4
$\begingroup$

I need to solve the following problem for a practical application.

Water, initially at temperature $T_0$, flows with a constant speed $v$ through a PVC tube of length $l$ and diameter $D$. The temperature of the tube is constant on its whole inner surface $T_t$. What is the temperature of the water $T_1$ at the end of the tube?

UPDATE: Real-world application I had in mind. (It may sound weird, you have been warned!)

Four hours a week my girlfriend is released from her mother duties and I'm taking 100% responsibility for our three-month-old daughter. This involves feeding her breast milk stored in the fridge. Unfortunately, I face two challenges:

  1. Arrival of hunger is very unpredictable. It may happen anywhere between 2 and 3 hours. This makes it really difficult to pre-heat the right amount of milk, as milk that is warmed up but not consumed immediately is to be treated as wasted.

  2. My daughter is not the most patient when it comes to hunger. To re-iterate, every second the milk is delayed she cries as if her existence was threatened. As her parent, this is uttermost painful and stressful.

So I thought of the following solution: Buy a 10-meter PVC pipe of 3mm inner section (outer section is 5mm). It would start in the breast milk container in the fridge, pass through a giant bowl of water at 40°C, potentially spiraling several times inside it, and end 1-meter lower where I sit in the sofa. In essence, this would be a gravity-fed pipe for which initial calculations revealed a speed $v=0.39m/s$ and a discharge rate of $167mL/min$, if we crudely assume that breast milk has similar properties to water.

Not sure this was tried before, but it seems promising to me. The flow-rate is sufficiently high to feed an infant in real-time. The only remaining question was whether the milk warms up fast enough to reach infant consumption temperature. Fortunately, if we plug the numbers in the answer given by Gert, then spiraling 1 meter through the water tank would already bring the milk to reasonable consumption temperature.

$\endgroup$
5
  • 1
    $\begingroup$ Is the flow laminar or turbulent? Is the velocity profile parabolic or flat (with slip at the wall)? Have you learned how to solve conductive and convective heat transfer problems? $\endgroup$ – Chet Miller Jul 23 '16 at 12:51
  • $\begingroup$ While I am slightly familiar with the terms, I do not possess enough knowledge to solve the problem myself. What would be some reasonable real-world assumptions? $\endgroup$ – user1202136 Jul 23 '16 at 14:07
  • 3
    $\begingroup$ Are you at liberty to say what the practical application is? Gert gave you an EXCELLENT physics derivation for this problem, but some of your starting assumptions will probably be difficult or impossible to meet (e.g., constant temperature on the inner surface of the PVC pipe). $\endgroup$ – David White Jul 23 '16 at 17:06
  • $\begingroup$ @DavidWhite: Thanks for the thumbs up. I addressed the problem of a slightly unrealistic constant temperature on the inner surface of the PVC pipe in an edit. But knowing the practical application always helps, of course. $\endgroup$ – Gert Jul 24 '16 at 2:36
  • $\begingroup$ You might want to experiment with tube lengths a bit. Your data shows your flow is laminar (low Reynolds number), not plug (turbulent, as assumed). Also, by assuming the inner wall temperature to be constant (see end of my answer) we're neglecting resistance to conduction by the PVC itself. All this leads to an underestimate of the required length. So start off maybe 30 % longer re. calculated, as a starting point . $\endgroup$ – Gert Jul 24 '16 at 17:36
5
$\begingroup$

Real world assumptions:

  1. No temperature gradients over the cross-section of the pipe.
  2. Plug flow (turbulent flow).
  3. Water heat capacity $c_p$ and density $\rho$ are temperature invariant.

Consider an infinitesimal mass element $dm$ at temperature $T(x)$ travelling down the pipe at speed $v$. We apply Newton's law of cooling to it:

Cooling tube.

$$\frac{dQ}{dt}=hdA\big(T(x)-T_t\big)$$ Where $Q$ is heat energy, $h$ is the heat transfer coefficient (water to PVC) and $dA$ the surface area of the mass element.

$$dA=\pi D dx$$

$$\frac{dQ}{dt}=h\pi D\big(T(x)-T_t\big)dx$$

When the mass element's temperature has dropped by $dT(<0)$, then:

$$dQ=-c_pdmdT$$ Divide by $dt$: $$\frac{dQ}{dt}=-c_p\dot{m}dT$$ Where: $\frac{dm}{dt}=\dot{m}$ is the mass throughput of the water. So we get: $$h\pi D\big(T(x)-T_t\big)dx=-c_p\dot{m}dT$$ For plug flow: $$\dot{m}=\frac{\pi D^2}{4}\rho v$$ With $\rho$ the density of the water. Slightly reworked we get: $$-\frac{4h}{c_p\rho vD}dx=\frac{dT}{T(x)-T_t}$$ Lets call: $$\alpha=\frac{4h}{c_p\rho vD}$$ Now integrate between relevant boundaries: $$\int_0^l-\alpha dx=\int_{T_0}^{T_1}\frac{dT}{T(x)-T_t}$$ $$-\alpha l=\ln\frac{T_1-T_t}{T_0-T_t}$$ $$\Large{T_1=T_t+(T_0-T_t)e^{-\alpha l}}$$ Note that for $l \to +\infty \implies T_1 \to T_t$.


Note also that due to the simplifying real world assumptions this is a 'quick and easy' solution. For laminar flow with temperature gradients $\frac{\partial dT}{\partial y} \neq 0$ and $\frac{\partial dT}{\partial z} \neq 0$, a quasi-2D Fourier heat PDE would have to be solved, which is mathematically far more demanding. Despite the simpler solution's limitations, this is the model widely used by engineers for the design of e.g. tubular heat-exchangers.

The OP's choice of setting $T_t$ as the constant temperature of the inside wall of the tube is a little awkward. Normally we would set $T_t=\text{constant}$ for the cooling fluid surrounding the PVC tube. That then requires the calculation of a thermal resistance $U$ of the pipe, as shown here. In the calculation above, $U$ would replace $k$.

$\endgroup$
10
  • $\begingroup$ In my engineering experience, engineers use the solutions to the 2D Fouirier heat PDE to handle laminar flow heat transfer (or solve the equation numerically). These solutions are also typically presented analytically in heat transfer books in terms of the dimensionless internal heat transfer coefficient (Nussult Number) as a function of the Reynolds Number, Prantdl Number, and L/D ratio. For turbulent flows, there are analytic relationships and graphical relationships of Nussult Number as a function of Reynolds Number determined from experimental correlations. $\endgroup$ – Chet Miller Jul 24 '16 at 1:20
  • $\begingroup$ @ChesterMiller: you're quite correct and I'm aware of a number of web resources that provide solutions as you describe. But the approach above provides good approximations for high $Re$ numbers, where $\partial T/\partial r \approx 0$. It's also what I was taught, so many moons ago. I'm currently solving for the 2D Fourier equation but with $\partial v /\partial r =0$. Taking actual velocity distribution (laminar case) into account is only solvable numerically, I think. $\endgroup$ – Gert Jul 24 '16 at 1:46
  • $\begingroup$ I'm also a Chemical engineer, BTW. Quite a coincidence... $\endgroup$ – Gert Jul 24 '16 at 1:55
  • $\begingroup$ Beyond the hydrodynamic entrance region, if it is assumed that the effect of temperature on the viscosity are small, the velocity profile will be parabolic in r, and independent of z. The constant wall temperature heat transfer analytic solution for this problem is in most books on fluid dynamics and heat transfer. My preferred text is Transport Phenomena by Bird, Stewart, and Lightfoot. $\endgroup$ – Chet Miller Jul 24 '16 at 12:34
  • 1
    $\begingroup$ @user1202136: well, thanks! Yes, both phenomena are driven by simple gradients. Exponential decay functions then inevitable pop up. $\endgroup$ – Gert Jul 24 '16 at 17:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.