Assuring the Quality of Test and Calibration Results - ISO IEC 17025 Clause 5.9

suildur

Involved In Discussions
I am working to build an integrated QMS for a calibration lab and an inspection body which are established under the same company.

The proposed accreditation scope of the calibration lab is as below:
1. Heat cells
2. T/Cs
3. Pressure (up to 75 bar)
4. Non-automatic weighing scales (up to 150 kg)

The inspection body's scope is inspection of clean rooms and zones according to ISO 14664-1, -2 and -3.

Anyway, I have constructed a QMS documentation as belows:

1. QM
2. Procedure 1 - Control of documents and Records
3. Procedure 2 - Sales
4. Procedure 3 - Purchase
5. Procedure 4 - Customer Relations
6. Procedure 5 - Corrective and Preventive Action
7. Procedure 6 - Control of Nonconforming Operation
8. Procedure 7 - Management Review
9. Procedure 8 - Internal Audit
10. Procedure 9 - Training
11. Procedure 10 - Quality Assurance
12. Procedure 11 - Clean Room Inspections and Tests
13. Procedure 12 - Calibration
14. Procedure 13 - Control of Data
15. Procedure 14 - Control of Reference Equipment

The system structure is based on ISO 9001 plus 17020's and 17025's additional requirements.

So... My questions to 17020 and 17025 experts are as follows (begining from the most important):

ISO IEC 17025 clause 5.9 Assuring the quality of test and calibration results requires a procedure. The requirement states (in short);

5.9.1 There must be a quality control procedure where data are recorded, and, statistical methods are applied for the review of the results. That monitoring must be planned and reviewed. Monitoring can contain,

a. Building of internal quality control system by using certified reference materials,
b. Joining inter-lab test programmes, conducting proficiency tests etc.
c. Re-conducting of tests or calibrations with the same or different methods,
d. Re-conducting of tests or calibrations on the preserved material,
e. Research on the different specs of the material.

5.9.2 Quality control records/data must be analysed and planned operations are applied to correct the problem or to prevent reporting of nonconforming results when data are out of limits.


Okay, I clearly understand 5.9.1. I will write down the method for proficiency tests (where two of our employess will conduct the same calibration) and joining inter-lab test programmes (where our AB, TURKAK, or any other accredited laboratory plans a test programme we will join, too), however, the problem is about 5.9.2.

I understand there must be analysis if PT/ILC demonstrates a nonconformity. But, I have also seen some statistical methods (t-distribution, f test, z score, q test, chi square test, x cards) in a sample procedure.

So, shall I apply statistical methods on client's calibration results?

I mean,

1. Shall I apply statistical methods on each calibration step of an individual calibration of any client? For example, if we calibrate a T/C and use 10 calibration points, so I have 10 deviation values and an uncertainty value, shall I apply t-distribution on 10 deviation values?

2. And, shall I compare the uncertainty value to similiar T/C calibrations which we have conducted before?

Thanks in advance for answers.

P.S. I have checked the thread Compliance with ISO/ IEC 17025:2005 Clause 5.9 for Calibration Laboratories That doesnot help me, I think.
 

suildur

Involved In Discussions
Arent there any answers? S.O.S!!!

:)

Well, maybe kidding, but I really need some comments on this. Any comment or information is highly appreciated.
 

BradM

Leader
Admin
Well... I'm certainly not an expert here. But I will certainly try.

In general, I interpret that section that you have adequate quality control over the records. This might be 100% inspection, or may involve using SPC and control charts. Basically you have assurance to detect if something is out of specification, and a plan to deal with it when it is discovered.
 

suildur

Involved In Discussions
Well... I'm certainly not an expert here. But I will certainly try.

In general, I interpret that section that you have adequate quality control over the records. This might be 100% inspection, or may involve using SPC and control charts. Basically you have assurance to detect if something is out of specification, and a plan to deal with it when it is discovered.

Yes, maybe not an expert answer, but at least that's an answer.

Thanks for your kind reply...
 
M

Mgrace

Dear Suildur,

When reporting the uncertainty, t distribution is really needed because the uncertainty will be express like this....>>>> example : The uncertainty is ± 0.01°C with a coverage factor K=2 at 95% confidence level. "K=2 at 95% is from t distribution.

That is my idea and I'm not expert for that.
 

suildur

Involved In Discussions
Dear Suildur,

When reporting the uncertainty, t distribution is really needed because the uncertainty will be express like this....>>>> example : The uncertainty is ± 0.01°C with a coverage factor K=2 at 95% confidence level. "K=2 at 95% is from t distribution.

That is my idea and I'm not expert for that.

Yes, you are right, but, right for the expanded uncertainty :eek: As you probably know, every calibration method has its own method of uncertainty calculation, and, those are determined by internationally recognized reference documents if you are using standard referenced methods. After calculating the uncertainty you expand it by multiplying the uncertainty result with two. But, that's something different than what I was asking.

I asked, if I must apply any statistical method to check whether the distribution of the sample space (calibration step results) is meaningful or not, each time I calibrate something. I got an answer from another lab, that there is no need to apply those statistical methods for the calibration.

Thanks for your kind interest.
 
Last edited:
A

aweinerth

Suildor - Do you have a draft procedure that you can share for 5.9 - Assuring the Quality of Test Results? I have a calibration lab and in process for accreditation and would like to benchmark off of something that is already out there. Thanks in Advance.
 

suildur

Involved In Discussions
I do have it while I have already written it. But, it is in Turkish.

However, I would like to answer your questions for whatever you need.
 
Top Bottom