Use of control charts to monitor laboratory balance performance
- Thread starter nesrine
- Start date
There are many of us here who are proficient at control charts. What is your first question? (The more specific you are the better we can help you, including why you think control charts will help you / what problem are you hoping they will solve?)
Hello, thank you for replaying and for offering help, I want to use control charts as a mean of ensuring that my analytical and precision balances are under statiscal control and that any variation in the weighing is due to random causes. How I want to do it: simply by pick an object with a constant weight (in my case i'm using a standard weight (not calibrated) as it's not a metrological chart but a statistical one) weighing it 30 times minimum and use the data to draw the central line and both control and surveillance limites., each value is the result of one weighing ( I'am not using duplicates) and even after the chart is drawn i want the technician to simply weigh the standard weight (once) report the result on the chart immediatly and be able to decide if he can go on with his work and use the balance. now my questions are:
- Which standard might I use that covers this type of chart ( and if I am being honest I'm a bit confused with the types of charts) : I have the ISO 7870 and I'm not sure if it's "individuals control chart" or " Acceptance control chart". please let me know which type is best for a case like mine and please if you could suggest other reliable standrds please do so.
- I tried the individuals control chart but I couldn't go on as the data should follow a normal distribution, but mine did not and so I stopped there I did some research and found out you could statistically intervene but the statistics were above my pay grade and I need to be able to explain it in an audit (I can send the data or the excel file if someone would kindley check them for me).
- My last problem is metrological not statistical and it's regarding which standard weight should I use: at first I thought I should use the one closest to what the balance routinely weighs ( the balance is used for the same weighing evey day) so for weighing elements around 0.02 grams and 0.5 g so I chose small weights but after that I asked experienced people who kept giving different advices the first said no you should used 50% of the maximum weight your balance can weigh the second said you shoud maximize the error and the last one said no you were correct in the first place so now i'm confused again and I welcome advice regarding this point.
I hope I was able to clearly explain please let me know of further explanation or information I need to provide.
Yes a I, MR (aka X, MR, individuals chart) is appropriate for this use.
But first you need to understand - and accept - that control charts do not rely on or require a Normal distribution*. This is a myth - worse than the myth of the Yeti or Nessie. So you don’t need to test for Normality. You can read my paper “Theoretical Distributions and False Alarm Rates in Control Charts” in the resources section.
I would not recommend taking 30 sequential readings in a short period of time. 20-25 single readings over a period of time representing the times you would take your monitoring readings is the best (this is a Rational Subgrouping thing). The problem is that if you take 30 sequential readings in a very short period time you are not incorporating the variation that could naturally occur during your monitoring time periods.
If you would like to read more about these topics you can check out my “Essential References for Quality Engineering” in the resources section - there are numerous links to free articles and other references form pre-eminent experts in the topics. (Rather than random colleagues or google searches - including AI - that are not properly educated and experienced in control charts and other Quality Engineering methods)
There are no reliable ‘standards’, ISO or otherwise that will help you with control charts. They are all mediocre pablum. Check out Donald Wheeler’s works at SPCPress.com
As for which weight to use - the responses you are getting are metrology /calibration type responses. I would just use 2-3 typical weights that are used. You might find that a specific weight is ‘the canary in the coal mine’ so you should start by trying several weights. Physics is better understood through well structured study designs; in this case it’s a set of I, MR control charts.
Wheeler, Donald, “The Chart for Individual Values”, http://www.agileleanhouse.com/lib/lib/Topics/ProcessBehaviorCharts/TheChartForIndividualValues.pdf
Wheeler, Donald, “Individual Charts Done Right and Wrong”
Wheeler, Donald J., “A History of the Chart for Individual Values, The ultimate in homogeneous subgroup”, Quality Digest July 2017 https://www.spcpress.com/pdf/DJW317.Jul.17.History of XmR Chart.pdf
But first you need to understand - and accept - that control charts do not rely on or require a Normal distribution*. This is a myth - worse than the myth of the Yeti or Nessie. So you don’t need to test for Normality. You can read my paper “Theoretical Distributions and False Alarm Rates in Control Charts” in the resources section.
I would not recommend taking 30 sequential readings in a short period of time. 20-25 single readings over a period of time representing the times you would take your monitoring readings is the best (this is a Rational Subgrouping thing). The problem is that if you take 30 sequential readings in a very short period time you are not incorporating the variation that could naturally occur during your monitoring time periods.
If you would like to read more about these topics you can check out my “Essential References for Quality Engineering” in the resources section - there are numerous links to free articles and other references form pre-eminent experts in the topics. (Rather than random colleagues or google searches - including AI - that are not properly educated and experienced in control charts and other Quality Engineering methods)
There are no reliable ‘standards’, ISO or otherwise that will help you with control charts. They are all mediocre pablum. Check out Donald Wheeler’s works at SPCPress.com
As for which weight to use - the responses you are getting are metrology /calibration type responses. I would just use 2-3 typical weights that are used. You might find that a specific weight is ‘the canary in the coal mine’ so you should start by trying several weights. Physics is better understood through well structured study designs; in this case it’s a set of I, MR control charts.
Wheeler, Donald, “The Chart for Individual Values”, http://www.agileleanhouse.com/lib/lib/Topics/ProcessBehaviorCharts/TheChartForIndividualValues.pdf
Wheeler, Donald, “Individual Charts Done Right and Wrong”
Wheeler, Donald J., “A History of the Chart for Individual Values, The ultimate in homogeneous subgroup”, Quality Digest July 2017 https://www.spcpress.com/pdf/DJW317.Jul.17.History of XmR Chart.pdf
Good question from the Metrology standpoint.
Do you have the manufacturer / model of the balance so we can look up specifications, or can you at least provide the range and resolution of the balance to see if you are taking valid readings.
Similar threads
