We have hosted a live webinar about best practice for the control of the melting process in foundries, together with the Foundry Planet. If you have missed it, you can watch the recording here.
From those who attended the webinar we’ve had many insightful questions. We thought that it would be extremely useful to publish the answers to these over the course of two posts. In this first post, Wilhelm Sanders, our Product Manager for OES here at Hitachi High-Tech, gives his answers to questions surrounding calibration. These are quite technical in nature, but so were the questions!
Q: When you measure any CRM (certified reference material) on your spectrometer, what are your acceptance limits?
Before I answer this question directly, I’d like to give a bit of background information on CRMs and why they are essential to OES.
Background
In simple terms, the OES detector measures the intensity and wavelength of light emitted after a localized plasma is created on the sample by the spark. The measured parameters detected by the instrument (such as voltage) are dependent on the concentration or amount of the substance being analyzed. However, without calibration, the measured intensity and wavelength doesn’t tell us much about the sample. To get meaningful results on chemical composition, the measured values must be compared with certified reference materials (CRMs). This allows the instrument to interpret the detected results and deliver a quantitative chemical analysis.
CRMs are huge databases of certified reference materials and must be used for accredited test and calibration laboratories involved in quality assurance. The CRMs, based on a huge number of samples and tests, create calibration curves for the elements that the OES user needs to measure.
So, now we understand the role of CRMs in measurement, we can go back to the original question.
Answer
Supplied CRMs do have an uncertainty associated with them. This is based on the variation of the results supplied by the testing labs which were involved in the certification process. Firstly, the calibration of the spectrometer should be done with as many CRMs as possible to reduce statical variation. The more standards, the more accurate the curve. As a guideline, the uncertainty of the calibration curve should not exceed ± 2SR, where SR is statistical reliability.
SR can be found with the following formula:
If when measuring standards, you find they are very different from the calibration curve, then that reason should be investigated. For example, has the wrong sample been loaded, or wrong method used?
Q: Are there any ASTM or DIN standards that support your answer to question 1?
All laboratories that are accredited to DIN EN ISO / IEC 17025 must have shown evidence of determining the uncertainty of all measurement procedures, methodologies and instrumentation to retain their accreditation. For OES, there are two practical ways of achieving this:
Q: What is the definiation of a CRM or standard?
This is the official definition taken from Bundesamt für Materialprüfung, a German provider of certified reference materials.
“A reference material is a material or substance of sufficient homogeneity for which one or more property values are sufficiently well established to be used for the calibration of measuring instruments, the assessment of measurement methods or for assigning property values. Reference materials are indispensable for ensuring the accuracy and reliability of measurement results.”
Q: Can bar stock be tested many times on multiple machines to create a check sample instead of buying an expensive CRM?”
The DIN 51008-2 standard published by the German institute for Standardization gives these guidelines:
To ensure the functionality of OE spectrometers, control samples must be used. The only requirement is that they are comparable in precision to the recalibration samples. Their composition should be determined by ‘linking to the calibration’, where they become reference samples themselves.
Spectrometer control samples differ from recalibration samples in several ways. For example, they are equivalent to the analytical samples or are low alloyed to ensure they lie on the calibration curves generated by the reference samples. Control samples are also larger than CRMs, which makes them more suitable for routine use in the spectrometer. And they are 2 -3 times cheaper than CRMs. This makes them more suitable for checking the functionality of the spectrometer, using CRMs for this task is too expensive.
Creating a reference sample from a control sample
Using control samples for checking the functionality of the spectrometer can be done with measuring light intensity or sample composition. The process for creating a control sample based on composition is as follows:
Calibrate the spectrometer with the reference samples. Before the spectrometer has a chance to ‘drift’, the control samples must be measured at least six times. This links them to the calibration created with the reference samples, turning the control samples into reference samples.
These control samples can be used to check the spectrometer in the following scenarios:
Note: method d is particularly recommended when the spectrometer is used to measure samples that are very similar. For example, only low alloy and CrNi steel, or GG and GGG.
The elements of the control sample have reasonable tolerance limits within the determined value.
Statistical evaluation of calibration values
As explained by SUS samples, a large supplier of standards, the calibration values obtained with recalibration samples and control samples should be recorded and then evaluated by statistical methods. This gives insight into the operational safety, actual stability of the instrument and the frequency of the recalibration. Only recalibrating when absolutely necessary saves material of the recalibration samples.
More details on this can be found in our cast iron guide.
Find out more:
If you have further questions on OES for your application, please get in touch. And if you missed our webinars on OES, you can watch them here.
Watch the recorded webinar
