Final Memo

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Matthew Yu, Nick Schneider, Vijay Ramanujan

Final Memo

Design Concept

After constructing our three prototypes, the savonius turbine design was chosen due to its

simple and low cost construction, multi-directional input, high starting torque, low operating

speed and low maintenance compared to the darrieus turbine and five bladed lenz turbine. We

felt that the use of an airfoil, either by itself in the darrieus design or in conjunction with drag

differential as in the case of the lens turbine, would be ineffective under the short duration test

conditions of our lab and would simply add unnecessary weight to our model. Since the goal was

to produce the lightest turbine while still reaching a power output of 4W, the savonius design

seemed like the best option, but we could not be sure until data was gathered on all three

prototypes.

Data gathered in the wind tunnel allowed for us to quickly decide against the use of the

darrieus model, since even our small scale model didn’t provide the left necessary to get the

turbine spinning under low wind speed and we felt that it would only get significantly harder to

construct working airfoils at full scale, especially given the flimsy plastic we would soon be

working with to create our models. This left us to decide between the savonius and the lens

turbine, but it will be explained later through our mathematical model that both turbines had the

capability to produce 4 watts of power under the constraints of our test day conditions, so it only

made sense to choose the lighter savonius design.


Aerodynamic analysis & Test Results

A table containing our wind tunnel results is shown in figure 2. With that data, we

graphed the torque vs turbine speed and the coefficient of power vs tip speed ratio (TSR). From

the coefficient of power vs TSR, our turbine demonstrated a maximum Cp of 0.029558 when the

TSR is 0.2943 (figure 4). It is also clear that since the R-squared value is relatively low, our data

did not provide the best fit.

To perform a full scale aerodynamic prediction of the initial Savonius turbine, we applied

dynamic similitude, since the dimensionless parameters, TSR and Coefficient of Power in this

case, stay constant for both the test and the real application; these parameters allowed me to

apply scaling laws to determine the velocity we needed to run for the full scale turbine with

respect to the torque imposed on the shaft and the tip speed ratio needed for max power at the

scaled level. After measuring the maximum coefficient of power to be 0.029 and the tip speed

ratio to be 0.29 with respect to that coefficient of power, these results were applied to the

coefficient of power equation to scale up our model to the appropriate area. With the tip speed

ratio, we determined the angular velocity of the turbine and its resulting torque on the shaft from

figure 3. This data was essential to the DC motor analysis and gearbox sizing.

Given that we want the full scaled turbine to generate 4 watts of power, an average

running speed of 12 miles/hour, and the density of air, the full scale turbine should have an area

of 20.328 ft^2. Since the height of the ceiling limits or height to a maximum length of 4 ft, that

forced our design to have a length of 5.082 ft.


DC Motor Analysis and Gearbox Sizing

From our earlier analysis, we obtained the torque of the turbine Qt , and the angular velocity of

the turbine, ω t . We also knew values for K , K m , i0 , Rm as related to the motor.

We now have:

G = ω m /ω t = Qt /Qm

Which implies ω m = Gω t and Qm = Qt /G

Further, we have:

Qm = K m (i + i0 )

K ω m = i(Re + Rm )

Substituting:

Qt /G = K m (i + i0 )

K Gω t = i(Re + Rm )

Our only two unknowns in these simultaneous equations are G, i and Re . Thus, by fixing one of

these variables, we can find the other two. We choose to fix G, to get that:

i = Qt / (K m G) − i0

Re = (KGω t )/i − Rm
And using these, we can recognize that the power supplied to the external resistor, which we

aim to maximize, is given by i2 Re . Thus, we iterate through different acceptable values of G to

find the value that maximizes power, and choose the Re corresponding to it. In this case, we

used a gear ratio of 7:1, and an external resistance of 21.57 Ohms.

Conclusions

Our two blade savonius design with single stage gears was one of the simpler and more

robust models and produced roughly 2.0W of power at an external resistance of 20 ohms on final

test day. At roughly 3.90 kg, it was also one of the lightest designs as well, so had we been able

to achieve our desired goal of 4W of power output, our design would have been a contender to

win. Heading into test day, we knew that it would certainly spin due to the simple physics of

drag differential that savonius models are based on, so our big question was whether or not we

had given ourselves enough room for error in our value for desired power output to sufficiently

make up for the inevitable power loss within our system . As in any mechanical system, there are

many different ways to lose power to friction, hence the reason for choosing a desired power

output of 5W (rather than 4W). Some examples of frictional losses would be due to the gear to

gear interface (since mdf is certainly not a frictionless material) and the turbine itself not fitting

tight enough to the shaft allowing for minor binding along the length of the shaft as the turbine

spun. The issue regarding frictional losses to the gears was addressed to the best of our ability, as

we hand sanded each and every one of the 96 gear teeth between the two gears used to ensure

that minimal power was lost in this manner.


We believe one of the biggest factors in power loss was due to the flimsy plastic that we

used as a fabric for each blade. As you can see in the cad models below, we used only 7 dowels

on each blade to support the plastic fabric. The problem was that as the wind pushed on the

blade, the plastic would deform and pockets of air were created along the surface of the

supposedly low drag blade between the dowels. The aerodynamic profile of our blades

completely changed as surface defects began to form and our drag differential suffered. To fix

such an issue, we would need to have either used a different material along our blades, or we

would have needed to do a better job pulling the plastic sheets tight along the blades. The only

viable option in our case would have been to use a different material altogether, since the plastic

was wrapped around the rather flimsy dowels, so the tighter the plastic pulled along our blade

surface, the more the dowels would buckle, and the the looser the plastic fabric would become.

With that being said, it was observed that our turbine actually spun too quickly on test

day. Our turbine was massive in terms of size, and should have relied on generating high torque

at low spin speeds. After about halfway down the hallway, our turbine seemed to still be

speeding up, which leads us to believe that it never really reached its maximum torque potential

for 20 ohms of resistance that we specified at the beginning of the run. Intuitively, our solution

was to drastically decrease the amount of external resistance, probably down to 1 ohm, which

would hopefully allow us to be more efficient in power production. Our math calculations told us

that ideally we should run at 20 ohms, so we would need to go back through and determine

whether such an observation makes sense according to our data.

Appendix
Figure 1a, 1b, 1c, 1d. Isotropic, Front, Top view and Drawing of Turbine

Turbine Resistor

Dynamic Speed external

Pressure (H) (RPM) (ohms) voltage e (volts) Q (Nm) Lambda Cp

0.080 51.000 10.000 0.750 0.024 0.132 0.021

0.080 105.000 30.000 1.930 0.025 0.272 0.044

0.080 134.000 50.000 2.560 0.023 0.347 0.053

0.080 136.000 Infinite 3.200 0.017 0.352 0.039


0.080 165.000 100.000 3.180 0.021 0.427 0.059

Figure 2. Wind Tunnel Results

Figure 3. Torque against Turbine Speed


Figure 4. Coefficient of Power against TSR

You might also like