Paste or batter viscosity is one of the most important factors as concerns the quality aspects, the texture and structure of the final products, such as confectionery, wafer, pastry and bread, and the rheological technological aspects of paste or batter processing during manufacturing, the pumping properties and the risk of lump formation.

For confectionary and wafer manufacture, the factories are using a wide range of flours of varying quality in term of protein level, starch damage, particle size distribution and fiber content. Also the final paste viscosity obtained by means of these flours may considerably vary, even from one batch to another.

An aim of the present invention is to establish a method that allows to predict the final paste viscosity obtained from a given batch of flour. This method should be rapid, easy to execute, inexpensive, non destructive and adapted to be used in a continuous monitoring mode and at factor level. The invention should allow to ensure an on-line viscosity controlled preparation of the paste and the continuous treatment of the paste in order to obtain final product of perfect quality.

The method should also be able to be fully automized.

In order to realise these aims, the method according to the present invention is characterized by the combination of the following steps : - obtaining a near infrared measurement of the flour before preparing the paste,

- using said measurement in a pre-established correlation involving on one hand the near infrared measurements made on several flours and on the other hand the viscosities measured of the corresponding pastes prepared with said several flours in order to obtain an anticipatory viscosity value of the paste to be prepared with said given flour, - comparing said anticipatory viscosity value to a target viscosity value - adjusting at least the flour content and/or the content of the further ingredient of the paste to be prepared in order to obtain said target viscosity value, - preparing a paste with the adjusted flour and further ingredient contents, - measuring the viscosity of the prepared paste, - comparing the measured viscosity value to the target viscosity value, and - in case said measured and target viscosity values being significantly different, further adjusting said adjusted flour and further ingredients contents in order to obtain a viscosity control loop.

It has been recognised that there exists a surprisingly high, almost perfect linear correlation between the near infrared spectrum values of the flour and the final paste viscosity obtained with said flour. By combining this correlation in a method for automatic ingredients adjustment, for viscosity measurement control of the paste, and for re-adjusting the content of the ingredients, it is possible to obtain a continuous mode of paste and final product preparation that is inexpensive, non destructive, and can be adapted without high costs to existing production lines, plants and factories. Moreover, the products obtained by the method according to the invention are of perfect quality which is of particular advantage for the consumer.

Favorably, the near infrared measurement and the value of the corresponding viscosity measured on the paste are added to the pre-established correlation in order to complete and correct the latter.

The method gets therefore continuously more precised and allows to obtain a perfect functioning at factory level.

A favorable embodiment is characterised by the fact that said further ingredient is water and that the flour/water ratio of the paste is adjusted in order to obtain the target viscosity value using a pre-established correlation between the flour/water ratio of the paste and the viscosity of said paste and by the fact that said flour/water ratio is further adjusted if the measured and target viscosity values are significantly different.

These features allow to obtain a particularly easy viscosity control loop usable in any wafer and biscuit plant.

A variant is advantageously characterized by first planning to add as further ingredient a fixed content of enzyme in order to obtain an anticipatory viscosity value that approaches the target viscosity value, and further adjusting the flour/water ration in order to obtain the target viscosity value.

This combined enzyme and water controlled method for batter preparation allows to obtain a very precise viscosity controlled paste preparation.

The invention also concerns an installation for implementing the above define method and is at this purpose characterized by the fact that it comprises - a device for obtaining on-line near infrared measurement of a flour before preparing a paste with said flour and at least a further ingredient including a liquid ingredient, - a central control unit adapted for chemometric calculations, for calculating an anticipatory viscosity value of said paste using a pre- established correlation of near infrared measurements on several flours and of viscosity measurements of the corresponding pastes prepared with said several flours, for calculating at least an adjusted flour content and/or an an adjusted content of a further ingredient of the paste in order to obtain a target viscosity value,

- weighing and volumetric measuring devices for measuring the quantity of flour and of further ingredients, - a mixing apparatus for mixing the flour and the further ingredients, - an on-line viscosimeter for measuring the viscosity of the paste obtained, - said viscosimeter being connected to the central control unit which is adapted to compare on-line the measured viscosity value to the target viscosity value and to further adjust on-line said adjusted flour and/or further ingredients content in order to obtain a viscosity control loop, and - a paste acceptance device controlled by the central control unit and adapted to direct the paste either to further treatment devices in order to obtain the final product or to a readjustment cycle intended to take correcting steps in case that the measured viscosity value is significantly different from the target viscosity value.

These features allow to obtain an installation with an integrated loop for controlling the paste viscosity that is very precise, inexpensive and non destructive and that may be implemented easily in existing plants.

Other advantages will become apparent from the features set forth in the dependent claims and the description setting forth the invention hereafter in more detail with the aid of drawings which represent schematically and by way of example several embodiments and variants of the invention.

Figure 1 shows a recorded spectrum with the absorbance A as a function of the wavelength.

Figure 2 shows the correlation between NIR predicted viscosity values VNIR and measured viscosity values Vi.

Figure 3 is a diagram showing the viscosity without enzyme in relationship with the viscosity with enzymatic treatment having 400, respectively 800 mg enzyme per kilogram of flour.

Figure 4 represents a curve relating the viscosity reduction R to the content E of enzyme added to the flour.

Figure 5 shows a curve relating the paste viscosity to the water content of the paste.

Figures 6 to 10 show schematically five embodiments of installations for implementing the method according to the invention.

For wafers and other confectionery a typical paste or batter consists of wheat flour (Triticum aestivum L.) and water in weight proportions of about 50% of the total liquid paste, dough or batter. Variations in wheat flour quality lead to considerable changes in paste rheology and baking performance. Combination of enzymatic treatments and water level adjustments are used to obtain a target final paste viscosity and to avoid wheat protein aggregation and lump formation. The word paste relates to any mixture having predetermined contents of flour and water and may correspond to batters and doughs.

To investigate mechanism underlying paste viscosity a wide range of flours used in factories around the world (Australia, Brazil, Canada, China, Chile, Colombia, Ecuador, England, France, Germany, India, Japan, Malaysia, Poland, Ukraine and Venezuela) were studied and analysed.

The pastes or batters of each flour were prepared by taking 20g of flour to which 20mut of water were added. The components were homogenised in a mixing machine (Polytron PT 300 sold by Kinematica GmbH, CH-6014 Littau-Luzern) for 1 minute.

Paste viscosity were determined in a Rapid Visco-Analyser (RVA, Newport Scientific Pty Ltd, Warriewood NSW 2102 Australia) in which the contents were stirred at 1600 rpm for 2 minutes and then at 160 rpm for a further 28 minutes at 35°C. Final viscosity was measured after 30 minutes.

Near Infrared Spectroscopy (NIR) is the study of the interaction of electromagnetic waves and matter. The electromagnetic spectrum is located between 780 and 2500 nm.

NIR spectra of the flour samples were recorded using a NIR Systems model 6500 scanning monochromator spectrometer fitted with a sample transport accessory and a reflectance detector (Foss-NIR Systems Inc., Silver Springs, MD 20304-1915, USA). A lead sulphide detector was used for wavelengths above 1100 nm and a silicone detector for those beneath 1100 nm.

Figure 1 illustrates an example of recorded spectrum for one specimen showing the absorbance A A = log(1/R) = log (IO/I) as the surface of the flour, where R is the reflectance, lo is the intensity of the incident radiation I is the intensity of the reflected radiation emerging from the sample.

The example of NIR spectrum extends from 1075 to 2503 nm and data points have been acquired in 2 nm steps.

Such NIR spectra have been recorded for a considerable number of specimens of flours. For these specimens a paste or batter has been prepared and its viscosity measured according to the method described above.

A correlation between the NIR spectra and the paste viscosity has been made by cross-validation calibration.

All of the measured data points, that means absorbance values Aij, i being the specimen number and j being the wavelength, are recorded in a matrix Y having i = n lines and j = p columns. The viscosity values are recorded in a single column matrix X having i = n lines.

These two matrixes have then been treated by Principal Component Analysis (PCA) as disclosed in Chemometrics, Statistics and Computer Application in Analytical Chemistry by Matthias Otto, Wiley-VCH, 1999, ISBN 3- 527-29628-X.

By this method one obtains one NIR predicted viscosity value VNIR for each specimen, which is correlated to the corresponding measured viscosity value Vi.

All n predicted viscosity values VNIR and all n measured viscosity values Vi allow to establish by means of Partial Least Square Regression (PLS), see Chemometics above, a correlation line between the NIR predicted viscosity values VNIR and the measured viscosity values Vi as shown on figure 2.

Software executing such statistical treatment may be purchased under the names VISION and WINISI from Foss-NIR Systems Inc, Silver Springs, USA.

The flour NIR measurements and the paste viscosity measurements of over 50 specimens were reported in the diagram of figure 2 and show an excellent correlation y = 0.99x + 32.41, R2 = 0.98.

Thus, by measuring a considerable amount of flour NIR spectra and paste viscosities, it has been possible to establish a linear correlation between the flour NIR spectrum, giving a predicted paste viscosity value VNIR, and the measured viscosity value Vi.

Accordingly, by obtaining from an unknown flour the near infrared spectrum, by computing the infrared spectrum data points and by using the Principal Component Analysis (PCA) and the Partial Least Square Regression (PLS), it is possible to predict an anticipatory viscosity value VNIR of the paste, which is very useful to know, as one can adjust the manufacturing conditions and the paste composition to the value of the desired target paste viscosity, if necessary.

These adjustments may for example be made by planing and making an enzyme treatment of the flour with a predetermined content of enzyme if the anticipatory viscosity value is higher than the desired target paste viscosity value.

As shown in figure 3, a linear correlation exists between paste viscosity without enzyme (x) and paste viscosity with enzyme (y), for different enzyme levels.

For an enzyme treatment with a content of 400 mg of the enzymes protease and xylanase per kilogram of flour, the correlation y= 0.61 x is obtained. With a content of 800 mg, this correlation is y = 0.42x. Thus batter viscosity may be

reduced considerably by planning the addition of predetermined amounts of enzyme.

These enzyme amounts or contents may be obtained by using the curve represented in figure 4 showing a diagram relating the viscosity reduction, expressed by the ratio R = viscosity with enzyme/viscosity without enzyme, and the content E of protease and xylanase enzyme added to the flour. If for example the viscosity has to be reduced by 30%, a treatment with 280 mg enzyme/kg flour may give an excellent approach to the desired target value of paste viscosity.

A complementary way of adjusting the anticipatory viscosity value obtained by the NIR-spectrum to the desired target paste viscosity is based on the effect of the water content on final paste viscosity. This effect is illustrated in the diagram of figure 5 showing the curve relating the paste viscosity to the water content of the - 0. 2375x paste. This curve corresponds to the equation y = 2 E + 0,8 e R2 = 0. 99, and allows to predict and thus adjust the final paste viscosity to the desired target paste viscosity.

Preferably, it is proposed to plan first the enzyme treatment of the flour, to compare the thus modified anticipatory viscosity value to the value of the desired target paste viscosity and to plan the adjustment of the water content in order to obtain the exact target paste viscosity.

This method of anticipatory evaluation of paste viscosity applies perfectly to industrial utilisation at a factory level where the flour quality can be surveyed continuously by NIR measurements and the final paste viscosity monitored and adjusted if necessary to a desired target value. Interruptions of the manufacturing process can therefore be avoided, whilst preserving constant rheological properties of the paste.

A typical embodiment of installation and method for viscosity controlled paste preparation and final product manufacturing is shown in figure 6. This installation and this method are suitable for industrial application on factory level

especially as concerns NIR measurements on-line control of water and other ingredients addition, and on-line viscosity measurements.

The paste is prepared by means of flour 11 and other ingredients, including ingredients in powder form 12 and liquid ingredients 13, principally water. As described above, a near infrared measurement M1 is obtained by means of NIR on-line measurements of the flour. The measurement corresponds to absorbance measurements of a spectrum of reflected electromagnetic waves energing from the surface of the flour. The spectrum extends between 400 and 2500 nm and is measured in steps of 2 nm.

The values of the measured spectrum are delivered to a central control unit CCU, including a computer programmed for chemometric calculations and statistical evaluation of the measurements.

The flour weight and the weight of the other powder ingredients 12 are determined by means of the weighing instruments W1 and W2 and the weighing results M2 delivered to the central control unit CCU. The latter contains in its memory the above described correlation between the NIR spectra of a considerable number of flours and the measure viscosity values of the corresponding pastes containing all 50% (weight) flour and 50% (weight) water.

This correlation has been obtained throughout extensive calibration experiment.

The purpose of the calibration experiment is mainly to establish a mathematical relationship between spectrum and physical and chemical parameters under investigation previously determined by an independent technique such as viscosity measurements. In this particular case, a wide range of commercial wheat flours used in wafer factories around the world was selected.

A calibration-validation set of 50 wheat flour samples was selected from the collection of flours. The sample set was selected to cover a broad range of wheat flours used in wafer factories. A second sample set of 30 common wheat commercial flours was used for validation of the calibration equations.

Viscosity measurements were carried out with a Rapid Visco-Analyser (RVA4, newport Scientific). Flour (20g) was tipped into an Rapid Visco Analyser (RVA) canister, which contained 20 ml of water. It was then homogenised for 1 min. The container was fitted with a plastic paddle and then placed into the RVA.

The contens were stirred at 1600 rpm for 2 minutes and then 160 rpm for a further 28 minutes at 35°C. Final batter viscosity was measured in RVA Units (RVU) after 30 min. (1 RVU = 1 cP/12. 8).

NIR data were collected using a scanning monochromator spectrophotometer NIR Systems model 6500 fitted with a sample transport accessory (Foss NIR Systems Inc, Silver Spring, MD, USA). All reflectance replicate data were recorded as log (1/R) from 400 and to 2498 nm at 2 nm intervals between datapoints. Spectra collection, manipulation and wavelength selection were performed using routine operation and calibration software. Four- point Fourier smoothing was applied to the data prior to further processing.

Calibrations were developed using an IS13 version 1.04A software (Windows Infrasoft International WiniSI II, 1999) over the range 400-2498 nm which is the focus of this work. Prior to calibration, all spectra were scatter corrected using Standard Normal Variance (SNV) which takes each spectrum and scales it to a standard derivative of 1.0, and Detrend which removes the linear and quadratic curvature of each spectrum. A second derivative calculated over a 5- points (5nm) gap and a 5-points (5nm) smooth were applied. A calibration model was developed using a Modified Partial Least Squares (MPLS) algorithm and a cross-validation strategy called the Leave-one-out (LOO) validation method where some samples are excluded from the calibration set. This is repeated until all samples have been excluded once. The cross-validation strategy measures the prediction accuracy of the calibration model's prediction. The cross-validation results are shown in figure 2.

The performance of the model was determined by the following statistics : strandard error of calibration (SEC), standard error of cross-validation (SECV),

coefficient of determination of calibration (R2), coefficient of determination of cross- validation (1-VR) between RVA values and values estimated by prediction models developed from NIR scans. The NIR cross-validation calibration developed using flour viscosity showed significant correlation between NIR predicted viscosity and RVA measured viscosity: SEC = 11. 7 SECV = 12.35 R2 = 0.997 1-VR = 0. 89 The MPLS calibration fitted the other regression methods such as Principal Component Regression (PCR) and Partial Least Squares Regression (PLSR). The modification in MPLS involves the standardisation of variables after each iteration.

The high R2 and the low SEC values for batter viscosity obtained for the MPLS calibration equation could be linked to the possible measurement of starch, polysaccharides and protein contents by NIR. This study shows that NIR has a great utility as a rapid analytical technique to efficient predict batter viscosity in flour and as a tool for quality control of raw materials and selection of suited flours.

The NIR raw spectra wheat samples (figure 1) exhibit typical broad absorbance bands. Four wavelengths were significantly involved in batter viscosity prediction. These wavelengths corresponded to 1926 nm for CONH in 2nd overtone, 1076 nm for HN2 as deformed vibration, 2180 nm for N-H and protein bonds in 2nd overtone and 2322 nm for starch and cellulose. Two major components seem to be involved in batter viscosity: protein and polysaccharides.

This observation is in agreement with the effect of starch damage on batter viscosity, i. e. an increase in the percentage of starch damage at a constant level of water led to an increase in batter viscosity. In batter, the water uptake by damaged starch was significantly higher than from non-damaged starch. The effect of non- starch polysaccharides on batter viscosity has also been recognised.

An increase in the percentage of protein led to an increase in batter viscosity, up to 11 % d. m. protein. Above this value, a non-significant effect of protein level was observed.

Returning to figure 6, an anticipatory viscosity value is determined on-line by NIR measurement and chemometric calculation in the central control unit (CCU).

This anticipatory viscosity value is compared to a target viscosity value and the flow/water ratio of the paste to be prepared is adjusted in order to obtain said target viscosity value. This may be done by means of the correlation curve shown in figure 5 stored in the memory of the Central Control Unit (CCU).

If the predicted viscosity value is VP and the target viscosity value is VT, their difference AV = Vp-VT corresponds to a difference AW in the water content in figure 5. For example, a planned viscosity decrease AV = 1400-1000 cp = 400 cp corresponds to an increase of water content AW = 1.44% of the paste. Thus in the example of wheat flour, it is necessary to increase the water content from 50% to 51,44% (weight) in order to obtain a target viscosity of 1000 cp.

The central control unit (CCU) sends orders C3 to the volumetric valve W3 which measures the exact volume of water to be added to the measured weight W2 of flour. All ingredients 11, 12 and 13 are mixed in the mixing tank MT. The viscosity of the paste or batter is measured at the outlet of the mixing tank by an on-line viscosimeter OV, for example of the type sold by Hydramotion Ltd, YO 170NW, England.

The measured viscosity value M3 is delivered to the central control unit and compared to the target viscosity value VT.

If these two values are significantly different, the central control unit calculates a re-adjusted water content by means of the correlation curve of figure 5 and controls the volumetric valve W3 according to this re-adjusted flour/water ratio.

Moreover the measured viscosity value M3, preferably normalised to a flour/water ratio equal to 1, and the corresponding NIR measurement M1 are added to the calibration samples. The calibration is then automatically extended,

completed and corrected accordingly by chemometric calculation as explained above. Thus, by continuously increasing the calibration samples, the chemometric calculation and the predicted viscosity values get more and more precise.

According to the results of the online viscosimeter OV, the central control unit decides paste acceptance or refusal, a valve PA directs the paste to the baking device BD or to a readjustment cycle RC.

In or directly outside the baking device on-line colour measurements MS of the final product are made for example by a scanning spectrophotometer CM for wavelengths in the visible light. These colour measurements allow to adjust the baking time or temperature C4 for a given flour/water ration in order to obtain the desired baking degree.

On the final product, density measurement M6 are made by weight and volume measurement devices WV. The value of the density is used for fixing a maximum value of water addition WA in respect of the volumetric measurement of the water content.

Furtheron, immediately after the production or lateron, textural properties M7 of the final product are determined by measuring members TP measuring for example the hardness, the rigidity, the brittleness, the fissuration, the crustiness of the final product. The values of these measurements are delivered to the central control unit CCU and allow to acquire further data permitting future predictions of other technological parameters of the flour, the paste or the final product by establishing similar chemometric calculations involving the NIR measurements and said registered data.

The on-line method and installation of figure 6 comprises moreover a readjustment cycle RC. If the measured paste viscosity value M3 is significantly different from the target viscosity value, the readjustment cycle RC is designed for correcting this difference. If the measured viscosity is too low, the paste is directed to the baking device BD, but the baking time C4a is adjusted and increased in order to obtain the desired final product. If the measured viscosity is too high, the

paste is directed by the valve to the mixing tank and a readjustment of the water content of the paste is calculated as explained above with reference to figure 5.

The final product may be any product produced by means of a paste comprising any flour, such as wafers, biscuits, cakes, snacks crackers, cereal products, confectionery, pastry, pap, porridge, animal food. The quality of these final products is greatly improved by the invention which is therefore of particular benefit to the consumer.

The variant of the method and installation shown in figure 7 has a similar general principal. Also the same elements are indicated by the same references letters and numbers.

The principal difference is in the presence of a resting tank RT. The paste mixed up in the mixing tank MT is delivered to the resting tank RT where the paste is kept in a thermostatic state under stirring action for a predetermined resting time in which fermentation takes place. As disclosed above, the paste is then delivered to the on-line viscosimeter OV. In case that the measured viscosity value is significantly too high, the paste is directed by the value PA to the resting tank RT in this variant. A water readjustment is calculated according to the measured viscosity and instructions C3a are given to a second volumetric valve W3a to deliver the calculated amount of water to the paste contained in the resting tank RT in order to obtain a paste with a viscosity equal to the target viscosity value, as explained above with reference to figure 5. This second viscosity adjustment is normally rare and occurs principally when changing the crop or the miller supplier.

Thus this variant is characterized by a viscosity control loop based on the flow/water ratio, the other ingredients not having an effect on the viscosity control ; the paste is moreover subject to a resting operation and/or fermentation in a supplemental resting tank RT.

The embodiment represented in figure 8 has general layout which is similar to the variant of figure 7, but the viscosity is controlled by at least one enzyme added in fixed contents to the flour and having an influence on the paste viscosity.

As shown on the top of figure 8, the paste is prepared by means of flour 11, at least another ingredient 12 in powder form, a predetermined content M2a of enzymes 12a, for example protease and/or xylamase and a liquid normally water 13. In this embodiment, the contents M1, M2, M2a of flour 11, other ingredients 12 and enzymes 12a are fixed and by means of the NIR spectrum an anticipatory viscosity value can be calculated, by first obtaining an anticipatory viscosity value for a flour/water ratio equal to 1 without enzyme and by reducing said value by adding a predetermined amount of enzyme according to the correlation lines shown in figure 3, for example a content of 400 or 800mg/kg. The enzyme content is determined in such a manner that the anticipatory viscosity value with enzyme is approaching the desired target viscosity value, but is preferably still higher than said value. It is obvious that figure 3 may be completed by further correlation lines for other enzyme contents. It may also be possible to use a graph as illustrated in figure 4, allowing to determine the enzyme content for a desired viscosity reduction. The target viscosity value is then obtained by finally adjusting the water content C3 of the paste as explained before with reference to figure 5.

All other features of this second embodiment are identical to those of the variant shown in figure 7.

In figure 9 is represented a variant of the preceding embodiment. In this variant, the content of enzyme 12a is not fixed previously, but this enzyme content depends on the flour biochemical quality. It is possible to fix the flour/water ratio to certain value, for example equal to 1. By means of the NIR spectrum, an anticipatory viscosity value without enzyme is then obtained, which should be above the target viscosity value. The quantity of enzyme W2a is finally calculated by means of the correlation lines of figure 3 or the graph of figure 4 and instuctions C2a emitted from the central control unit.

The exact quantities of flour W1, of other ingredients in powder form W2, of enzyme W2a and of water W3 are delivered to the mixing tank MT, mixed up and further processed as explained with reference to figures 7 and 8.

In the case that the viscosity value M3 measured by the on-line viscosimeter OV should not be identical to the target viscosity value, the quantity of enzyme and/or of water of the paste may be adjusted accordingly for example by means of the graphs of figures 4,5 and 6.

In a further variant represented in figure 10, at least another ingredient 12b in powder form having an effect on paste viscosity, for example starch, non starch polysaccharides, proteins, gum, texture agents is added to the paste. For these ingredients, it is possible to apply an analogous calibration method as for the flour.

A NIR absorbance spectrum M1 b for a given ingredient, for example starch, is recorded and used in a pre-established correlation involving the NIR absorbance spectra of a significant number of samples of said ingredient and viscosities measured on pastes containing a predetermined amount of the corresponding sample of ingredient in order to obtain an anticipatory viscosity value of the paste to be prepared depending on the nature and the NIR absorbance spectrum of said ingredient.

Knowing this anticipatory viscosity value, it is possible to adjust the latter to the desired target viscosity value by modifying the flour/water ratio by giving a modified instruction C3 to the volumetric value W3.

In this variant, no resting tank is provided and the paste is delivered from the mixing tank directly to the baking device. All other particularities of the method and installation are identical to those disclosed with reference to figures 6 to 9.

Of course, the embodiments and variants described above are in no way limiting and can be subject of all desirable modifications within the framework defined by the independent claims.

The statistical treatment of the NIR spectrum and its individual absorbance values may be of any kind. The spectrum measured may be from 1100 nm to 2500 nm or any other limits, such as from 780 nm to 2500 nm. The steps of 2 nm may be fixed differently for example to 10 nm. It may also be possible to measure the

absorbance of only one or several selected significative wavelengths, instead of measuring a complete spectrum.

Different types of sources of the electromagnetic waves may be used, such as incandescent bulbs, light emitting diodes, tungsten lamps. Filters may be used in conjunction with any of these sources to eliminate unneeded radiation, such as visible radiation. This may contribute to prevent unnecessary sample heating and potential heat damage of the flour.

Instead of measuring the reflected intensity for obtaining the absorbance, it is possible to measure the intensity of a transmitted electromagnetic wave by using thin light-transmitting flour samples.

As concerns the enzyme and water adjustments, it is also possible to combine the enzyme content and the water content in one step. It may also be useful to compute directly the predicted viscosity for flour with predetermined contents of enzyme, for example 400 mg or 800 mg enzyme in 1 kg flour.

Treatments of the flour and paste with other enzymes or with other substances may also be programmed in order to adjust the predicted final paste viscosity.

The treatment of the paste for obtaining the final product may be made by any device or installation, such as baking ovens, steam treatment, cooking apparatus, microwave devices, drying devices, hot moulding devices, packaging devices.

Other final product measurements may be added, comprising any physical or chemical test devices. Any other intermediate treatment of the paste before the final baking operation may be added.

In order to control the viscosity of the paste, an anticipatory viscosity value may be calculated by means of correlations involving NIR measurements on one or several other ingredients different from flour and enzyme, such as starch, non starch polysaccharides, proteins, texture agents, and adjusting their contents, for example by the flour/other ingredients ration in order to obtain the desired target viscosity value.