Weight fluctuations tracking with temperature?

11 posts / 0 new
Last post
Dennis
Weight fluctuations tracking with temperature?

Time to resurrect some discussion on this site. I've noticed a daily weight increase of about 1 lb, which drops off overnight. In a healthy hive on a warm day, I'd have thought this was due to the bees bringing in water and or nectar. But this particular hive is barely clinging to life and I see almost no traffic in and out of the hive -- certainly not enough to account for any weight variation. It looks like the weight is tracking with temperature or day light. Anyone else see anything like this?

Image: 
Nate
Nate's picture
I can confirm temperature sensitivity

I had the exact same profile while using mine - I gave up last year after never being able to solve it. I think my cells were just not temp compensating like the specs said they were supposed to... so that said a relatively simple temperature compensation algorithm in the code would get rid of it.

Paul
Paul's picture
Dennis, Please post a graph

Dennis, Please post a graph that includes humidity. If you have any wooden supers (or wax or honey) on the scale it will absorb moisture. I think you will find that the correlation between weight and humidity is much stronger than weight/temperature.

To test the temperature sensitivity of the load cells, it is best to load them with something that doesn't absorb water, like iron or steel. I recommend testing each scale before it is deployed by loading it with barbells, chain, etc. and letting it run in place with temperature swings. It is OK to upload data during the testing.

Thanks!

Dennis
Same date range with Humidity

Paul --
I see no obvious correlation between humidity and weight. The load cells are mounted on an aluminum frame and fully enclosed with the exception of the leveling foot of each cell, which protrudes through a plastic cover. The fluctuations in weight track with daylight. So if I didn't know better, I'd have assume this was the result of bees bringing in water. But that particular colony has failed -- no queen or brood and only a couple hundred bees who aren't leaving the hive at all. So the weight should be stable.

Image: 
Paul
Paul's picture
Uneven heating of load cells

We use "4 wire" load cells. I think each cell has 4 strain gages connected in a bridge configuration. For temperature compensation to work best, I think that all 4 stain gages must be at the same temperature, which is usually true. However, if part of the cell is exposed to the sun, and part in the shade (which results in uneven heating) the temperature compensation may not work.

I have no problem with adding temperature compensation to the software but: 1. I want to get it as good as possible with out it and 2. to be done "right" I think it needs temperature probes that are tightly coupled (thermally speaking) to the load cells.

Just my 2 cents.

Paul

Emil
Emil's picture
HX711

I think this is the HX711. I did some test with different load cells and also more HX711's, and I can provoke it with just holding on the heat sink of the HX711. I have 2 scales, both RPi in a small box, but one with Power on all the time (emil) (the HX711 is almost on the same temp all the time) this works fine, and one which is powered up once a hour for a few minutes and this follows the outdoor temp (norway1), it has been off since january because it's out of power, and there is a lot of snow...

Dennis
HX711 weight fluctuations

If the weight fluctuations are actually driven by the temperature of the hx711, I suppose I could just try insulating the box containing the r-pi. But that'll have to wait until I can recover the system, which appears to have been scrambled by a recent power outage that involved a bit of flickering on and off. I suppose I should get in the habit of periodically backup up the system. sigh...

MarkG
I've been testing some hives

I've been testing some hives and also have weight movement as per the pic's. Both hives are located out of direct sunlight in my shed. I've been reading the above posts and the concern over humidity and temp.

Hive 001 has 2 x concrete Besser blocks weighing 24.9kg
Hive 002 has plastic drums of water weighing 64kg
Inside Hive temp was added yesterday afternoon - but of course the sensor is just dangling near the weights, as is the Rpi themselves.
(I've a 3rd but only loaded with weight yesterday afternoon - so too early to comment at present.)

WRT hive001, it is possible that 2 concrete blocks could absorb moisture from the air - but I'm not convinced it's almost a cup of water in variation :)

WRT hive002, can't be from the influences of humidity on the actual weight, and the variation, with a baseline of 64kg, from 59kg to a peak of 103kg had me wondering for a fleeting moment if a large Koala had broken into the shed and sat on the scales :)

Hive 002 is also located against a wall/window which faces the rising sun....but is not in direct sunlight. So rising temp would seem the only variable.

Clutching at straws... is it possible thermal expansion of the Aluminum scale frame together with the rubber feet stuck to the concrete floor is causing stress in addition to the real weight?

My next step is to box the RPi and reduce the temp swings of RPi and Hat as per earlier posts, however am open to thoughts on how to stabilize weight so that meaningful interpretation can be undertaken when in the field.

Image: 
Hive001- over 2 days
Hive002 - over 2 days
Paul
Paul's picture
Weight fluctuations

Can you swap the weights so:
Hive 001 has plastic drums of water weighing 64kg and
Hive 002 has 2 x concrete Besser blocks weighing 24.9kg

Also, can you post the same graph but with humidity for bh001?

I think there are two separate issues here. I'll start with bh001 because it is easier.

Yes, I think the blocks can absorb and loose quite a bit of water, depending on the swing in RH. (Thats why I asked to see the RH graph and for you to change weights to something not as absorbent.) You'd be amazed how much a wooden super with empty drawn comb can move, or even just a board.

bh002. It has definite problems. The weight is very noisy. I suspect a bad connection on the analog side of the HX711. That would be the RJ11 connectors on the load cells - try re crimping them, very, very hard, or the 6 pins on the HX711 board or Hive Interface board - try re flowing the solder connections if you can. I'll try to post pictures of which pins could cause this problem.

The load cells work by very tiny changes in resistance. Any other changes in resistance in the circuit, like contact resistance, an insulation displacement connector (RJ11) not fully crimped, or a cold solder joint that changes with temperature can cause large changes in the readings.

You could swap HX711 boards between the two systems (and hive interface boards, too) to try to find the problem.

MarkG
I can swap the weights over

I can swap the weights over tomorrow morning (GMT+10) easy enough. When I do I'll encase the Besser blocks in a plastic bag to reduce fluctuations.

Humidity graph for bh001 is attached....I grabbed this from hivetool.net, however see that it's not fully current. Have just logged in remotely into bh001and grabbed an additional screen shot from localhost with up to date data - past 3 days so it also overlays the data on hivetool.net.

I'll review the connections and HX711 board on bh002 tomorrow also.

Hive 003 has now accumulated some data - it's sitting on a wooden table in the same shed as the others. It has 39.9kg comprising of 1 x Besser block (14.1kg) wit the remainder being water in drums.

Given that the Besser block absorbs moisture, I'm assuming a 150gm variance is likely to be acceptable.

Image: 
Hive 001 with humidity - 2 days
Hive 001 with humidity - 3 days (localhost)
Hive 003 with humidity - 1.5 days
MarkG
An update - With the porous

An update - With the porous Besser Blocks encased in plastic....

Over 2.5 days:
Hive bh001 has a weight range of 63.92 to 64.23kg. That's 310gms across 64kg or approx 0.5% variation.
Hive bh002 has a weight range of 39.97 to 40.10kg. That's 130gms across 40kg or approx 0.25% variation.

The graphs below, with temp and humidity, accentuate the variation, but with variation at <0.5% I'm thinking I should be satisfied, irrespective of what is causing the variation. :)

As for hive bh002, I swapped the Hivetool Hat for one not yet used, checked over the scale connections but still had noise issues. I've now replaced the scales and early indications are that the noise has reduced and scales are behaving as designed.

Image: 
Hive bh001 - 2.5 days
Hive bh003 - 2.5 days
Log in to post comments