[Closed] Verifying a weigh scale (I'm really sorry!)

Given that the tolerances are expressed as tiny percentages then surely your test weights need to be as heavy as possible to make those tiny percentages be as measurable as possible.

If you place 6 separate 500g weights and it read 3000.00 then that'd be a better test than the 50 g weight surely?

It can only held to be as accurate as its output scale, surely?

Ie, if its listed tolerance is 0.00048g but it's actually out by ten times that, 0.0048g, your scale will still correctly read 50.00g. So it doesn't matter.

[EDIT] Tolerances are presumably far stricter than they need to be to achieve the end result for just this reason. If a bridge needs to support five tons, you don’t build it to support 5.0001 tons, you build it to support ten.

Get someone in to do an onsite calibration and get a test certificate ?

EDIT -

In the spec there is no “tolerance” on accuracy, it only quotes combined error as <=0.016% of rated capacity. This is <=0.0004 8 g!! This is less than a grain of rice weighs. The weigh scale indicator is 000.00 so how can I demonstrate that it is operating to specification?

You cant, you need more accurate equipment.

Pharma equipment plus a very fine tolerance for accuracy - i'd be looking for a specialist to come out, do the calibration and issue a certificate if I were PM in that project.

Talk nicely to Trading Standards - they should have testing facility and/or contact who do.
That or speak to the manufacturer about calibration and testing.

It's not weighing APIs or anything critical. It's actually keeping track of the weight of a litre bottle of H2O2 as it decreases over time as it feeds a decontamination system.

We will get it calibrated by Mettler Toledo when it gets to its final destination. There's no point doing it now as it's going to be relocated three times.

Got a weigh scale (0-3 kg) integrated in a piece of pharma equipment and the client wants to see it verified during factory acceptance test.

the words in bold all have very specific meanings and presumbaly mean that someone has determined that you should be working to a defined standard (GMP or similar) in which case there should probably be an OQ,IQ,PQ plan somewhere by someone who gets paid far too much for arguing the toss about semantics...

it only quotes combined error as <=0.016% of rated capacity. This is <=0.0004 8 g!!

are you sure? the rated capacity is 3000 g. So 0.016% of that is 0.48g.

If I place a 50 g weight on the platform and it reads 50.00 would this verify accuracy?

yes, but I feel you might be worrying about precision but asking about accuracy.

We will get it calibrated by Mettler Toledo when it gets to its final destination. There’s no point doing it now as it’s going to be relocated three times.

except when it arrives at the final destination and fails the test and you can't confirm it was ok when it left you using the same method. Use the same test they will use to accept it to release it. I presume you are using traceable calibration weights?

Have a look at any info you can find about gauge (or gage 🙁 ) capability. This essentially gives the variance/repeatability of the gauge. (Also known as a type-1 test, following this then a type 2 or type 3 test will be needed on-site at the end user)

You should be looking at the cgk over about 50cycles at least to infer capability. However the spec of 0.00048 is going to hard to agree to, was this agreed at POS or requested afterwards?

https://support.minitab.com/en-us/minitab/18/help-and-how-to/quality-and-process-improvement/measurement-system-analysis/how-to/gage-study/type-1-gage-study/methods-and-formulas/methods-and-formulas/

What does the scale manufacturer say? They should be certifying the kit and giving their test data

Edit:don't know

Lots of questions, cos I'm nosey rather than saying the existing method is wrong - are the scales currently on a test rig and won't be fully "calibrated" by Mettler-Toledo until the scales are finally installed in process? Why have you decided to go with measuring mass rather than level (using say a conductivity probe etc)? Do you really need .00 level of resolution when it's water feed for a decontam process? Will anyone be monitoring the scales indicator during the process? As suggested, you'd need to test the scales over their range, even though they may not be used over its full scale, so check them in X gram increments up to its maximum range and reduce back to zero - where X gram could be 50g, 100g or whatever. Keep the weight on for a while to see if the scales read steady - but another question would be, how fast is the decontam process and what's the response time of the scale indicator? It's for measuring 1kg / 1 litre of water, so why go for a 3 kg / 3 litre scale ? Unless it's going to be used to measure more in the future? It's not a critical item for measuring API's you mentioned, so does its "accuracy" matter when it's purely just indicative? I'd go back to the designers / process engineers and ask questions - did they suggest a test / calibration procedure? If the client wants the scales tested during FAT, what are their test and future calibration requirements in terms of GMP and what are their quality requirements on the scales? Is there a functional spec for the decontam process? This will hopefully give you some guidance.... Happy thinking !

Read USP <41> and <1251>

Linearity is a pretty useless test, it's uncertainty is the smallest cumulative value. So don't piss about putting a range of weights on.

Look at sensitivity, repeatability and corner loading errors from a practical point of view.

Also, product location is going to be critical to obtain the best minimum balance weight, but be wary of setting this as a pass fail criteria. What's your minimum expected weight, go for that with a safety factor of 2 (which will double the USP safety factor)

Mettler Toledo have an excellent white paper at MT.com on this.

Edit to say, minimum weight on a 3kg top pan is going to have a theoretical minimum of 8.3 or .83g depending on the readability. Maybe even 83 if it's not got much in the way of readability.

Edit, oh and that test you are doing with 10 replicates is a repeatability test. Used to calc minimum balance weight.

The first weighing of the sequence is your sensitivity (or accuracy in the terminology you are using). Need a traceable weight.

Get an M class weight, although the Clint would be happier with an F class (they are getting expensive so I'd get a vendor in to do it, should be about 300 quid).

As you already have a contact with Mettler I’d suggest asking them for a copy of their “ good weighing practice “ guide.
In an ideal world and under GxP regs you should probably be doing some sort of routine verification before use already, are you sure it’s not this that your client wants to see in practice, although the fact they are asking suggests that in a recent audit visit they saw something being weighed but had no evidence of the routine verification.

I would be checking what the specification is for the performance of the scale within the equipment it's integrated into, not the rated performance of the component that's been chosen from the market. Verify against what it needs to do, not what it could do in a different context.

It has a 3kg capacity but the equipment only needs a bit over 1kg, perhaps it has 0.016% accuracy but only only needs 1%.

The verification of the scale in context could be more difficult. Maybe the important criterion is that it's calibrated so that the scale is at zero when the bottle is empty, or at 25% when there's 25% left. Is the bottle refilled or replaced when empty - because not all bottles will weigh the same. Is the relationship between the reading on the scale and what that represents to the user correctly stated in relation to the density of H2O2?

Thanks so far peeps.

I'm still wrestling with this though!

are you sure? the rated capacity is 3000 g. So 0.016% of that is 0.48g

I wish that was the case.

In the spec, rated capacity is expressed in kg, not g. Therefore repeatability is given as %RC <=0.016 so it's 3/100 x 0.016 = 0.00048 g.

Which cannot be seen on the indicator or HMI.

Isn’t it 0.00048KG?

I've no idea, but a long time ago i used to make flight cases for these guys who do just that: https://www.budenberg.co.uk/

Mixing units up there. 0.00048kg = 0.48g

Well within the resolving capability.

Doomanic. I love you!

Thanks for that pal. I'll sleep tonight.

No worries. Pointing out other people's errors is my stock-in trade, just ask my co-workers... 😀