27.12.2012 Views

Newton's law of cooling revisited - Cartan

Newton's law of cooling revisited - Cartan

Newton's law of cooling revisited - Cartan

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

1074 M Vollmer<br />

<strong>cooling</strong> time, i.e. transition temperature difference, where the dominant <strong>cooling</strong> changes from<br />

radiation to convection. For 10 W m −2 K this happens at �T = 137 K (after ≈840 s in a T(t)<br />

plot)) and for 30 W m −2 K it happens at �T = 415 K (after 112 s). Qualitatively it makes<br />

sense that the higher this transition temperature difference, the larger the range <strong>of</strong> validity<br />

<strong>of</strong> Newton’s <strong>law</strong> <strong>of</strong> <strong>cooling</strong>. For the convective heat transfer <strong>of</strong> 100 W m −2 K and ε = 0.1<br />

(see figure 5), convection would dominate radiation right from the beginning, which easily<br />

explains the linear plot.<br />

From the theoretical analysis, it is clear that radiative <strong>cooling</strong> should lead to deviations<br />

from Newton’s <strong>law</strong> <strong>of</strong> <strong>cooling</strong> above critical temperature differences, which in some <strong>of</strong> the<br />

discussed cases were below 100 K. This raises the question <strong>of</strong> why many experiments reported<br />

the applicability <strong>of</strong> Newton’s <strong>law</strong> for a temperature difference range <strong>of</strong> up to 100 K. The answer<br />

which is proposed here is simple: if one waits long enough, any <strong>cooling</strong> process can probably<br />

be described by a simple exponential function.<br />

To my knowledge, this statement has not been proven theoretically for the <strong>cooling</strong><br />

<strong>of</strong> objects involving nonradiative heat transfer. The idea behind it may be motivated for<br />

convective <strong>cooling</strong> <strong>of</strong> simple shaped objects irrespective <strong>of</strong> the size or Biot number by the<br />

following argument. The <strong>cooling</strong> <strong>of</strong> objects such as spheres, cylinders, or plates <strong>of</strong> any size<br />

(not just small objects), which start at a given initial temperature and which are in contact<br />

with a fluid (radiative heat transfer is neglected) can be described by a series expansion <strong>of</strong><br />

exponential functions [2]. It is found that for sufficiently long times, a single term in the series,<br />

i.e. a single exponential function, describes the temperature distribution within the objects. If a<br />

suitable average temperature is defined, the single exponential function will therefore describe<br />

the <strong>cooling</strong> process.<br />

With this result from pure convective <strong>cooling</strong> in mind, theoretical <strong>cooling</strong> curves involving<br />

nonlinear radiative heat transfer were analysed by using series expansions <strong>of</strong> exponential<br />

functions as fit functions.<br />

Figure 7 (left) depicts the theoretical temperature plot for αConv = 10 W m −2 K, ε = 0.9<br />

and 40 mm Al cubes <strong>of</strong> figure 5. These theoretical data were fitted (figure 7 (right)) using a<br />

third-order exponential fit <strong>of</strong> the form<br />

�T (t) = T0 + A1 · e −t/τ1 + A2 · e −t/tτ2 + A3 · e −t/tτ3 . (11)<br />

The agreement <strong>of</strong> such a simple fit is extremely good as can be seen from figure 8, which depicts<br />

the difference between the theoretical plot and the third-order exponential fit. Disregarding<br />

the first 10 s, deviations are below 1 K.<br />

The fit includes three functions with different amplitudes and time constants (the constant<br />

term <strong>of</strong> 0.07 is practically zero and unimportant for the discussion). The first one has the<br />

smallest amplitude (≈162) and a time constant <strong>of</strong> only 78.7 s. The second contribution has<br />

an appreciable amplitude (≈253) but also a relatively small time constant <strong>of</strong> 297 s. The<br />

third contribution has an amplitude slightly larger than the second one (281) but decays with<br />

a much longer time constant <strong>of</strong> about 1010 s. Due to exponential decay an amplitude is<br />

suppressed to less than 1% after five time constants. Hence, the first contribution already<br />

contributes less than about 1 K after 400 s <strong>of</strong> <strong>cooling</strong>. Similarly, the second contribution<br />

becomes less important with time. For times longer than 1500 s it only contributes less<br />

than 1.7 K. This happens at a temperature difference <strong>of</strong> about 65 K, which means that for<br />

longer times, i.e. smaller temperature differences, a single exponential provides a very good<br />

approximation to the <strong>cooling</strong> curve. Hence, if we assume that it is a general property <strong>of</strong> the<br />

combined convective and radiative <strong>cooling</strong> that the resulting T(t) curve can be approximated by<br />

a superposition <strong>of</strong> exponential functions with different time constants, one does indeed expect<br />

that after the exponentials with small time constants have died away, a simple exponential

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!