Sensors - what and where

[Guest Alphonsox]

I've been a little distracted for the last few weeks and have been somewhat taken by surprise by the prompt arrival of the plastering crew. They are starting upstairs which means I have a window of a couple of weeks to finalise the wiring and such downstairs.

Currently I don't have any sensors in place and probably need some. I intend to use  DS18B20 temperature sensors for slab temperature measurement (how many points do I need to monitor ? ) 

What else should I be monitoring ? What have other people done ?

[reddal]

Hi,

I wish I'd done more of this on my build - there are loads of sensors tied in with the systems (heating, UFH, etc) but they typically aren't very accessible - you can't log the data and put it all together to see what is really going on.

If doing it again I would try to find a way to install independent sensors and datalogging of as many of these as practical :

• temperatures in each zone of UFH
• room temperatures in several places
• external temperature
• internal humidity
• hot water tank temperature (+thermal store temperature in my case)
• temperature of inputs to the thermal store (solar and stove in my case)
• circuit by cicuit electricity usage
• and probably more!

There is some complexity to the datalogging systems that I don't really know anything about - but I'd want to be able to get the detailed raw data onto a computer to analyse it there.

If you can find a way to setup a system so that adding additional sensors costs no more than a bit of wire and a few pounds for each sensor - then I would put in loads!

- reddal

[ToughButterCup]

I'll be following this like a hawk.

Ian

[tonyshouse]

I built in thermocouple probes, lots, data log the results.

[reddal]

4 minutes ago, tonyshouse said:

I built in thermocouple probes, lots, data log the results.

How does the datalogging work?

[worldwidewebs]

I hope this isn't a stupid question, but why do you want to log all this information? What are you going to do with it?

[tonyshouse]

Livorno data logger, puts results into a spreadsheet that can then draw graphs, I use a bespoke bms, does the Sam with more power, channels and flexibility. Thermocouples from here here

[SteamyTea]

14 hours ago, Alphonsox said:

What else should I be monitoring

 

1 hour ago, worldwidewebs said:

What are you going to do with it

It is always worth knowing why you want to monitor something as this denotes what you can do with the data.

Let us take a simple slab as an example.

You can just put a sensor in the centre (but which centre) and sample the temperature every day at the same times i.e. every hour.

On its own, that will not tell you much.  All it will record is the temperature at that point at those times.  It has nothing to relate to.

So let is add another temperature to do room temperature.  So where do we put this one.  Hanging from the ceiling so that it is in the centre of the room (assume it is a simple rectangle floor plan), or hide it in a corner, maybe at seat height, or low down, under a sofa or fridge.  Will it be affected by sunlight, draughts, mechanical air extraction?

Once you have sorted those out, you can now start comparing your two data sets.  There are 3 main ways to do that:

Temperature Difference

Binned Mean Temperature Data Sets (quite tricky to set up)

Time Series Data

They will all give you different results, so you need to know what you are doing it.

But that still only gives you a comparison between two points at any given time (slab and air temperature), so it is probably worth getting another sensor to do external temperature (or a local weather station to get data from).

Even then, it is only comparing temperatures and with a relatively large margin of error (there are statistical methods to show the bounds that are easy to set up and use).

The trouble with just measuring temperature is that you will not know if, it is the internal air temperature, or the external temperature that is affecting your slab temperature, or the slab temperature that is affecting your internal air temperature, as they both may be affected by the external temperature, or the insolation, or maybe the wind speed (has a greater affect than anything else at my site).

So you can see that is quickly becomes tricky to know what to measure, and when, where and how.

Now two simple scenarios:

Control internal air temperature from slab temperate

Establish relationship between slab temperature and internal air temperature

 

To control internal air temperature from the slab temperature is quite simple, you just plot slab temperature along the x-axis of a chart, and internal air temperature along the y-axis, draw a line of best fit (spreadsheets are excellent at this) and you will end up with an equation (stick to a simple straight line y=mx+c).  As you have forced the 'controlling' factor, the slab temperature onto the x-axis, by knowing that temperature, you can just read of the chart what the air temperature should be (there will be margins of error).

So if the slab gets cold, you heat the air, if it gets hot, you cool or replace the air.  Simple, but crude.  Much easier to just control the air temperature from the air temperature, it will be more accurate.

To establish relationship between slab temperature and internal air temperature is really just a matter of collecting enough data and plotting the slab temperature against the internal air temperature, but without, or allowing for, external factors.  So no heating, cooling, bulk air movement (sealed room), then look to see what the time difference is between changes.  So you can start by slicing the data up into seasons, months, weeks, hours of daylight, intensity of sunlight, wind direction, wind speed, rainfall, curtains opened or closed, room usage etc, and just compare that to the 'sealed room' data.  This will usually need the logger set up differently, so rather than sample at fixed times, you sample at fixed temperature differences i.e. every half degree.  Then you look at the lag times, these will usually overlap because of the differing thermal properties between the slab and the internal air.  You can then start to interpret the data and see where the variances are, but you still need to know why you are doing it.  So assuming that it is for internal temperature control, you can create a large data array that allows you to set the heat input parameters i.e. how much and how long to heat, from the historical data i.e. slab is 2°C below optimum, I need 3 kWh of thermal energy in the room in the next 7 hours, or whatever.

 

So after that ramble, decide why you are doing it first, then how you will do it, then what you will do with the data.  Before long you will be an expert at conditional data manipulation.

 

 

Edited May 26, 2016 by SteamyTea

[Guest Alphonsox]

Thanks for that Steamy - The temperature data will be primarily for room temperature control, I don't know exactly how this will be done at the moment and probably won't do until we have lived in the place for a few seasons. My expectation is that it will either be a dead simple thermostat (which I think is what Jeremy ended up with) or a ridiculously complex predictive system based on historic data, current temperature measurement and predicted changes over the next 24 hours. Given that I may want to do the latter I need to get the sensors in place now - retrofitting is likely to be impossible.

[DeeJunFan]

Apart from the smoke, CO and CO₂ I will only require one temp sensor

Sensor cost - Approx £50,000

Sensor Value - Priceless

SWMBO can sense temp variations of 0.01 degree and reacts to them within 0.1 of a second.

Data logged (Forever)

Aggregated Data (this house is freezing)

 

Edited May 26, 2016 by DeeJunFan

[SteamyTea]

2 hours ago, Alphonsox said:

Given that I may want to do the latter I need to get the sensors in place now - retrofitting is likely to be impossible.

Put in 3 or 5 sensors in a diagonal line from the brightest corner to the darkest corner.  That way you can get a gradient across the slab and eventually just use the one that gives you the closes match to what you need to control.  Make sure you know where they are, so if you place any furniture or extra runs over them, you know.

If you have MVHR, put the air temp sensors in the room outlet, will almost certainly give a better result, except above a cooker.

[reddal]

2 hours ago, SteamyTea said:

Before long you will be an expert at conditional data manipulation.

As Steamy explains there is a lot you can do with detailed data. If anyone has this kind of data but doesn't feel like they have the data analysis skills to extract the value from it then I will have a go for you if you give me all the raw data.

Compared to the giants around here there aren't many topics I can contribute expertise on - but data analysis is something I can at least have a go at!

- reddal

[SteamyTea]

Can send it to me too and if I get time, I will show you some different results

[daiking]

Does anyone have a load cell in their pan?

[SteamyTea]

Called a weighing scale on WeightWatcher day isn't it

[Woodgnome]

Quick question, got screed being laid on Tuesday and completely forgot about sensors!

Should the sensors all run to a single point, or maybe say 2 or 3 points depending on how many sensors I put in the slab? 

[jack]

Look into one-wire sensors.  They're cheap and you can run several of them off one cable.  You do need a way of reading the values - we have a one-wire module in our home automation system.

[SteamyTea]

You can get the 1-Wire temp sensors pre-wired into a string.  May be easier, quicker and in the end, cheaper to get some lengths of that.

http://www.homechip.com/1-wire-controllers/sensors/t-string-temperature-sensors.html

 

Have a word with these people, I have always found them helpful.

https://www.homechip.com/

Edited August 13, 2016 by SteamyTea