There are
a lot of terms that means the same,
pre-calibration or NIR starter calibration or pre-built calibration or pre-installed calibration orcalibration package or pre-developed calibrations or pre-calibrated NIR or global calibrations or nir global calibration package or factory calibrations or universal near-infrared (NIR) calibrations or local calibrations or ready-to-use NIR calibrations or off-the-shelf calibrations or factory-calibrated or pre calculated model or start-up calibrations or calibration equations or
prefabricated nir calibrations or calibration library or mathematical model.
That are Calibration models that are prepared and developed by a calibration specialist. They have
collected a lot of samples
over years and measured them with NIR and analyced it with
reference methods.
The NIR spectra are then
calibrated against the reference values. This is called a NIR calibration or calibration model or sometimes calibration curve or calibration equation.
Normally a precalibration is delivered as a file that is compatible to the used NIR analysis software. Such a calibration file does not contain the spectra nor the reference values.
So how can that work?
The only thing that is in the file is a description what it is for (e.g. protein in feed) and the chemometric model that is represented and stored as list of vectors and matrices.
You can’t visualize them, it’s a
black-box file. You have
no insight of how the calibration is done, how are the settings, how is the prediction performance.
You can not extend the calibration with your data to
adjust it to your purpose or specialty.
Most often the pre calibration
files are protected, so you can use it only with a
paid license to your software or even to your instrument serials number.
These are some (not well known)
limitations you will discover if you got one.
But such
starter calibrations are
very useful to have
a fast and easy start with a new NIR spectrometer. That’s the main reason why pre-calibrations are available. The second reason is that a collection of spectra can be reused to build such pre calibrations.
Predicting the future?
Are very old spectra useful to predict the future? To adjust a calibration model with newly collected data,
the calibrations grows and contains more and more
redundancy.
That means there are very similar spectra with the same concentration range.
So
which spectra
can be removed to make the calibration better? You maybe never ask this because often you hear, that the more spectra you put into a model the better it will be.
Why to remove some spectra?
- reduce not needed redundancy
- makes the calibration smaller and less complex
- makes the calibration better fit to the current situation of now and the near future
- remove long past seasonal data if you have natural products because nature is changing
- and of course bad outliers should be removed
Custom NIR calibrations
Build your own calibrations that
perfectly fit to your specific sample matrix of your products and your preferred raw materials from your local suppliers.
Nature grows differently depending on the
geographical region, by
seasons and year by year. As you know that
NIR-Spectroscopy is not an absolute method, then you have to think about to
calibrate these current changing effects into your models.
If you own the spectra and the reference values then your are able to
build your own calibration models and
re-calibrate them when needed. So you have the
full control on Calibration updates (also known as moving models).
Conclusion
A NIR-instrument can
only measure NIR spectra. So
the usefulness of NIR comes in with calibrations. That is
very important to know when buying such an instrument. For a fast start you can use pre-built calibrations. Good reliably calibrations are offered from
third party to
quite high prices that level is similar to a cheaper NIR-Instrument!
To continue successfully it is
highly recommended to develop your own customized calibration (multivariate calibration model) with your own data from your own products, especially with the use of natural resources. Therefore you need knowledge about chemometrics and multivariate analysis (MVA), spectroscopy and the software used to get the calibration optimized.
It is worthwhile to create your own calibrations, because you can
calibrate product characteristics that are not covered by the proposed pre-calibrations.