It's Physics time!
=
Ok, what is "loudness"? To say it easy, the amount of noise something produces is the amplitude of a compression wave of air. When it hits a surface, it transmits through that material and/or is reflected by it. The transmitting sound is distorted in the process before exiting on the other side.

Now, we measure sound pressure - which correlates with the ampliude of the compression wave - in decibel, or dB for short. dB is a logarithmic scale, where 3 dB means 2 times the energy and \$\sqrt 2\$ the amplitude of the wave. In other words: something that is 10 times as loud has 100 times the energy and is 20 dB louder.

It also has a neat chart that tells us how many dB something creates when it is a set distance away on [wikipedia][1]. Combat could from own experience (a sword clashing onto a helmet next to me) reach the noise of a vuvuzela horn at 1 meter distance (120 dB, 20 Pa) but usually will be a bit lower than what a Jackhammer does at the same distance (100 dB, 2 Pa). 

Dish out the math already!
=
So, let's assume at 1 meter (\$r_1 = 1\text{m}\$) it's as loud as a noisy road in 10 meters distance, so \$p_1 = 0.632 \text{Pa}\text{ or }90\text{dB}\$.

What is that pressure at the door? \$p_2 = p_1\frac{r_1}{r_2}\$ tells us. let's assume the room has 2 meters to the door. So \$p_2= 0.632\times\frac{1\text{m}}{2\text{m}}=0.316 \text{Pa}\$. This is equivalent to a ca 84dB at ear.

Let's assume the door does not distort the sound or reduce the sound pressure in a meaningful way (it does, but one can't calculate that easily at all). So we have a 0.316 Pa noise emitted from the door to be heard in 20 feet (=6.096m). That's \$p_3=0.316\text{Pa}\frac{1}{6.096}=0.052\text{Pa}\$, so about the sound of a loud conversation or low key yelling (ca. 68dB). 

A curtain (which again should dampen the sound but does not for the sake of simplicity), and another 10 feet (=3.048m) later, the sound pressure is down to \$p_4=0.017\text{Pa}\$, which is around 58.5dB. Remember, a normal conversation is 40 to 60dB, so this could be barely audible against the ambient noise of a lively room (of 70dB+).

But what if they are louder?
=
If we assume the people yell and scream at the top of their lungs and produce 120 dB or 20 Pa at the 1m Range, the sound travels still 11.144 meters, so the sound pressure drops - assuming no doors and curtains but one straight line without obstacles - \$p_2^*=20\text{Pa}\frac{1}{11.144}=1.8\text{Pa}\$ or ca. 100 dB, and taking each of the obstacles as a new source for the sound, just about 0.54 Pa, ca 88.5dB. This does not take into account the losses of the curtain and door, but if they would not be there, this yelling and screaming would be audible clarly.

That doesn't make sense to me!
=
Why does the formula seem to quirk out for iterative setup in comparison to cummulative? The cummulative assumes we have one corridor of sound. The iterative assumes we transmit the sound through some perfect, non dampening material between three areas of a set length. Think like a tube in which you put in thin layers of foil: even their mere presence does dampen the sound.


tl;dr:
=
At this distance the sounds *could* be drowned out by the noise of the lively mess hall if one takes into account the dampening properties of a door, the hallway and the curtain, but if for some reason they all would be silent for just a moment, they will hear the basic level of combat noise even over the ambient 70dB (like a TV) of a mess hall. In any case they should at least hear the spikes of the death screams or barbarians raging, fireballs detonating and other particular loud noises.

If the guards don't notice the basic combat sound for some reason it doesn't however stop a sloppy guard that is late for lunch break from passing the door that the players fight behind and hearing it through the door and then run to alert the rest.

In that case the closed door doesn't help the players at all, as their own combat noise would drown out these alerted guards advancing down the corridor and taking position on both sides of the door. Running in armor should be roughly 70dB at 1 meter. Assuming the heroes are 2 meters in the room and the goblins rub their bellies against the door in going into position, this is 63dB at the player's positions and *that* surely will drown out in their own combat noise of 90dB at least. But again, the door and air in the hallway dampens that sound before pressing through the door, so... *if* you rule that somebody manages to alert the garrison, and they manage to be quick enough to come while still fighting, the players should not hear them coming.

<sup>Disclaimer: The calculation makes a few assumptions and simplifications and takes a few shortcuts that result in the numbers being unrepresentative for real calculations, especially since the dampening of various materials (including air) is among other things frequency and pressure dependant. Also, the basic assumption does not include the channeling, dampening or directing properties of a stone corridor and the wall geometry and their decoration. But these could cancel out with the door and curtain properties. It does however allow an estimation as the results should be within the same order of magnitude.</sup>
  [1]: https://en.wikipedia.org/wiki/Sound_pressure#Examples_of_sound_pressure