively. We knew that these equations were extremely accurate because each R2 value was within .001 to 1, which would mean that the function fits the data closely. Table 2: Open Gap Functions Configuration Trend Line Function Baseline Y=0.0007x2 + 0.0181x Bubble Y=0.0011x2 - 0.0009x Fins Y=0.001x2 - 0.0032x Delta Y=0.0012x2 - 0.0057x R2 Value .9991 .9994 .9997 .9994 Table 3: Closed Gap Functions Configuration Trend Line Function Baseline Y=0.001x2 + 0.0011x Bubble Y=0.0011x2 - 0.0031x Fins Y=0.0011x2 - 0.0049x Delta Y=0.0013x2 - 0.0089x R2 Value .9999 .9999 .9990 .9991 Table 4: Blunt Body Functions Configuration Trend Line Function Baseline Y=0.0018x2 + 0.0084x Bubble Y=0.002x2 + 0.0018x Fins Y=0.0018x2 + 0.0054x Delta Y=0.0036x2 - 0.0356x R2 Value .9995 1.0 .9999 .9980 11 Reynolds Number Calculation: In order for us to determine the necessary airspeed for our model to represent a full scale truck, we first had to calculate the Reynolds number: Re = V*w/v (2) We multiplied an airspeed V of 24.6 m/s (55 mph) by the characteristic width of a full scale trailer wf, 2.44 m (8ft), then divided by the kinematic viscosity of air, 1.56x10-5 m2/s. The resulting Reynolds number was 3.84x106. Using the theory of similitude from our Fluid Dynamics course, we could calculate the necessary airspeed for the model to actually simulate a real truck at highway speeds. The Re was multiplied by the kinematic viscosity, and then divided by the characteristic length of the model truck w, .0635 m (2.5 inches). The resulting airspeed U was 944.1 m/s. Drag at Highway Speed: Next, we were able to calculate the drag that the model truck would experience at the adjusted highway speed. For our model to represent a truck at 24.6 m/s (55 mph), it would have had to be going 944.1 m/s. It would have been impossible to test the model at this airspeed because the wind tunnel had a maximum speed of 62.6 m/s (140 mph). That is why we found the above equations relating airspeed to drag for the model. Table 5 lists the drag forces that the model would have experienced at 944.1 m/s for each of the configurations. Table 5: Predicted Drag on Model at 944.1 m/s Configuration Trailer Device Drag (N) Open Gap Baseline 641.0 Bubble 979.6 Fins 888.3 Delta 1064.2 Closed Gap Baseline 892.4 Bubble 977.5 Fins 975.8 Delta 1150.3 Blunt Body Baseline 1612.3 Bubble 1787.7 Fins 1609.5 Delta 3175.2 12 We wanted to do more statistical analysis to see if there was another, perhaps improved, relationship between airspeed and drag on the model. This is why we decided to see how lines would fit to the last few points of the data, excluding the zero values and lower fan speed readings. We did this because we noticed that the graphs started to appear more linear than quadratic, and we could also tell by the voltage steps leveling out. This meant that the difference between the values of DeltaVolts of two points was the same as the difference between the point next to it. Tables 6, 7, and 8 include the trend line functions for only the higher speed readings and the calculated drag at 944.1 m/s. Table 6: Open Gap Functions Configuration Trend Line Function Baseline Bubble Fins Delta y y y y = = = = 0.085x - 1.6291 0.0967x - 2.1586 0.102x - 2.6518 0.1076x - 2.6337 .9994 .9999 .9998 .9965 Table 7: Closed Gap Functions Configuration Trend Line Function Baseline Bubble Fins Delta y y y y = = = = 0.0954x 0.0954x 0.0998x 0.1193x - 2.2536 2.1993 2.4256 3.0682 y y y y = = = = R2 Value 1 1 .9994 .9999 Table 8: Blunt Body Functions Configuration Trend Line Function Baseline Bubble Fins Delta R2 Value 0.2039x - 5.186 0.2039x - 5.0412 0.3037x - 9.4851 0.308x - 7.9012 R2 Value .9946 .9994 .9918 .9994 Calculated Drag (N) 78.6 89.1 93.6 98.9 Calculated Drag (N) 87.8 87.9 91.8 109.6 Calculated Drag (N) 187.3 187.4 277.2 282.9 Drag Coefficient Calculations: After calculating the Drag that the model would experience at speeds representing a full scale truck at highway speeds, we then calculated the drag coefficient using the equation: Cd = Fd / (½ ρAV2) (3) Here, ρ is 1.184 kg/m3, A is 0.0057 m2, and V is 944.1 m/s, and D is the calculated drag from above. Tables 9, 10 and 11 show the calculated values of the drag coefficient for each attachment and 13 configuration, including both statistical methods; quadratic and linear. The bold values came from the formulation with the highest R2 value. Table 9: Open Gap Drag Coefficients Configuration Quadratic Cd Baseline 0.2134 Bubble 0.3258 Fins 0.2954 Delta 0.3539 Linear Cd 0.0261 0.0296 0.0311 0.0329 Table 10: Closed Gap Drag Coefficients Configuration Quadratic Cd Baseline 0.2968 Bubble 0.3251 Fins 0.3245 Delta 0.3825 Linear Cd 0.0292 0.0292 0.0305 0.0364 Table 11: Blunt Body Drag Coefficients Configuration Quadratic Cd Baseline 0.5362 Bubble 0.5944 Fins 0.5353 Delta 1.056 Linear Cd 0.0623 0.0623 0.0922 0.0941 Percent Differences: Here, we calculated the differences in the drag coefficients between the baseline configuration and the respective trailer attachments from the blunt body tests based on the recorded data and not the calculated drag forces. We did this because the differences between drag using the quadratic and linear functions were too dissimilar. Also, the blunt body shape was the only one that we could predict the drag force. Table 12 displays the measured drag from the test model at 70% fan speed, or 44 m/s, and the coefficient of drag, as well as changes in the coefficients. We chose the 70% fan speed data because we thought the higher speeds were more inaccurate due to the model vibrating in the wind tunnel. Table 13 shows the percent difference in drag coefficients for the calculated drag for the model simulating highway speeds at 944.1 m/s. Table 12: Percent Change in Drag Coefficient at 44 m/s Blunt Body Drag @ 44 m/s Drag Coefficient (N) Baseline 3.801 .5875 Bubble 3.910 .6043 Fins 3.720 .5749 Delta 5.566 .8603 14 % Change +2.85 % -2.14 % +46.4 % Table 13: Percent Change in Drag Coefficient at 944 m/s Blunt Body Drag Coefficient % Change Baseline 0.5362 Bubble 0.5944 +10.9% Fins 0.5353 -0.168% Delta 1.056 +96.9% Discussion and Conclusions: From our experimental results, we concluded that the testing of our scaled model produced some inconclusive results that showed the baseline model truck having the least amount of drag force. For our first scenario with the gap open between the truck and the trailer, we calculated a drag force of 641 N, which was what the model truck would experience at 944.1 m/s. Therefore, the result of the drag coefficient was 0.2134. When the bubble piece was attached to the rear of the trailer, we observed an increase in drag force, which was 338.6 N. This opposed the results from full scale real-world trucks. For real-world trucks, this bubble attachment generated a decrease in drag, making the semi truck more fuel efficient. When we tested the truck with the trapezoidal fin attachment, the drag force was 888.3 N, increasing the drag coefficient from 0.2134 to 0.2954. This attachment had also been tested on real-world trucks but its decrease in drag is not as efficient as that of the bubble reference. For our purpose, this trapezoidal attachment also created an increase in drag. Finally, our last test for the open gap was testing the triangular attachment. This attachment resulted in the least amount of drag reduction, and also as a safety factor we would not recommend the fabrication of it. The drag force created form the attachment was 1064.2 N. The increase in drag coefficient went from 0.2134 to 0.3539. For our testing of the model truck with the gap sealed between the trailer and the truck, we concluded that the results were again erroneous. When the truck was tested without any modifications to the trailer, the drag force was 892.4 N. The drag coefficient was calculated to be 0.2968. When the bubble piece was attached at the rear of the trailer, its drag force was 977.5 N, increasing the drag coefficient from 0.2968 to 0.3251. The trapezoidal attachment resulted in a drag force of 975.8 N and increasing the drag coefficient to 0.3245. The triangular piece, as before, resulted as the most inefficient modification to the trailer. 15 Finally, when we analyzed the blunt body performance, we found that its baseline drag force was 1612.3 N at a scaled value of 944.1 m/s. It experienced a much higher drag force because the blunt body was not as streamlined as the model truck. Its drag coefficient at this speed was 0.5362. The predicted force on the blunt body was 2.02 N at 44 m/s, and the actual drag force obtained was 3.801 N, producing a drag coefficient of 0.5875. This is very comparable to the drag coefficient of a full scale truck. One source stated that the drag coefficient of a modern tractor trailer, weighing around 40 tons and traveling at 24.6 m/s, produced a drag coefficient of approximately 0.6. When we attached the bubble piece, the drag force was increased to 1787.7 N at 944.1 m/s, which also increased the drag coefficient to 0.5944. As for the trapezoidal piece, we concluded that the drag force was reduced to 1609.5 N at 944.1 m/s. This decreased the drag coefficient from 0.5362 to 0.5353, which is a 0.168% reduction. The blunt body model is nearly the best representation of a full scale truck that is obtained from our experiment. After analyzing the data, we concluded that our experimental testing of the scaled model truck provided inconclusive results. A reason for these results being questionable is that our model truck was too aerodynamic and streamlined for the wind tunnel testing area. Also, the size of the model was just not big enough to give results that can be applied to real life applications. Using a larger model would result in a lower calculated air speed using the similitude theory. If we had chosen a larger scaled model, then the truck would have been too large for the testing area and the side wall effects from the wind would have affected our results too greatly to draw any conclusive results. As for the testing of our blunt body object, we saw a reduction in drag of 2.14% at 44 m/s, when our trapezoidal fin modification was attached. This result was superb because this application is being used on some real life trucks today, and is what we expected. Future Work: With our experimentation complete, we concluded that not all scaled wind tunnel testing can be applied to real life applications. One must consider the comparison between Reynolds numbers and Mach Numbers in order to find relevant parameters to be used for experimentation. For our project, we established that the testing area of the wind tunnel is undersized for this type of simulation of a semi truck. With other projects developing to decrease drag on semi trucks, a larger 16 testing area must be developed in order to use scaled testing. The larger the model the more accurate and applicable this scaled testing could be. References: 1 http://www.llnl.gov/str/May03/McCallen.html 2 Çengel, Yunus A. 2007 Heat and Mass Transfer: A Practical Approach 3rd ed. McGraw-Hill: New York, NY. 3 http://www.engineering.sdsu.edu/~hev/aerodyn.html. Department of Mechanical Engineering, San Diego State University. 14 February 2007. 4 http://www.solusinc.com/transport.html 5 http://www.oakridgehobbies.com/toybox/toy_newray_die_cast_trucks.html. The Oakridge Corporation. 14 February 2007. 6 http://www.greencarcongress.com/2006/11/study_improveme.html. Mike Millikin. 14 November 2006. 7 http://whyfiles.org/shorties/190truck/. University of Wisconsin. 10 November 2005. 8 http://www.aerodyn.org/WindTunnel/ttunnels.html. A. Filippone. (1999-2004). 17
© Copyright 2024 Paperzz