Combining radiators with UFH - some surprising conclusions

[sharpener]

Puzzled by some previous threads about mixing I have run a model of a heating system that combines radiators designed for an average temp of 45C with UFH designed for avg 30C.

 

The model assumes you have an HP which produces 5kW at 5 deg delta T (this keeps the numbers simple), and want 2.5 kW out of each side, as might be the case in a house with UFH downstairs and rads upstairs. The return pipe flows are allowed to mix on the way back to the HP.

 

A few simplifications have been made: the heat loss from the emitters is taken as proportional to the avg temp which is not quite true for the rads, and also at low flow rates the temp drop profile along the UFH loops is not linear but an exponential decay. But I think the general shape of the curves is right and the conclusions are valid.

 

Here is the graph of the various temps as a function of the proportion of the flow going to the UFH. The surprise is that for an equal 2.5kW/2.5kW split you only need 7.6% of the flow to go to the UFH circuit, because its emissivity per degree is much higher and the equilibrium flow temp is also 17 deg higher than it expects.

 

For a 50:50 division in the flow you only get 1/3 of the heat into the radiators and 2/3 goes into the UFH. As you can see, this does not vary a great deal over the whole central portion of the diagram.

 

 

 

 

 

I don't think the conclusion would be much affected by the conventional UFH arrangement of using a circulating pump and blending valve to temper the actual flow temp to the UFH loop but I have yet to complete the modelling of that arrangement.

 

Notes on the model: It arrives at the equilibrium temperature values by solving the following simultaneous equations (which give rise to some very complicated algebraic expressions in r, Krads and Kufh).

 

(i) The HP produces 5kW at 5 deg delta T - this defines the relationship between flow and (combined) return

(ii) Each circuit dissipates an amount of heat equal to the average temp between its inlet and outlet multiplied by a constant K, from which the temp drop across each cct can be calculated

(iii) The two circuits together dissipate exactly the 5kW ouput of the heat pump

(iv) The combined return temp is the sum of the individual return temps weighted by the respective flow rates (Method of Mixtures).

[JamesPa]

Looks interesting but Im not sure I get what is going on from a plumbing perspective, can you clarify.

 

Are you feeding water at 45C both the the rads and the UFH, however because the UFH flow rate is much lower the average UFH temperature is lower than the average rad temp.  Or are you mixing 45C water down to 30C.

[nod]

Bit of over thinking going on 

I think most have a combination of GF UFH and rads in the bedrooms 

[PhilT]

Slightly off topic but 45 is too high for an optimal new rad system. My average this year so far is around 37 with 100% rads. Design temp was 40.

[akjos]

From a theory and calculation perspective I think your conclusions make sense. But in reality this setup would be unpractical.

If I understand correctly, based on your following statement:

12 hours ago, sharpener said:

I don't think the conclusion would be much affected by the conventional UFH arrangement of using a circulating pump and blending valve to temper the actual flow temp to the UFH loop

you have modelled the system to receive 45'C flow on both Rads and UFH. Your UFH return is 20'C because the flow rate needs to be very low to only output 2.5kW.

UFH has a huge surface area, and feeding it 45'C it makes sense to only require a tiny portion of the overall flow rate.

 

But in reality, you don't want to feed 45'C to UFH and have a 20'C return. There will be big fluctuations across the floor, cold spots, hot spots and flooring most probably wouldn't support it.  UFH should be at 5-7'C delta T to be comfortable, that's why the 'more conventional' setup has a separate pump and mixes the flow.

 

Ideally we'd run the whole system on a lower temp, oversize rads and have everything running on a lower flow temp with both rads and UFH giving an acceptable delta T.

[JohnMo]

At the start of the heating season I was trying to optimise running the house and summer house UFH systems on a common flow temp without mixers. 

 

Turned up the heat pump flow temp to 32 from 29.5, also increased the dT on the house UFH to decrease output as far as I could and keep the heat pump happy enough.

 

Basically the house overheated after about 18 hrs was up to 24 degrees, even though the thermostat stopped the demand for heat, the floor had absorbed lots of heat. The summer house wasn't warm enough.

 

If you want two flow temps use a mixer.

[Dunc]

Forgive some daft questions:

Why is Tret bottomed out at 20 when r is either 0 (Tret from UFH =20) or 1 (Tred from rads =20)? 

Again at the extremes of the graph, why are TretUFH or TretRads different from TretCombined when r is either 0 or 1? i.e. when the entire flow (r) is going to either rads or UFH alone why is the "combined" return temp not the same as the one open circuit?

ta.

 

[sharpener]

3 hours ago, nod said:

I think most have a combination of GF UFH and rads in the bedrooms 

 

That is exactly the situation I am trying to model as it corresponds closely to my barn conversion retrofit except the numbers are 6 + 6 kW.

 

3 hours ago, JamesPa said:

Looks interesting but Im not sure I get what is going on from a plumbing perspective, can you clarify.

 

Are you feeding water at 45C both the the rads and the UFH, however because the UFH flow rate is much lower the average UFH temperature is lower than the average rad temp.  Or are you mixing 45C water down to 30C.

 

I am taking a common feed from the HP to both rads and UFH, at whatever flow temp is necessary to dissipate a total of 5kW (dark blue top line). At r=0 this has to be 72.50(!) to get rid of all the heat via the rads. At the 2.5:2.5 power point it needs to be just under 47C (bc the flow rate through the rads is higher than the design condition so the delta T is less).

 

(At such a low flow rate the equations are actually satisfied by a UFH return temp of about 12C which is clearly not attainable hence I have put in a lower limit of 20. Why this occurs I have not yet worked out, but see below regarding the UFH temp profile. OH teaches maths at Cambridge University and has independently solved the equations in the model to reach the same result but can't shed any light on this.)

 

1 hour ago, akjos said:

But in reality, you don't want to feed 45'C to UFH and have a 20'C return. There will be big fluctuations across the floor, cold spots, hot spots and flooring most probably wouldn't support it.  UFH should be at 5-7'C delta T to be comfortable, that's why the 'more conventional' setup has a separate pump and mixes the flow.

 

Have now modelled this too using heat loss proportional to average temp. If s times the flow to the UFH is mixed back from the UFH ret with a pump and mixer then the ratio Tf/Tu is given by

 

2(s+1) + Ku(2s+1)

    (2(s+1) - Ku)

 

which tends to 1 + Ku for large s so in this case where Ku = 0.25 (kW/C) the limit is 1.25

 

If we assume s might be 4 this gives 12.25/9.75 = 50/39 = 1.256

with no recirculation at all k = 0 so Tf/Tu is 2.25/1.75 =  1.286 so not a lot different. 

 

But I cannot reconcile this result with the temps from the overall model and I think the problem is the "average temp" heat loss assumption breaks down for the UFH where an exponential decay temp profile along the loop is more likely in RL.

 

1 hour ago, akjos said:

Ideally we'd run the whole system on a lower temp, oversize rads and have everything running on a lower flow temp with both rads and UFH giving an acceptable delta T.

 

Yes of course but in my retrofit situation to get to 45C flow I have already had to re-size the rads by a factor of 1.9x and to match  to 30C would require a further 2.5x increase in size!

 

[sharpener]

43 minutes ago, Dunc said:

Why is Tret bottomed out at 20 when r is either 0 (Tret from UFH =20) or 1 (Tred from rads =20)? 

 

Equations actually result in a figure of less than 20 in these cases which is not physically possible, see my recent post.

 

45 minutes ago, Dunc said:

Again at the extremes of the graph, why are TretUFH or TretRads different from TretCombined when r is either 0 or 1? i.e. when the entire flow (r) is going to either rads or UFH alone why is the "combined" return temp not the same as the one open circuit?

 

At the lhs the pnk and green lines do converge though the pink is not that easy to see, soz (the pic is a screenshot of a pdf print of the xls). At the rhs they don't bc I now notice there is a data point missing from the pink line, the missing value at r = 1 is 17.50, same as light blue. This gives an avg temp of 40 to the UFH which is what is required to dissipate twice the design value (2.5kW at 30C). HTH.

[sharpener]

1 hour ago, JohnMo said:

If you want two flow temps use a mixer.

 

I think the main reason you need a mixer and pump is to achieve a flow rate within the UFH loop(s) sufficient to distribute the heat evenly along their length.

 

From the heat loss equation point of view it is the volume supplied at the higher temperature that matters as it defines the quantity of heat input, for me the surprising point is how little flow you need. A thermostatic mixer will of course achieve this by closing the inlet so most of the flow to the manifold is what is recirculated from its return pipe.

[JohnMo]

1 hour ago, sharpener said:

 

I think the main reason you need a mixer and pump is to achieve a flow rate within the UFH loop(s) sufficient to distribute the heat evenly along their length.

 

From the heat loss equation point of view it is the volume supplied at the higher temperature that matters as it defines the quantity of heat input, for me the surprising point is how little flow you need. A thermostatic mixer will of course achieve this by closing the inlet so most of the flow to the manifold is what is recirculated from its return pipe.

A small heat runs out flow rate by the you get to +/-1m3/h. My 6kW heat pump was flowing just enough to drive 9 (7 in house and 2 in summer house) loops from about 15m away from the manifolds. Had a dT of 4 to 6 depending on room.

 

But the return temperature from a thick screed/concrete floor will dominate max flow temp you can achieve during a heating cycle, without a mixer. Mine rarely has a return temp above 27 to 28 (electronic mixer set to 35), so no matter what flow temps I may ask the heat pump to do my flow is capped at about 32 to 33.and takes about 8 hrs to get there.

 

So trying to spread the heat evenly between radiators and UFH without some form of mixer isn't going to work well.

 

If I drop my mixer to control to 28, so I get plenty of floor water recycling the heat pump does increase flow temp because the house flow temp is less dominant. But I have to run the heating most the time.

[ReedRichards]

Isn't a combination of radiators and UFH simply the wrong thing to do?  You are almost certainly going to need a higher flow temperature for the radiators so you don't get any economic benefit from the UFH.  If you like UFH then install it throughout and save on running costs.

[ProDave]

21 minutes ago, ReedRichards said:

Isn't a combination of radiators and UFH simply the wrong thing to do?  You are almost certainly going to need a higher flow temperature for the radiators so you don't get any economic benefit from the UFH.  If you like UFH then install it throughout and save on running costs.

And, if you build the house well enough, with enough insulation and air tightness, then as many of us have proved (even up here in the Highlands) then you way well not need any heating upstairs, just UFH downstairs.

[sharpener]

1 hour ago, ReedRichards said:

Isn't a combination of radiators and UFH simply the wrong thing to do?  You are almost certainly going to need a higher flow temperature for the radiators so you don't get any economic benefit from the UFH.  If you like UFH then install it throughout and save on running costs.

 

This is a counsel of perfection. Many on here including me have to start from having a property with rads already fitted upstairs, and the upheaval of installing UFH in their place makes it impractical.

 

And as @ProDave says we do not need much heat upstairs, there is no insulation in the bedroom floors so they are 16+ without any heating on.

 

My propossed solution to this is - when the rads are required at the same time as the ground floor UFH - to run them from a 55C thermal store charged up at off-peak rates, while the UFH is supplied at the lower flow temp in real time. The house is long and thin and I don't want all of it heated all of the time.

 

That is if the installer ever gets back to me now that Vaillant have issued their final approved drawings for it.

[sharpener]

Has anyone got a formula for how UFH heat dissipation is related to the circulating water temp?

 

I have found trawling through various websites the following rules of thumb:

• dissipation should not exceed 100W/m^2 for concrete floors and 75 for wood
• pipe spacing should be 200 mm for boiler systems, and 150mm for HPs with a flow temp of 35C (this gives 5m and 6.67 metres of pipe/sq m of floor respectively)
• Volume of water in system is 0.12 litres/m pipe run (which is 0.6 and 0.8 litres/ sq m respectively)
• Flow rate in l/min should be loop length in m divided by 40. (Combined with the previous result this gives a constant dwell time of 4.8 mins independent of length.)
• Max length should be 110m for reasonable pressure drop (At the flow rate given above the flow velocity is 0.4m/s and the head required for 110m is about 2.8m)

All well and good. A bit of additional info in this thread.

 

But nowhere can I find the flow temp required for a given heat output to the room for typical construction and floor coverings, any ideas?

 

[JohnMo]

 

 

I have these, different outputs for different pipe spacing and mean flow temperature 

 

[John Carroll]

Fig 2 is interesting. At pipe centres of 200mm, the floor heat output at 45C is 75W/m2 (floor area), 40C is 58W/m2 & 35C is 43W, if one was to use a 20C required room temperature and use a exponential of 1.15 then it matches the outputs reasonably well,                                                                                                                                                                                                            for example the output at 40C = ((40-20)/45-20))^1.15 X 75, 58W/m2 and at 35C = ((35-20)/(45-20))^1.15 X 75, 41.7W/m2.

 

 

 

Edited April 6 by John Carroll

[sharpener]

Thanks @JohnMo that is immensely helpful.

 

The immediate takeaway is that over the whole MWT range 30 - 50C the o/p appears to be directly proportional to MWT-20 (so is obviously intended for a room temp of 20C).

 

Also by changing to 150 mm centres and adding 33% in terms of pipe length you only get a 10% increase in o/p.

 

In my case my HP installer has assumed I have 150mm centres (which I somehow doubt) in the 1995 floor in my living room, and come up with a 62 W/sq m heat loading to get a room temp of 18C. Using yr helpful nomogram this equates to a MWT of 37C.

 

However his total of 2522W does not sound enough for a 40.6 sq m room in a stone barn conversion. And to get to 20C when it is freezing outside I will need 10% more, so MWT = 41C (43 if the pipes are actually at 200mm centres). Fortunately I am not reliant on the UFH as I have already added 2 rads to get faster warmup, so using them in addition will I hope push the flow temps back into the region of sensible CoPs.

 

Edit having now read @John Carroll's post I agree about the 20C target but am not convinced we need a higher power than 1.00. For example for the 200mm centres and 30 ... 50C I had written down 29, 43, 60, 75, 90 W/m^2 which is pretty much linear. Is there a theoretical basis for the 1.15, like there is for the 1.3 or whatever applied to rads?

 

 

Edited April 6 by sharpener
Cross-posted with @John Carroll

[John Carroll]

I have no theoretical basis, I just see that the UFH output isn't a straight line which may have indicated a not quite linear relationship.

[JohnMo]

On 07/04/2024 at 08:39, John Carroll said:

I have no theoretical basis, I just see that the UFH output isn't a straight line which may have indicated a not quite linear relationship.

Fig 2 has a curve already built in.