Maciej Socjusz is an application specialist at FOSS and has worked with calibration of NIR instruments for feed applications for many years. Here are his six steps to a fast and successful implementation.
1) Start by ensuring online access to measuring results and reference data
Ensure that the instrument can be connected to networking facilities, for example FossConnect™. This provides access to and from a support person who can work quickly and without disruption to daily operations.
2) Make the validation before starting to use the NIR instrument
It may sound obvious, but if overlooked, it will delay the implementation. Being able to see the full dataset acquired during the validation allows the specialist to make relevant statistical analysis to decide, for example, what requires a bias adjustment and what are outliers and mistakes.
3) Get off to a good start with timely advice
It is important that the application specialist is present at or immediately after installation otherwise users may get into bad habits when sampling and measuring which will be difficult to reverse.
4) Focus your sampling on the critical raw materials and parameters
According to ISO, a minimum of 20 samples must be used to validate a calibration. If the calibration includes multiple sample types (like all the vegetal ingredients) and up to six parameters, 20 samples will only cover a few sample types. It is therefore recommended to start with the most critical raw materials and parameters. Equally, it is important that the samples cover the variation of the parameters.
5) Use the instrument feedback to select samples for reference analysis
When using the calibrations, the FOSS instrument will warn you of outliers when there is an uncertainty about the validity of the results. Being an outlier means that the sample type is not represented in the calibration or it is simply a wrong sample. If it is not represented in the calibration, the relevant samples should be measured by reference chemistry to test if the NIR results are correct. If there is a sufficient number of outliers, FOSS will create hotfix of the calibration. If the validation shows issues for individual samples or parameters, the global calibrations may need to be adjusted based on the samples.
6) Continue to do random sample checks during steadystate operations using network facilities
After the initial adjustment, the Calibration will be stable and it only needs to be validated by a random sample set once a year. Once again, remote access is a fantastic tool to help the application specialist to differentiate between instrument, operator or calibration issues. Likewise, uploading and reformatting data is performed quickly as a matter of routine without interrupting the work of operators. The regular check ensures that the instrument is always
up to date, so should an actual problem occur, it is then easier and quicker to identify and fix it instead of starting with a retrospective analysis of performance.