Bill Freeman wrote:
Thermal conductivity is of minor importance for a heat exchange liquid
which is flowing in a turbulent flow. The primary heat transfer
resistances
from the engine to the air are the convective heat transfer coefficients
from the metal to the liquid and vice versa. This is what Jeff was
pointing
our in his nucleat boiling comments. The point is that you can't transfer
enough heat from the rotor housing to the coolant without local boiling.
The bubble of coolant vapor forms (absorbing a HUGE amount of
energy in phase change) and then as it is swirled away from the heat
source it immediately recondenses in the cooler fluid stream (dumping a
HUGE amount of energy by phase change). This process gives MUCH
higher heat transfer than normal turbulent flow convection can support.
Conduction would only be of first order importance if you were in
a static situation.
Bill
-----------------
Bill,
What you say is true for the area around the hot spot inside the engine
block, where nucleate boiling is needed for the high flux, however I will
maintain that high thermal conductivity is important in the overall cooling
SYSTEM, which includes the heat rejection (radiator) as well as the heat
absorption.
In standard heat transfer equations (for turbulent flow), there are three
dimensionless groups that allow one to see the effects of fluid properties,
Nusselt number = h*D/k where h is the inside heat transfer coefficient, D is
the characteristic diameter of the tube, and k is the thermal conductivity of
the fluid. HIGH Nusselt number is desired, to get a high film coefficient.
The Nusselt number has been shown to correlate to two other dimensionless
numbers, the Reynolds number, and the Prandtl number. (Again I am addressing
the convective heat transfer from the fluid to the radiator tube wall to get
the heat OUT of the system.)
Reynolds number is D*G/mu, where D is the characteristic diameter of the
tube, G is the mass flow rate, and mu is the viscosity of the fluid. We want
Large Reynolds number for high Nusselt number, therefore want low viscosity.
Prandtl number is Cp*mu/k where Cp is the heat capacity or specific heat.
These can be worked around for liquids in turbulent flow (inside the radiator
tubes), and the film coefficient can be expressed as:
h = 0.023*((G^0.8*k^0.66*Cp^.33)/(D^0.2*mu^0.47))
Ignoring the coefficients in the equation above, what this tells us is that
for good heat transfer from the liquid to the tubes in the radiator, we want:
High flow rate through the tubes (G)
High thermal conductivity (k)
High heat capacity (Cp)
Small tubes (D)
Small viscosity (mu)
If one looks at the table that I sent to Paul, the best fluid is water,
followed by a 50/50 Ethylene glycol water mix. All the other heat transfer
oils will not perform as well in the radiator.
Again, in the engine around the sparkplugs where the flux is high, you want
the nucleate boiling and the above information is secondary. But the overall
SYSTEM needs to be balanced, and getting the heat out of the fluid into the
air is equally important.
Bill Schertz
Your analysis is correct, but I believe that the range of k (fluid
conductivity) is
relatively small between the different fluids (e glycol/H2O and H2O, primarily),
so while k is significant, it is of secondary importance since (I think this
is correct,
but perhaps not) it varies over a limited range and you can cause the Nusselt,
Reynolds and Prandl numbers to vary much more by varying flow rates, tube
diameters and viscosity. The issue is one of "is it a variable with a
significant
range that I can control". I think the answer for k is "less so than many
other
variables". Of your variables, I might take some issue with the order, but the
key point is that I can make the flow rate vary at least one order of
magnitude,
I can vary the tube diameter at least one order of magnitude, while the
viscosity
and Cp and k vary over perhaps 2-4X. Also note that the exponents of the
key physical properties are less than unity. i.e. not ultra critical. I warn
young engineers to first look at the exponents of any new equation of interest.
Find any second order or third order terms and watch them like a hawk.
They'll either kill you or make you a hero, depending on which side they
come down on. Less that unity exponent terms are unlikely "killers".
So, certainly you want to choose a coolant with all the things going in your
favor as much as you can, but I submit that we STILL use water at least 75%
because it is cheap and readily available. That the Cp is high and viscosity
OK, and k reasonable are just frosting on the cake. They also mean that we
won't be going very far looking for better coolant, but we WILL design our
cooling systems to use water ("fixed" for it's two big disadvantages - it
freezes
and corrodes the hell out of stuff) cause it is cheap, available and good
enough.
The fact that it is actually VERY good is neat, too. If it were a worse
coolant,
we would probably still use it, and just have bigger radiators and water pumps.
So I agree, the system needs to be balanced, and we absolutely need a good
enough heat transfer to the air - the final step. I think that my point was
that
nucleate boiling is mandatory to get a high enough heat transfer from the very
hot parts of the engine to the fluid, since phase change heat transfer is about
3 orders of magnitude higher than conduction. Regardless of the size of the
radiator, if there were no nucleate boiling, the engine would overheat locally.
Higher k helps on radiator size, and cannot be ignored, but is not a show
stopper if it varies moderately.
I think we are really agreeing about 97% here, with a slight disagreement on
semantics and "why", both of which are not worth much more effort.
Thanks for the good commentary. Excellent points.
Bill
The AirCraft Rotary Engine NewsLetter. Powered by Linux.
http://home.earthlink.net/~rotaryeng/ http://www.linux.org