DE102007027569B3 - A method for qualitatively classifying surfactant-containing compositions - Google Patents

Abstract

The present method, by combining suitable spectroscopic techniques, such as NIR spectroscopy, to obtain analytical measurements in the form of raw spectral data and multivariate methods for numerically-mathematical processing of raw spectra data and correlating the processed data with pre-calibrated classes, enables qualitative classification of surfactant-containing compositions at low cost and time.

Description

The The present invention relates to a method for qualitative classification surfactant-containing compositions according to the preamble of the claim 1.

Under Surfactant-containing compositions are in particular washing and cleaning agents to understand, that is, any compositions and products used for cleaning objects or surfaces are intended or intended to assist their cleaning, but also Personal care products, this means, any compositions and products for cleaning the body.

Either for the Household as well as industry are a variety of surfactant-containing Products available, such as detergents, fabric softeners, hand and machine dishwashing detergents, Cleaning agent for cleaning surfaces, machines or vehicles and personal care products, such as shower gels, hair shampoos and bath products.

So according to the Federal Environment Agency were in terms of detergents and cleaners At the end of 2000, a total of around 54,000 products from around 4,500 companies were sold the German market. In 1998, more than one million tons Detergents and cleaners consumed in German households, including alone about 800,000 tons of household detergent.

The individual products typically contain a variety of components, depending on the purpose of the product from an extensive pool selected suitable ingredients become.

So can in detergents, among others: alcohols, bitter substances, Bleaching agents (such as perborate, percarbonate or hydrogen peroxide), Bleach activators (eg tetraacetylethylenediamine (TAED)), citrates, Fragrances / perfume oils, enzymes, dyes, Lubricants, fillers (usually sodium sulfate), preservatives / biocides, corrosion inhibitors, alkaline agents (such as sodium carbonate / soda), optical brighteners, Phosphates, phosphonates, polycarboxylates, soaps, silicates, surfactants (anionic, cationic, nonionic and amphoteric), discoloration inhibitors, wash strips and zeolites (to prevent limescale).

In Cleaning agents can including: abrasives, adsorbents, Alkalis, biocides, chlorinated hydrocarbons, fragrances, enzymes, Dyes, hypochlorite, complexing agents (eg citrates, phosphates, Phosphonates, carboxylates, EDTA or NTA), preservatives, Solvents, oils and fats, Polymers, organic and inorganic acids and their salts, surfactants, Waxes and softeners.

Orderly according to the consumption quantities are the most important ingredients in Detergents and cleaners: Surfactants, zeolites, sodium carbonate (soda), sodium sulfate, sodium perborate (Tetrahydrate), sodium percarbonate, phosphates, silicates, alcoholic Solvent, Polycarboxylates, sodium citrate, tetraacetylethylenediamine (TAED), Enzymes, fragrances, phosphonates, carboxymethylcellulose, optical brighteners, Polyvinylpyrrolidone (PVP), nitrilotriacetic acid (NTA) (Source: Industrieverband Body care and Detergent e. V. (IKW), September 2000).

Consequently For detergents and cleaners, the surfactants and the so-called builders (builders such as zeolites, soda and, to a lesser extent, phosphates) the most quantitatively significant Ingredients.

In Personal care products can including: abrasives, adsorbents, Antioxidants, chelating agents, fragrances, emulsifiers, dyes, Film former, gelling agent, preservative, solvent, buffering agent, foaming agent and regulators, surfactants, viscosity regulators and various herbal or synthetic active ingredients for cleansing and care the areas to be treated with the product such as skin, hair or Teeth.

at the production of such surfactant-containing products is in the course the manufacturing process from the raw materials to the finished packaged product Again and again the need for individual batches of intermediate or Check final products for your identity and / or quality.

by virtue of This product variety is usually very time-consuming and expensive Analytical methods connected. In general, this will be chromatographic and / or spectroscopic methods used, such as gas chromatography, liquid chromatography, Ion chromatography, mass spectroscopy or absorption spectroscopy. To identify individual ingredients are for separation disturbing Components about it In addition, often elaborate sample preparation steps necessary.

So will in the DE 10 2004 009 306 A1 discloses a method for the analysis of surfactants with polyethylene glycol groups. For this purpose, the surfactant is treated with a stripping gas to produce a gas phase which contains volatile compounds contained in the starting material, which are then analyzed in a gas measuring cell by means of infrared spectroscopy. But this is just the analysis of impurities contained in surfactants possible.

In contrast, the describes DE 195 38 932 A1 a fiber optic sensor for the detection of surfactants in liquids. The sensor is surrounded by the liquid to be measured so that upon irradiation of UV light into the sensor and the presence of UV absorbing surfactants in the liquid to be measured, the light emerging at the other end of the sensor has a characteristic change. This provides a fast and continuous measurement method for UV-absorbing surfactants.

A device which can also be used in surfactant production to ensure product quality is used in the DE 199 45 392 A1 disclosed. Thus, among other things, the determination of surfactant concentrations in aqueous solutions is possible. The measuring principle used in the device is based on the fact that certain dyes change their fluorescence properties as a function of the concentration of surfactant in the aqueous solutions, so that the measurement of the fluorescence of the dye indirectly gives an indication of the concentration of surfactant in the solution.

This Depending on the type of surfactant to be tested, the addition of a suitable selected dye to the sample to be examined.

task The present invention therefore provides a fast and reliable Process for the qualitative classification of surfactant-containing compositions. This qualitative classification includes both an identification of a individual product as well as a review of the quality of the product with regard to a consistent composition. Under product here are both intermediate and also to understand end products.

The Task is solved by the characterizing features of claim 1.

• a) Bereitstellen einer Mehrzahl an Klassen tensidhaltiger Zusammensetzungen mit jeweils einer Mehrzahl an Kalibrierungs-Materialproben der tensidhaltigen Zusammensetzungen pro Klasse mit einer Mehrzahl bekannter Klasseneigenschaften;
• b) Einstrahlen eines Messlichts aus einem vorgegebenen Wellenlängenbereich auf die Materialproben der tensidhaltigen Zusammensetzungen;
• c) Erfassen des durch die Materialproben der tensidhaltigen Zusammensetzungen hindurch tretenden, reflektierten, reemittierten und/oder gestreuten Messlichts;
• d) Bestimmen des wellenlängenabhängigen Verhältnisses von eingestrahltem zu erfasstem Messlicht (Spektrum) für jede Materialprobe der tensidhaltigen Zusammensetzungen einer Klasse;
• e) numerisch-mathematische Aufbereitung der Spektrendaten der einzelnen Materialproben der tensidhaltigen Zusammensetzungen;
• f) Korrelieren der Spektren einer Mehrzahl von Materialproben der tensidhaltigen Zusammensetzungen mit einer vorgegebenen Klasse der tensidhaltigen Zusammensetzungen;
• g) Erstellen einer Datenbank aus den aufbereiteten Spektrendaten mit unterschiedlichen Klassen von tensidhaltigen Zusammensetzungen, basierend auf den gemes senen Materialproben der tensidhaltigen Zusammensetzungen der einzelnen Klassen zur Kalibrierung einer Klassenkorrelation;
• h) wenigstens einmaliges Wiederholen der Schritte b) bis e) mit wenigstens einer Materialprobe einer tensidhaltigen Zusammensetzung mit wenigstens teilweise unbekannten Eigenschaften; und
• i) Bestimmen der Klassen von tensidhaltigen Zusammensetzungen, zu welcher die unbekannte Materialprobe einer tensidhaltigen Zusammensetzung zuzuordnen ist, mittels einer Klassenkorrelation der gemessenen Spektren unter Verwendung der erstellten Kalibrierungsdatenbank aus Schritt g).

In particular, the method according to the invention for qualitatively classifying surfactant-containing compositions comprises the following steps:

• a) providing a plurality of classes of surfactant-containing compositions each having a plurality of calibration material samples of the surfactant-containing compositions per class having a plurality of known class properties;
• b) irradiating a measuring light from a predetermined wavelength range on the material samples of the surfactant-containing compositions;
• c) detecting the reflected, re-emitted and / or scattered measuring light passing through the material samples of the surfactant-containing compositions;
• d) determining the wavelength dependent ratio of irradiated to sensed measurement light (spectrum) for each material sample of the surfactant-containing compositions of a class;
• e) numerical-mathematical preparation of the spectral data of the individual material samples of the surfactant-containing compositions;
• f) correlating the spectra of a plurality of material samples of the surfactant-containing compositions with a predetermined class of the surfactant-containing compositions;
• g) creating a database of the processed spectral data having different classes of surfactant-containing compositions based on the measured material samples of the surfactant-containing compositions of the individual classes for calibrating a class correlation;
• h) at least one repetition of steps b) to e) with at least one material sample of a surfactant-containing composition having at least partially unknown properties; and
• i) determining the classes of surfactant-containing compositions to which the unknown material sample is to be assigned to a surfactant-containing composition by means of a class correlation of the measured spectra using the created calibration database from step g).

With the method according to the invention it is through the combination of suitable spectroscopic methods to obtain analytical measurements in the form of raw spectral data and multivariate methods for the numerical-mathematical preparation of the Spectra raw data and correlation of the processed data with advance calibrated classes, possible for the first time surfactant-containing compositions at low cost and time to classify qualitatively.

A numerical mathematical processing of spectral data of wine and coffee is used in the DE 101 24 917 A1 described. However, the identification of the spectroscopically examined drinks was achieved by the variations in their compositions due to the natural substances, in particular seasonal variations and differences in origin, variety, processing and storage.

For the purposes of the present invention, the term "class" is understood as meaning a group of surfactant-containing compositions having defined properties, ie class properties, whereby a class can also be defined so narrowly that it contains only a single surfactant-containing pro so that the product can be identified by its affiliation to that class.

For example, as "class properties" the name of the manufacturer, name of the product, intended use (eg. B. heavy duty detergent).

Consequently For example, a class can be defined so narrowly that it is only a specific one Product of a manufacturer includes, but can also be broader be, z. B. all shower gels of a manufacturer, if all products In this class, they have their own common characteristics which are their own Demarcation opposite enable other classes and thus their identification.

to Calibration of the method according to the invention be first of a plurality of surfactant-containing products as calibration material samples, of which a plurality of class properties are known, respectively measured a spectrum. The measurement is usually repeated several times, to buffer statistical fluctuations. Usually, at least 10 samples / class property used. This data will be then numerically-mathematically processed to reduce the amount of data and to focus on the essential characteristics of the spectra.

For the procedure The present invention is basically any photometric Suitable measuring method, with the reproducible and meaningful spectra, what changes reflect in the composition of surfactant-containing products, can be obtained.

According to the present invention, it is preferable to include a spectrum in the near-infrared region (NIR spectrum) of the surfactant-containing products to be classified, as a rule no further sample preparation is necessary. For this example, a commercially available NIR-VIS spectrometer can be used. These work at wavelengths from 700 nm to 2500 nm, ie at wavenumbers from 14.300 cm ^-1 to 4000 cm ^-1 .

through Near-infrared spectroscopy (NIRS) can generally be H-C-, H-O-, Recorded H-N and H-S bonds and thus above all a multitude of organic ingredients. The fundamental vibrations of these groups are in the mid-infrared range, the first harmonics in the range of about 1550-2000 nm and harmonics higher Orders and combination vibrations in the range of 800-1600 nm.

With NIRS can both solid and liquid Samples are measured. For solid samples, the diffusely reflected Radiation (remission) measured. Liquid samples can be both in transmission as well as in transflection, d. H. through diffuse reflection after passing through the liquid Medium, to be measured. In the Transflexion is behind the zu measuring sample arranged a reflective surface, for. B. one Teflon surface at which the irradiated measuring light is reflected diffusely and thus Run through the sample twice.

to better reproducibility of the method according to the invention, both the calibration material samples as well as those to be classified Material samples at a constant temperature, preferably at about 23 ° C, measured.

Depending on the photometric measurement method used, the raw spectral data obtained reflect one or more of the following exemplified properties of the material sample: carbonate content, phosphate content, silicate content, content of oxidizing substances, content of anionic, cationic, amphoteric or nonionic surfactants, content of alkylpolyglycosides, content of C _X sulfonates.

The numerical-mathematical preparation of the raw spectra of data can with a commercial one Software, eg. B. BCAP V 6.00 [BÜHLER AG, ANATEC, CH-9240 Uzwil, Switzerland].

The Class correlation can also be done with a commercial one Software, for example Nircal 3.0 [Buchi AG, CH-9230 Flawil 1], z. B. by means of principal component analysis and clustering. The The result can be, for example, in the form of a cluster image as 3D plot where the axes are the main components represent.

there the NIRS method is correlated with the measured samples in such a way that multivariate methods like principal component analysis, clustering, artificial neural networks are applied to this processed data, to be able to make a statement from the flood of data, whether an unknown, also NIR-spectroscopically measured material sample a surfactant-containing composition to a particular class belongs or not.

Under multivariate methods are understood as evaluation methods that are more as only one measurement signal of the same sample to arrive at an analysis result get. These methods include u. a. multilinear regression (MIR), principal component analysis Principal Components Analysis (PCA), Principal Component Regression (PCR = Principal Components Regression), method of the partial smallest Squares (PLS: Partial Least Squares), clustering methods and artificial Neural Networks.

For the artificial neuronal networks, in particular the following algorithms are considered: backpropagation networks, Dynamic Learning Vector Quantization (DLVQ algorithm), Radial Basic Functions (RBF networks), especially with Dynamic Decay Adjustment algorithm (DDA algorithm) trained RBF networks (RBF-DDA network).

The PCA leads a division of the original Data matrix in two matrices, the factor values and charges (Loadings) are called. In the original data room is a Vector so chosen that with a projection of the data on it, the largest possible part the variance is mapped. This vector is the first major component. Orthogonal to the first main component is a second and optionally one orthogonal to the first and second main components, a third one Main component created, with the second and third main component preferably much of the first and the first and second main components to illustrate not yet described variance.

The Coordinates along the first major component contain the essential ones Information of the data, the second and third main component essentially the scatter again.

This The process is repeated until either the number of major components which corresponds to the dimension of the output data or a specific termination criterion is reached.

The thus obtained main components are linear combinations of the original ones Dimensions. They are linearly independent of each other, creating a defined number of main components less redundant information contains as the same number of output variables.

Further describe the main components thus obtained as possible much of the variance of the output data from the main components present not yet described. This usually causes that the first three to five main components reproduce the essential part of the information in the data record.

Mathematically seen, the principal component analysis is an eigenvalue problem whose fundamental solution the expert is known.

The The result of the principal component analysis is thus a transformation of the N-dimensional original Dataroom, which states that the first dimensions are the essential, strongly contributing to the total variance contributing data parts and the last dimensions practically only reflect the noise component. In this way, the structure of the spectroscopic data in question is by plotting the first main components against each other representable. They then stand as two-dimensional, preferably 3D images For example, for visual evaluation by the user, the let it be remains to choose such a representation.

at the calibration can then the so-called tolerance circles of the illustrations are chosen so that they can be adapted to specific classes as needed to: to facilitate the classification.

Based on the measured calibration material samples of the surfactant-containing Compositions of each class is used to calibrate a Class correlation a database from the processed spectral data with different classes of surfactant-containing compositions created.

Under Using this calibration database, it is now possible to use the Material sample of a surfactant-containing compositions with at least partially assign unknown properties to a class.

Further preferred embodiments The present invention is defined in the subclaims.

When surfactant-containing composition according to the present invention is preferably a detergent, especially a heavy duty detergent, Color detergent or detergent concentrate, a hand dishwashing detergent, in particular a normal or a balm hand dishwashing detergent, a Geschirrspültab, in particular normal, 3 in 1 or 4 in 1 dishwashing tablets, a detergent for household or industrial use for cleaning surfaces, machines or vehicles, in particular a floor cleaner, a car wash or Aircraft detergent, or a surfactant-containing personal care product such as a shower gel used.

The wavelength range used in the method according to the invention is preferably in the NIR-VIS range, in particular in the range of about 700 to 2500 nm (4000 to 14,300 cm ^-1 ), more preferably in the range of about 1000 to 2200 nm (4500 to 10,000 cm ^-1 ).

To the Irradiating the measuring light onto the material sample and / or for recording after a reflection, transmission or transflection of the sample emitted radiation, an optical waveguide can be used.

The area of the material sample that interacts with the radiated measuring light, ie the optical layer thickness of the material sample, is preferably between about 0.2 and 5 mm, more preferably about 0.5 or about 3 mm.

at the method according to the invention includes the numerical-mathematical processing of the spectral data preferably a data reduction, in particular by: normalization, Smoothing, especially after Savitzky Golay, 1st derivative, 2nd derivative, multiplicative Scatter Correction (MSC), Invert, Square, Mean Centering, Kubelka Munc transformation, absorption, baseline correction, addition a constant, Shift Negative to Zero, principal component analysis (PCA), Principal Component Regression (PCR) or Regression by Method least squares (PLS).

The Correlation of numerically-mathematically processed spectral data preferably includes multivariate methods. Preference can be here a main component analysis (PCA) and / or a Smoothing, in particular according to Savitzky Golay, and / or at least one series development, in particular a Taylor series development and / or artificial neural network algorithms, in particular backpropagation networks, dynamic learning Vector Quantization (DLVQ algorithm), Radial Basis Functions (RBF networks), in particular with the Dynamic Decay Adjustment Algorithm (DDA algorithm) trained RBF networks (RBF-DDA network).

Further Preferably, the correlation is done by clustering. In the Calibration can also the tolerance ranges of the individual clusters are set.

at the method according to the invention are further per surfactant containing a measured material sample Composition recorded about 3 to 20 spectral scans. Further For calibration, preferably at least about 10 material samples the surfactant-containing compositions used per class property.

there are preferred about 70% of all Per sample of measured material samples of surfactant-containing compositions for calibration and about 30% used to validate the procedure.

A advantageous graphic representation of a received with the inventive method Classification result is a 3D plot on a screen.

Additional benefits and features of the present invention will be apparent from the Description of exemplary embodiments as well as from the drawings.

It demonstrate:

1 : Original spectra of 11 different heavy duty detergents taken in diffuse reflection mode.

2 : Pretreated spectra of raw spectral data 1 ,

3 : Cluster image as a 3D plot of the heavy-duty detergent products.

The inventive method for qualitatively classifying surfactant-containing compositions will now be based on various commercially available heavy duty detergent products illustrated.

For this purpose, with a polarization interferometer (Büchi AG, Switzerland, optical resolution: 12 cm ^-1 , relative reproducibility: 0.5 cm ^-1 ) crosswise in a wavenumber range of 4000 to 10,000 cm ^-1 in the transflection mode of 11 different commercially available heavy duty detergent products NIRS Spectra taken using an optical glass fiber (silica glass, Infrasil, Bes Optics Inc., Warwick, UK, length 2 m, optical thin film 0.5, 1.0, 1.5, 2.0, 2.5, 3, 0 mm) was used. In order to largely compensate for inhomogeneities, 10 scans were performed to obtain averaged spectra.

The original spectra of the 11 different heavy-duty detergents (raw spectral data) are in 1 displayed. The pretreated spectra of the 11 heavy-duty detergents with the most important vibrations are in 2 displayed. The following products were used as heavy-duty detergents: VWM1: Clever; VWM2: Omo; VWM3: heavy-duty detergent from Lidl; VWM4: Bravil; VWM5: White giant; VWM6: heavy duty detergent from Schlecker; VWM7: Splendid; VWM8: Tandil; VWM9: Potzblitz; VWM10: aliomat; VWM11: Sentimat.

To create a qualitative model, the chemometrics software NirCal 4.0 (Büchi AG, Switzerland) was used. The software allows selection of spectra and wavenumber range and mathematical pretreatment of spectra and statistical analysis using cluster analysis, principal component analysis (PCA) and least squares regression (PLS). The samples were thermostated at 23 ° C before recording the spectra in a water bath (model PC / 4, Julabo, Seelbach, Germany). The recorded spectra were randomly in a so-called Learning set (75%), ie calibration samples, and a validation set (25%), ie samples for testing the calibration equation. The optimal number of factors for the prediction was determined by cross-validation.

• 1) BIAS; Die durchschnittliche Abweichung zwischen den vorhergesagten Werten (y_n) und den aktuellen Werten (x_n) im Lernset sollte möglichst klein sein (bei Null liegen).
  
• 2) PRESS, "Predicted Residual Error Sum Square" ist die Summe der Fehlerquadrate zwischen den vorhergesagten und den Referenzwerten. Der PRESS-Wert des Validiersets sollte möglichst klein und ähnlich dem des Lernsets sein. PRESS = Σ(xn – yn)2
• 3) "Standard error of estimation" (SEE); Die Standardabweichung der Unterschiede zwischen Referenzwerten und NIR-Ergebnissen im Lernset.
  
• 4) "Standard error of prediction" (SEP); Wie unter 3), allerdings für das Validierset
  
• 5) Der Korrelationskoeffizient (R²) sollte möglichst nahe bei 1 liegen.

The best quantitative regression model was chosen based on the following statistical parameters:

• 1) BIAS; The average deviation between the predicted values (y _n ) and the current values (x _n ) in the learning set should be as small as possible (zero).
  
• 2) PRESS, "Predicted Residual Error Sum Square" is the sum of the squares of error between the predicted and the reference values. The PRESS value of the validation set should be as small as possible and similar to that of the learning set. PRESS = Σ (x n - y n ) 2
• 3) "Standard error of estimation"(SEE); The standard deviation of the differences between reference values and NIR results in the learning set.
  
• 4) "Standard error of prediction"(SEP); As under 3), but for the Validierset
  
• 5) The correlation coefficient (R ² ) should be as close as possible to 1.

In the 3 illustrated cluster image as a 3D plot of the various heavy duty detergent products clearly shows that a distinction can be made between the individual products and thus that a sample to be examined by the method according to the invention can be clearly assigned to a class.

Claims (15)

Method for qualitatively classifying surfactant-containing Composition comprising the following steps: a) Provide a plurality of classes of surfactant-containing compositions each with a plurality of calibration material samples of the surfactant-containing Compositions per class having a plurality of known class properties; b) Irradiation of a measuring light from a predetermined wavelength range on the material samples of the surfactant-containing compositions; c) Detecting the by the material samples of the surfactant-containing compositions passing, reflected, re-emitted and / or scattered measuring light; d) Determining the wavelength-dependent ratio from irradiated to detected measuring light (spectrum) for each material sample the surfactant-containing compositions of a class; e) numerical-mathematical Preparation of the spectral data of the individual material samples surfactant-containing compositions; f) correlating the spectra a plurality of material samples of the surfactant-containing compositions a predetermined class of surfactant-containing compositions; G) Creating a database from the prepared spectral data with different classes of surfactant-containing compositions, based on the measured material samples of the surfactant-containing compositions the individual classes for the calibration of a class correlation; H) at least once repeating steps b) to e) with at least a material sample of a surfactant-containing composition with at least partially unknown properties; and i) determining the classes of surfactant-containing compositions to which the unknown material sample a surfactant-containing composition is assigned, by means of a Class correlation of the measured spectra using the created calibration database from step g). Method according to claim 1, characterized in that that as surfactant-containing composition a detergent, in particular a heavy-duty detergent, color detergent or detergent concentrate, a hand dishwashing detergent, in particular a normal or a balm hand dishwashing detergent, a dishwasher basin, especially normal, 3 in 1 or 4 in 1 dishwashing tablets Detergents for household or industrial use for Cleaning surfaces, Machines or vehicles, in particular a floor cleaner, a car wash or Aircraft detergent, or a surfactant-containing personal care product such as a shower gel is used. Method according to one of claims 1 to 2, characterized that the class properties of the individual classes of surfactant-containing compositions the following group at least partially includes: name of the manufacturer, Product name, purpose. Method according to one of claims 1 to 3, characterized in that the wavelength range is in the NIR-VIS range, in particular in the range of about 700 to 2500 nm (4,000 to 14,300 cm ^-1 ), preferably in the range of about 1000 to 2200 nm (4500 to 10,000 cm ^-1 ). Method according to one of claims 1 to 4, characterized that the measuring light by means of an optical waveguide on the material samples passed the surfactant-containing compositions and / or of them is recorded. Method according to one of claims 1 to 5, characterized the optical layer thickness of the material sample is between about 0.2 and 5 mm, preferably about 0.5 or about 3 mm. Method according to one of claims 1 to 6, characterized that the material samples of the surfactant-containing compositions for Measurement be thermostated, preferably to about 23 ° C. Method according to one of claims 1 to 7, characterized that the numerical-mathematical preparation of the spectral data includes data reduction, in particular by: normalization, Smoothing, especially after Savitzky Golay, 1st derivative, 2nd derivative, multiplicative Scatter Correction (MSC), Invert, Square, Mean Centering, Kubelka Munc transformation, absorption, baseline correction, addition a constant, Shift Negative to Zero, principal component analysis (PCA), Principal Component Regression (PCR), Regression by Method least squares (PLS). Method according to one of claims 1 to 8, characterized that the correlation of the numerically-mathematically processed spectral data multivariate methods, preferably a principal component analysis (PCA) and / or smoothing, in particular according to Savitzky Golay, and / or at least one series development, in particular a Taylor series development and / or artificial neural network algorithms, in particular backpropagation networks, Dynamic Learning Vector Quantization (DLVQ algorithm), Radial Basis Functions (RBF networks), in particular RBF networks trained with the Dynamic Decay Adjustment Algorithm (DDA algorithm) (RBF-DDA network); includes. Method according to one of claims 1 to 9, characterized the correlation is done by clustering. Method according to one of the claims 10, characterized in that that the tolerance circles of each cluster during calibration are adjustable. Method according to one of claims 1 to 11, characterized that about 3 to 20 spectral scans / material sample of a surfactant-containing Composition to be included. Method according to one of claims 1 to 12, characterized at least about 10 material samples of the surfactant-containing compositions per class property for calibration. Method according to one of claims 1 to 13, characterized that about 70% of all Per sample of measured material samples of surfactant-containing compositions for calibration and about 30% used to validate the procedure become. Method according to one of claims 1 to 14, characterized that the classification result as a 3D plot on a screen is pictured.