Multivariate Statistical Software for Research and Development
CAMO multivariate statistical software products and solutions assist scientists in conducting Research and Development. Our software products have been used for the following Research and Development activities:
| Unscrambler Online | ||||||||
 Useful Links
|
- Chemical and Property Analysis
- Method development
- Microbiological Analysis
- Product Development
- Qualitative and Quantitative Analysis
Examples
Chemical and Property Analysis
In the production of decaffeinated coffee the manufacturer needs to know how much caffeine is still in the coffee to ensure product quality. For this purpose they must be able to tell whether the caffeine concentration is below or above 0.1 %. The usual way to determine the caffeine concentration is to use HPLC. This method is very accurate but has some disadvantages, e.g. it's very costly. The coffee company wants to replace this method with a method that is faster, easier and cheaper to use. A method that fulfills these requirements is Near Infrared Reflectance (NIR) spectroscopy. For these reasons the coffee company is interested in making a model, which can give the caffeine concentration of a given sample from the NIR measurements.
Method: Scatter Correction: Since the samples consist of ground coffee beans, we expect some scattering effects in the measurements due to different particle size and packing. We can either use Multiplicative Scatter Correction or take the second derivative of the spectra. MSC is a method designed to remove both additive and multiplicative noise effects in reflectance spectroscopy.
Classification: First we run a PCA model on all the samples, both the calibration and the validation set, to see if they span the same variation. Furthermore we want to see if it is possible to recognize the different coffee sorts, especially to see whether the pure coffee samples are different from the others. For validation of the PCA model the quick leverage correction method is used, since we are only interested to find the similarities of the samples.
Calibration: Now we are ready to make the PLS model from the spectra and the HLPC data. For this regression we use the test set validation. After a few seconds the model is ready and we use the ready-to-use model plots to examine the model named PLS 1.
Benefit of Analysis: By using The Unscrambler®, it was possible to first classify samples and then make a model that made it possible to predict the concentration of caffeine with very good accuracy. Outliers were detected using The Unscrambler® graphics that allow easy interpretation of data. We may find which samples and wavelengths are the most important for the model. From these results it is easy to optimize the model to get the best prediction ability. Furthermore it is simple to select fewer wavelengths that can be used with a filter instrument.
Qualitative & Quantitative Analysis
Perrigo lactose classification with NIR measurements
Situation: A set of NIR spectra were measured for both lactose anhydrate and monohydrate forms. Spectra have a large variation in intensity though they are from same chemicals. Need to classify two forms based on their spectra.
Need: The objective is to see if The Unscrambler® is able to create a model (PLS .cal file) used with AIRS 3.0` for discriminate analysis purposes (raw material Identification). The current potential use of The Unscrambler® would be for raw material Identification methods. Unscrambler
Solution: Data pre-processing spectral normalized by mean is necessary in this case. Both SIMCA and PLS DA work extremely well.
