WO2005045421A1

WO2005045421A1 - Improved method and apparatus for measuring bitterness in beer and brewing samples

Info

Publication number: WO2005045421A1
Application number: PCT/DK2004/000768
Authority: WO
Inventors: Jakob Christensen; Lars NØRGAARD; Lars Munck; Anne Marie Ladefoged
Original assignee: The Royal Veterinary And Agricultural University
Priority date: 2003-11-05
Filing date: 2004-11-05
Publication date: 2005-05-19

Abstract

The invention concerns an improved method and apparatus for the quantitative determination of the bitterness content of beer, or the process stream of beer production. The method employs UV/VIS spectroscopy to perform a direct determination of absorbance by a sample from the brewing process stream, without necessitating a sample preparation step or extraction of bitterness components. The absorbance at one or more wavelengths is used to determine bitterness content by comparison with the multivariate absorbance measurements of beers of known bitterness content by regression analysis.

Description

Improved method and apparatus for measuring bitterness in beer and brewing samples

Field of invention

The invention relates to an improved method and an apparatus for measuring the bitterness in beer or in brewing samples, a method which optionally can be used in-line, on-line, at-line and/or off-line.

Background of the invention

Bitterness is an essential quality parameter of beer, and the determination of bitterness is a routine analysis in breweries. The bitter compounds in beer originate from hops, which are included in the brewing process for their flavouring, aroma-enhancing and bacteriostatic properties (DE Biggs, JS

Hough, TW Young: Malting and Brewing Science. Volume 1: Malt and Sweet Wort, Volume 2: Hopped Wort and Beer. ". Edition 1981, Chapman and Hall, London). Hops are the cone-like flowers of the female hop vine (Humulus lupulus). Within the hop flower there is a soft resin, comprising alpha acids and beta acids that are the precursors of the bitter compounds in beer. Alpha acids from the hops, which are water-insoluble, undergo a thermal isomerization during the wort boiling step of the brewing process, when they are converted to water-soluble iso-alpha acids. The three alpha acids, humulone, cohumulone and adhumulone each give rise to two epimeric iso- alpha acids, as depicted below:

Hops can be added in the brewing process in a non-isomerized or isomerized form, or as hop oils. Non-isomerized hop products, in the form of hop cones (dried hop flowers), plugs, pellets (powdered hop flowers) or hop extract (alpha acid extract of hop flowers), are added to the wort before boiling. Isomerized hop products, such as pellets and extract, can be added later in the brewing process or to the final beer. Hop oils can be added to give beer hop aroma without imparting any bitterness. Hops are the most expensive raw material used in brewing beer, and hence their addition in the brewing process is carefully controlled. The alpha acid content of hops not only varies between different hop varieties, but can also vary between batches, which in turn determine the amount of iso-alpha acids formed during wort boiling and the final bitterness of the beer. Hence, there is a need to monitor and adjust for bitterness content during the brewing process.

Beer is commonly brewed from malted barley that is milled and mixed with hot water to form a mash (DE Biggs, JS Hough and TW Young; Malting and Brewing Science. Volume 1: Malt and Sweet Wort, Volume 2: Hopped Wort and Beer ", Edition 1981, Chapman and Hall, London). During mashing, starch and proteins in the malt are converted to fermentable sugars and amino acids by malt hydrolases. The fermentable extract, or sweet wort, is then strained through the spent grain at the bottom of the mash, and transferred to the brew kettle, where it is brought to a boil. Hops are added at different times during wort boiling to impart bitterness and/or aroma. The sweet wort is then cooled and aerated, and brewers' yeast is added. The yeast produces alcohol and carbon dioxide and other byproducts by fermentation of the sweet wort. After fermentation, isomerized hop products can be added to the "green beer" to increase the bitterness before it undergoes maturation. The last step in the brewing process can include filtration, and carbonation. Finally, the beer is moved to a holding tank where it stays until it is bottled or kegged. The brewing process is best optimized if bitterness analysis is performed on the wort, following boiling, and at later steps in the process, such as after the fermentation, to decide whether additional hops should be added before maturation, and to control the quality of the final product. The traditional and internationally approved method for bitterness determination in beer involves the extraction of iso-alpha acids from acidified beer into iso-octane, followed by a centrifugation step, and photometric measurement at a wavelength of 275 nm against a reference of pure iso- octane (European Brewery Convention, 1998, Analytica-EBC, 7.8. Nϋrnberg, Germany). The optical density of the acidified solvent extract is multiplied by a factor to produce an analytical value, measured as Bitterness Units (BU): BU = Optical Density at 275 nm x 50. This method is costly, time-consuming and subject to laboratory error. A faster method for determination of bitterness in beer, that can be performed on-line or at-line, is therefore highly desirable. More detailed and specific analysis of bitter compounds in beer is possible using HPLC (Raumschuh et a/.1999, J Am Soc Brewing Chemists, vol4: 162-165). This method is also time consuming, and is not currently amenable to on-line analysis in the brewing process. US 4 751 185 discloses a method for the detection and quantification of individual hop bittering components. The method comprises the steps of separating the bittering components from the samples, for example by liquid chromatography and detection of the resulting constituents photometrically at around 270-302 nm. This approach is also not suitable for on-line or at-line detection.

Chemometrics is the science of relating measurements made on a chemical system or process to the state of the system via application of multivariate mathematical or statistical methods. Chemometric methods are very suitable for the analysis of spectral data, and are extensively used especially within the area of near infra-red (NIR) spectroscopy. Chemόmetric methods, such as Partial Least Squares Regression (PLS) and Neural Networks, have been utilized to extract spectral information that can be correlated to physical/chemical quality measurements, and have been incorporated in fast on-Iine/at-line instruments for quality control and process monitoring and control. NIR instrumentation dedicated for breweries is commercially available for fast monitoring of alcohol content and original gravity of beer (Fellows, 1995, The Brewer p.187-192). Nørgaard L. et al. (2000) in Applied Spectroscopy 54: 413-419, discloses a chemometric method for determining the amount of original extract of beer. The method comprises degassing a sample of undiluted beer, obtaining a full-spectrum in the near-infrared (NIR) region and applying various regression models to predict original extract.

WO 95 21 242 discloses a method for monitoring the color and bitterness in beer. It suggests an automated, and possible in-line method for monitoring bitterness in beer, using fluorescence spectroscopy evaluated with multivariate data analysis. Compared to ultraviolet-visible spectroscopy (UV- VIS) fluorescence techniques are complex, generally more expensive and the calibration development and calibration transfers can be cumbersome to perform. Therefore, there is a need for an easier, cheaper and more convenient method and apparatus for determining the bitterness in beer, which can be used both in-line, on-line, at-line and off-line with sufficiently high accuracy and precision.

DE 101 08712 A discloses a method for determining the spectral profile of a beer sample, based on a spectroscopic measurement of absorbance from above 300 nm in the visual and near-infrared range. Thereby it is possible to detect changes in a beer sold to the consumer that may result from mishandling during delivery or erroneous labeling of the sold product. The method includes a multivariate data analysis in form of Principal Component Analysis (PCA) to analyze the obtained spectral data. PCA is used to compare and classify the measured spectral profile of a beer to a library of known and well described beers, but is not designed or capable of making a quantitative determination of bitterness or any other quality parameter of a beer.

Krϋger et al. (1989) in Monatsschrift fur Brauwissenschaft 42(8): 312-316 describes a commercial spectrometer, measuring absorbance in the 200 to 900 nm range. The instrument can be programmed to measure bitterness in bitterness units based on measuring absorbance at one single wavelength (275 nm) of a solvent extract prepared from a beer sample. Thus, the method is an automated implementation of the standard EBC method (European Brewery Convention, 1998, Analytica-EBC, 7.8. Nϋrnberg, Germany).

Summary of the invention

Accurate measurement of bitterness compounds formed during the brewing process and in the final beer is an important parameter for quality control. The process stream in a brewery, which includes the wort, fermenting wort, ^" green beer and mature beer, contains a complex mixture of compounds, many of which absorb light in the Ultraviolet-visual (UV-VIS) range. Detecting specific compounds, based on their light absorbance in the UV-VIS range is often hampered by a low signal-to-noise ratio due to other absorbing compounds in the mixture. In beer, the absorbance of iso-alpha-acids is considered to contribute with around 1 % to the overall absorbance at 275 nm. Other substances like flavonoides, polyphenols, and amino acids that are abundant in malted barley and the brewing process stream also absorb light in this area. Analysis of bitterness compounds in beer and brewing samples by UV-VIS absorbance has thus traditionally depended on their prior selective solvent extraction from beer. Alternatively, fluorescence spectroscopy has been a method of choice for the detection of fluorescent compounds in complex mixtures due to generally lower background interference.

The invention provides an accurate method for determining the bitterness components during the brewing process, both in-line, on-line, at-line and offline, without requiring any extraction or separation of the bitterness components in the analyzed samples. The method for determining bitterness, provided by the invention, is based on a combination of UV or UV-VIS absorbance spectroscopy and chemometric analysis. The method of the invention surprisingly shows that chemometric data analysis of the UV-VIS light absorbance of bitterness compounds in a brewing sample can provide a more accurate determination of bitterness units than analysis of their fluorescence emission spectrum.

The method of the invention, based on UV-VIS spectroscopy, provides a number of specific advantages over known fluorescence spectroscopy methods. The instrumentation is less complex and expensive, and the calibration development and calibration transfers are simpler to perform.

Further, the equipment for UV-VIS absorbance spectroscopy is cheaper and easier to calibrate, repair and maintain.

According to the present invention there is provided a method for performing a quantitative determination of the bitterness content of a brewing sample comprising the steps of:

- measuring absorbance of the brewing sample at two or more wavelengths in the region of about 190 nm to about 800 nm, and

- comparing the measured absorbance with stored multivariate absorbance measurements of brewing samples with known bitterness content by means of a regression analysis, and

- making a quantitative determination of the bitterness content of said brewing sample.

According to a further aspect of the invention is provided an apparatus for quantitative determination of the bitterness content of a brewing sample, comprising a data processing unit (5, 19) with a program code means adapted to: - receive data comprising absorbance of said brewing sample measured at two or more wavelengths in the range of about 190 nm to about 800 nm, and - compare by means of regression analysis said received data with stored data comprising stored multivariat absorbance measurements of brewing samples with known bitterness content, and - make a quantitative determination of the bitterness content of said brewing sample, when the program code means are executed on the data processing unit.

In a preferred embodiment of the invention, the apparatus further comprises a measuring unit adapted to: - irradiate the brewing sample at two or more wavelengths in the range of about 190 nm to about 800 nm (1 ,17), and - measure absorbance at said two or more wavelengths by said brewing sample (2,18), and - transfer data comprising said measured absorbance to said data processing unit (5, 19).

The measuring unit according to the invention may further be adapted to: - transfer light to and from the brewing sample (16).

Brief description of the drawings

Figure 1 shows the UV-VIS absorbance spectra from 240 nm to 600 nm of 21 beer samples measured with a UV-VIS spectrophotometer. DB is Dark Beer, LB is Light Beer and A is Absorbance.

Figure 2 shows the bitterness units (BU) of each of 21 beer test samples determined according to the EBC method (abscissa values), plotted against the BU predicted by a PLS model with six components, based on UV-VIS spectra according to the present invention (ordinate).

Figure 3 shows the bitterness units (BU) of each of 21 beer test samples determined according to the EBC method (European Brewery Convention, 1998, Analytica-EBC, 7.8. Nϋrnberg, Germany), (abscissa values), plotted against the BU predicted by a PLS model with four components, based on fluorescence spectra according to the method of WO 95/21242 (ordinate). Figure 4 shows an off-line apparatus.

The apparatus comprises a measuring unit comprising a UV-VIS light source (1), a light detector (2) and a processing unit with means for data acquisition, spectrometer control and the quantitative measurement of bitterness content of a brewing sample (5). A sample container can be placed in the apparatus in a receiving compartment (3) positioned in the light path between the light source (1 ) and light detector (2). The sample container can be a quartz cuvette or vial, or a stop-flow or flow-through cell and has a width that is equal to the optical path length. The apparatus may have an inlet and /or outlet (4) for the sample.

Figure 5 shows an on-line apparatus.

The apparatus comprises a measuring unit comprising a UV-VIS light source (17), means for transferring light (16), a measuring probe (15), a light detector (18) and a processing unit for data acquisition, spectrometer control and the quantitative measurement of bitterness (19). The brewing sample flow (13) is illuminated from the measuring probe (15). Transfer of light to and from the measuring probe (15) can be conducted using optical fibres (16). The measuring probe (15) can be designed as a transmission, transflectance or reflectance probe. The measuring unit (spectrometer) and processing unit situated in area (12) and the measuring probe in area (11 ) are connected by the means for conducting light (16), where the two areas may be physically remote.

Description of preferred embodiments

In this application the following definitions are used:

Absorbance: The degree of attenuation of a light beam. The absorbance can be measured either in transmission or reflection/transflection (pseudo absorbance) mode. With respect to the present invention, absorbance is determined by illuminating a sample and detecting transmitted or reflection/transflection light.

At-line: A measurement performed on a sample taken from the process stream or tank, where the measuring equipment is setup close to the process line.

Bandwidth: A measure of the band of wavelengths detected by the instrument for a given wavelength.

Beer tank: A tank used for processing or storage of a brewing intermediate (e.g., wort boiling; wort storage; fermentation; lagering) during the brewing process.

Brewing pipeline: Any pipeline in the beer production process that transports a brewing intermediate or finished beer product containing iso- alpha acids. Brewing sample: A sample taken anywhere in the beer production process including the wort, boiled wort, fermenting wort, green beer, and the final product of the brewing process, beer (for example ale, lager, porter) . BU: Bitterness units, according to the EBC method for determination of bitterness in beer (European Brewery Convention, 1998, Analytica-EBC, 7.8. Nϋrnberg, Germany): Optical Density at 275 nm of the solvent extract of a beer sample multiplied by a factor of 50.

Chemometrics: The chemical discipline that uses multivariate mathematical and statistical methods to design or select optimal procedures and experiments, and to provide maximum chemical information by analyzing chemical data. In the present context chemometrics is used as a means of dealing with complex UV-VIS data to extract information about the bitterness content of the analyzed beers. Examples of chemometric methods include Partial Least Squares Regression (PLS), Multiple Linear Regression, Principal Component Regression, Ridge Regression, Neural Networks. Discrete spectrum: A spectrum measured at multiple individual wavelengths.

In-line: A measurement performed directly in the process stream or in the tank. Off-line: A measurement performed on a sample taken from the process stream or tank, where said measurement is performed in a laboratory. On-line: A measurement performed in e.g. a bypass stream connected to the process stream or tank of the beer production process, where said bypass stream flows though the measuring chamber. However on-line is not limited to this setup, and any setup, where measurements are performed on a stream in direct contact to the process stream or tank is to be regarded as an on-line measurement.

PLS: Partial Least Squares Regression (PLS) is a method that correlates an independent data matrix (X, e.g. spectral data) to a dependent matrix or vector (y, e.g. BU reference data) by several regression steps. Regression analysis: A mathematical procedure correlating two matrices (X and Y/y) in a linear or non-linear model. PLS is an example of a regression analysis. Sample preparation: Refers to any kind of treatment of the brewing sample prior to determination of bitterness, such as dilution (e.g. by water), pH adjustment by any buffer, degassing, addition of antifoaming reagents (e.g. iso-octanol), etc. Step size: The distance that is used for digitizing the spectrum. If every 5^th nm is recorded the step size is said to be 5 nm.

Undiluted: Using the raw brewing sample without adding any further solvent or chemicals to the sample during measurement.

UV-VIS spectroscopy: Ultraviolet-visual (UV-VIS) spectroscopy is the absorption of ultraviolet and/or visible light by one or several molecules causing the excitation of electron(s) from a ground state to an excited state. The electronic transitions seen in UV-VIS spectroscopy are e.g. σ to σ* (alkanes), σ to π* (carbonyl), π to π^* (carbonyl, alkenes, alkynes), η to σ* (oxygen, nitrogen, sulphur) and η to π* (carbonyl). The wavelength range covered by UV-VIS spectroscopy is from 190 nm to 800 nm. Wort: The liquid extract which is obtained from the mash during the brewing process. Apparatus

The light source for apparatus according to the invention is preferably capable of irradiating the beer sample at any wavelength in the area from about 190 to about 800 nm and the light detector of the apparatus is preferably capable of detecting and measuring light at any wavelength in said area.

In a preferred embodiment the light source and the light detector of the apparatus are capable of measuring in the entire range from about 190 to about 800 nm. Preferably the apparatus is capable of measuring in a selected subsection of the UV-VIS such as 200-600 nm, more preferably 220-400 nm and most preferably 240-300 nm. Optionally the apparatus can be constructed to only measure in one or more of the following regions 190- 240 nm, 240-275 nm, 275-350 nm, 350-430 nm, 430-550 nm, 550-650 nm and 650-700 nm.

In another preferred embodiment the light source and the light detector are constructed to irradiate and measure at discreet wavelengths. The smaller the steps the closer to a continuous spectrum the discreet measurements become. Preferably the size of the steps are within the range 0.001 nm to 20 nm, preferably 0.01 nm to 5 nm, more preferably 0.02 nm to 2 nm and most preferably 0.1 nm to 1 nm. Examples of such useful steps are 5, 2, 1 , 0.5, 0.1 , 0.05, 0.01 nm. If a faster measured spectrum is desired, the step can be further enlarged or if a more detailed and more accurate spectrum is desired smaller steps can be taken.

Optionally the apparatus is constructed to measure the spectrum with varying step lengths using small steps in one area and large steps in another area. Hereby it is possible to combine the advantages of a fast spectrum with the advantages of an accurate spectrum. By these variations it is possible to obtain any operating profile which takes into account the need for a certain time limit for the results, and the need for a certain accuracy of the determination. In a preferred embodiment the apparatus is constructed to be able to measure the spectrum with varying bandwidth using any combination of bandwidth e.g. a small bandwidth in one area and a large band width in another area. Suitable bandwidths are listed in the description of the method according to the invention.

In another embodiment of the invention the apparatus is constructed to irradiate and detect light according to the preferred subsection and distribution or bandwidth within said subsections as described for the following method.

Examples of a light source useful for irradiating the brewing sample are a diode, laser, Xenon lamp, deuterium lamp, halogen lamp, tungsten lamp, Nernst glower, hollow cathode lamp or any other light source known by a person skilled in spectroscopy.

Examples of means useful for wavelength selection in the apparatus are either a prism, grating, filter or any other wavelength selector known by a person skilled in the art.

Examples of a light detector useful for detecting light transmitted by an irradiated brewing sample and thereby measure the absorption of light by the sample is either a photodiode, photomultiplier, phototube, photocell, photoconductor or any other light detector known by a person skilled in the art.

Examples of means useful for transferring light from the light source to the sample and from the sample to the light detector is an optical fiber, fiber optic cable, light pipe or any other light transmitter known by a person skilled in the art. Absorption measurement may be recorded using any kind of data processing device such as a computer or a chip or any other processor known by a person skilled in the art.

Optionally the apparatus has means for measuring the absorbance as transmittance, reflectance, transflectance.

The measured absorption data can be transferred from the measuring unit to the data processing unit by any kind of wire or wireless communication equipment, all known to a person skilled in the art.

The data processing unit comprising a program code means can be any kind of data processing device such as a computer or chip, which is capable of making a quantitative determination of the bitterness content of a brewing sample.

The processing unit may include any circuit and/or device suitably adapted to perform the above functions. In particular, the processing unit may comprise a general- or special-purpose programmable microprocessor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Programmable Logic Array (PLA), a Field Programmable Gate Array (FPGA), a special purpose electronic circuit, etc., or a combination thereof. Preferably the processing unit is capable of using chemometrics as the method for quantitative determination. Preferably the measuring unit is an integral part of the apparatus. Further it is preferred that the processing unit is a storage medium.

In a preferred embodiment of the apparatus, it can be used in both in-line, on-line, at-line and off-line detection. The used detection mode is chosen by the user. The apparatus according to the invention is capable of quantitatively determining the bitterness content of both an intermediate of the process stream and a finished product of the brewing process and is more precise than any previously disclosed apparatus for determining the bitterness content of beer.

One embodiment of the apparatus according to the invention can be adapted to perform off-line sampling as depicted in Figure 4, and or to perform on-line sampling as depicted in Figure 5. The UV-VIS instrumentation for both adapted forms of the apparatus is assembled from standard components including a light source and wavelength selecting means. Irradiation from the light source of the apparatus may be transferred to a sample, optionally by way of an optical fiber mounted in the apparatus. For off-line sampling, the sample is contained in a sample container positioned in a receiving compartment of the apparatus, which may be linked to an inlet and/or outlet connection for the sample (figure 4). The sample container can be a quartz cuvette or vial, or a stop-flow or flow-through cell and have a width that is equal to the optical path length. For in-line sampling (figure 5) irradiation from the light source is transferred to a measuring probe placed in direct contact with the process stream comprising the sample. Transmitted light from the irradiated sample, located either in the sample container (off-line) or in the process stream monitored by the measuring probe (on-line), is transferred, optionally by means of an optical fiber, to a suitable light detector e.g. a photodiode array or a photomultiplier. The spectroscope of the UV-VIS instrumentation is coupled to a processing unit comprising a computer system where software for data analysis is installed as well as software for instrument control. The absorbance of light by the irradiated sample is detected and measured by the spectroscope and this absorbance data is transferred to a data processing unit. The data processing unit comprises a program code means and is capable of performing the following steps: receiving data comprising measured absorbance at two or more wavelengths in a range of about 190 to about 800 nm by a brewing sample, and comparing by means of regression analysis the received data with stored data comprising multivariate absorbance measurements at two or more wavelengths in a range of about 190 to about 800 nm of brewing samples with known bitterness content, and making a quantitative determination of the bitterness content of said brewing sample, when said program code is executed on the data processing unit.

The output of a measurement is a quantitative measure of the bitterness in terms of BU of the sample, as well as an approval/rejection of the measurement according to an implemented outlier detection method.

Spectral monitoring of brewing samples

Iso-alpha acids have chromophores with maximum absorption around 275 nm. The UV absorbance of iso-alpha acids in a brewing sample is normally determined after their chemical extraction from the brewing sample, which separates them from other UV absorbing components in the sample. The method of the present invention however allows the determination of bitterness content to be performed directly on the brewing sample. By measuring the UV-VIS absorbance of the brewing sample at more than one wavelength, the contribution of the bitterness compounds to the UV-VIS spectra can be mathematically extracted using these measurements in a multivariate calibration to reference BU values of a set of beers, determined according to the EBC method (European Brewery Convention, 1998, Analytica-EBC, 7.8. Nϋrnberg, Germany).

Although all iso-alpha acids contain the same chromophore, the UV-VIS spectra of the various iso-alpha acids are found to differ. Thus, by measuring absorbance at more than one wavelength it is possible to detect and discriminate between many different iso-alpha acids.

In one embodiment of the method according to the invention there is a sample preparation step prior to the measuring step. Sample preparation, by way of example only, can be one or more of the following: dilution (e.g. by water); pH adjustment by any buffer; degassing; addition of an antifoaming reagent (e.g. iso-octanol).

A UV-VIS spectrum of a brewing sample is recorded in the range of about 190 nm to about 800 nm. The measurement can be performed on an undiluted brewing sample, but prior sample preparation is optional.

In a preferred embodiment of the invention an analysis based on the full spectrum is used for determination of bitterness content. In another preferred embodiment only a part of the entire UV-VIS spectrum is used in the analysis. The wavelength ranges used are preferably the range 200 to 600 nm, more preferably the range 220 to 400, and even more preferably the range 240 to 300 nm.

In another preferred embodiment of the invention full spectrum analysis is performed on more than one predetermined subsection of the entire UV-VIS range.

Examples of said subsections, useful in the measurement of bitterness, include 240 to 275 nm in combination with 400 to 430 nm; 240 to 275 in combination with 650 to 700 nm; or all three ranges in combination. In fact there are numerous subsections which can be used according to the invention, and therefore the subsections named are only for illustrative purposes and not meant as a limitation.

The spectrum can be measured and recorded as a discrete spectrum. The measurement of said discrete spectrum can be performed in steps, preferably with a size of the steps within the range 0.001 nm to 20 nm, more preferably 0.01 nm to 5 nm, more preferably 0.02 nm to 2 nm and most preferably 0.1 nm to 1 nm. Examples of such steps are e.g. 5, 2, 1 , 0.5, 0.1 , 0.05, 0.01 nm, but the steps are not limited to these sizes. In fact any step size can be used to record the discrete spectrum, principally depending on the desired density of data points.

Each measurement also has a bandwidth. Suitable bandwidths are within the range 0.001 nm to 20 nm, more preferably 0.01 nm to 5 nm, more preferably 0.02 nm to 2 nm and most preferably 0.1 nm to 1 nm. Examples of such bandwidths are e.g. 5, 2, 1 , 0.5, 0.1 , 0.05, 0.01 nm, but the bandwidth are not limited to these sizes.

In another embodiment an even narrower wavelength range or possibly only a few selected wavelengths within the entire UV-VIS range are used in the determination of bitterness. In any of the previously mentioned ranges for measurement, a predetermined number of single wavelengths, preferably more than 2 wavelengths, more preferably more than 5 wavelengths, even more preferably more than 10 wavelengths, can be use in the determination. The single wavelengths can be evenly distributed in the used wavelength range, they can be randomly chosen or they can be located at selected wavelengths known to be relevant absorption wavelengths for bitterness components. An example of such a relevant wavelength is 275 nm or 430 nm. In those cases where several subsections are used in the determination of bitterness, the predetermined number of single wavelengths can be evenly distributed among the subsection or unevenly distributed. Furthermore, the wavelengths used in a subsection can also be either uniformly distributed within the subsection or non-uniformly distributed concentrated in an area of the subsection. Furthermore the distribution within a given subsection can differ from any other subsections.

By concentrating the single wavelengths between pre-selected subsections and/or within certain subsections it is possible to compensate for difficult measuring conditions e.g. in a range with heavy background absorption. Another relevant measuring area can be used to increase overall precision when combined with a few wavelengths in a relevant measuring area, or concentrating wavelengths in a more optimal area can be used to reduce the number of wavelengths needed to perform an acceptable determination of bitterness, i.e. reducing the amount of data to be analyzed.

In an even more preferred embodiment of the invention the UV-VIS spectra are measured at more than one wavelength chosen freely in the entire UV- VIS range. More preferably the wavelength is chosen in one or more narrow ranges such as 250 to 320 nm.

Analysis of absorption spectra by chemometrics/multivariate data analysis

Routine bitterness analysis of brewing samples, according to the invention, is performed by predicting the bitterness from the recorded multivariate absorbance measurement. The developed multivariate calibration model is used for the prediction of BU in new samples by inner product vector multiplication of the recorded spectrum and the regression coefficients estimated from spectra for beers with known bitterness content. In the total prediction time, the time used for the spectral acquisition is the limiting factor, which is within seconds for fast modern instrumentation. Since the detected spectral measurement is multivariate it is possible to implement outlier detection for e.g. deviating samples or measurement errors. The outlier detection can be based on a multitude of different statistical and mathematical tests most of which are based on analysis of residuals and/or Mahalanobis distances. The possibility of outlier detection is an advantage of the present method compared to the traditional method which makes the present method suitable for process monitoring and trouble shooting.

The samples used for the multivariate calibration modeling should span the variation expected in the samples to be analyzed i.e., if the pH, color or temperature are expected to vary among the samples, the calibration samples on which the model is build should span the same variations. The outlier detection facility of multivariate methods can reveal samples with variations that are not contained in the ^"calibration samples. Possible data pre-transformations, such as 1^st and 2^nd derivatives or ratio values of the UV-VIS spectra can be used for further optimization of calibration models. Optionally, calibration models can also be optimized by developing separate models for separate beer types, colors, pH or composition.

Model development: Calibration.

The multivariate regression model can provide a set of regression coefficients that can be used for prediction of the bitterness concentration in a new sample from the measured UV-VIS spectrum. The calibration can be accomplished by, but is not limited to, the following steps:

The recorded spectra are arranged in a matrix designated X with the dimension n samples x v variables, and the corresponding reference measurements are arranged in a vector called y with the dimension n samples x 1.

The equation to be solved during the calibration step is

y = Xb Eq. 1

where y and X is described above and the vector b (dimension 1 x v) contains the regression coefficients to be estimated.

A multitude of multivariate methods exist that provide solutions for this equation and in this case we focus on the PLS model.

In Step 0 a pre-transformation (if any) of X is performed and X is autoscaled or mean centered. If necessary further data transformations can be performed in this step (e.g. derivatives, scatter corrections, baseline corrections, non-linear transformations etc.). Step 1 in the PLS modelling is to estimate the first loading weight vector w (v x 1) in the equation

X = yw' Eq. 2

where y and X are described above and w has the dimension v x 1. The least squares solution to Eq. 2 is

w = X'y(y'yV¹ Eq. 3

w is then normalised to length one: w=w/|w|.

In Step 2 the corresponding first score vector, t (n x 1 ), in

X = tw' Eq. 4

is estimated by the least squares solution

t = Xw(w'w)^"1= Xw Eq. 5

Then in Step 3 the first loading vector, p (v x 1 ), in

X = tp' Eq. 6

is estimated by the least squares solution

p = X't(t't)-¹ Eq. 7

The relation between X and y for the first PLS factor is expressed in the equation (also called the inner relation)

y = tbinner Eq. 8 where bj_nner is a scalar and is estimated as

Dinner = t'y(t't)-¹ Eq. 9

Then the original data are reduced as given in the following equations

X_new = X - tp' Eq. 10

Vnew = y - tbinner Eq. 11

and the second set of vectors (w, t, p) and b_inner for the second PLS factor can be calculated by going through Eq. 1 to 9. This is repeated A times in order to extract A PLS factors.

The relation between b in Eq. 1 and binner is given by

b = W(P'W)^"1 b_inner Eq. 12

where W contains the A w vectors, P contains the A p vectors and bj, contains the A binner scalars for the A factors.

Model development: Prediction

When b is estimated it can be used for the bitterness content prediction of new samples. In the case of a new spectral measurement of an unknown beer sample, the recorded spectrum can be designated x'un_kn_own (1 x v). The prediction can then be formed as a two step procedure:

In Step 0 the same pre-transformation as for the calibration data is performed (if any) and the new spectrum is autoscaled or mean centered in the same way as the calibration data. Also transformations like derivatives, scatter corrections, baseline corrections, non-linear transformations are preferably performed comparable to the transformations performed during calibration. In Step 1 the actual prediction of beer bitterness is performed

Vunknown ⁼ X unknownb Eq. 13

where b is given in Eq. 12.

Model development: Validation

The estimation of the regression coefficients is called the calibration step, and the test of a developed model on new beer samples is called the validation step. A standard method for evaluating the performance of the model is the cross validation technique. In cross validation (CV) one or several samples are excluded from the data set. Then a regression model is calculated on the remaining samples (Calibration) and this model is used for the prediction of the excluded samples (Prediction). This procedure is repeated until all samples have been excluded once. In the special case where one sample is excluded at a time, the method is called full CV or leave-one-out CV. The prediction error obtained by CV is used as a measure of how well the model performs. The equation used for calculation of the Root Mean Square Error of Cross Validation (RMSECV) is

RMSECV =

where BU_re erence is the BU value measured by the EBC method (European Brewery Convention, 1998, Analytica-EBC, 7.8. Nϋrnberg, Germany), for sample i, B\J_predicted is the BU value predicted by the invented method and n is the number of samples used in the modeling.

Example 1 - BU reference determination of beers Beer samples Twenty-one beers with different bitterness levels were investigated. Bitterness values from 3.1 to 35 BU with a mean of 18.2 BU and a standard deviation of 8.3 BU are represented, and major differences in beer colours are also represented in the sample set including sixteen light beers and five darker beers. All the beers have a pH value around 4.5. All BU analyses were performed one day before the spectroscopic analyses, and performed according to EBC standards (European Brewery Convention, 1998, Analytica-EBC, 7.8. Nϋmberg, Germany).

Example 2 - Spectral analysis of the same beer samples Spectral method

The beers were degassed and their temperature was controlled to 25°C. 100 μl degassed beer and 2.90 ml water (also thermostated) were mixed in a 10 mm quartz cuvette. UV-VIS absorbance was measured at wavelengths from, and including, 240 nm to 600 nm, at intervals of 0.5 nm. Water was used as a reference sample. The cell path length was 10 mm, but it is possible to measure directly on the undiluted samples with no sample dilution by using a smaller path length e.g. 0.1 to 5 mm.

UV-VIS spectra of beer samples

In Figure 1 the UV-VIS absorption spectra from all 21 beers are shown. An obvious distinction between dark and light beers is observed as the dark beers absorb more light than the light beers. However, all beers were included in the models developed.

Example 3 - PLS model of data from beer UV-VIS spectra and BU reference data

PLS regression models were developed between the obtained UV-VIS absorbance spectra and the measured BU values. Models were validated by full cross validation. The PLS model was developed using 2^nd derivatives of spectral data in the wavelength range of 240-600 nm. In Figure 2, a Predicted versus Measured plot for the developed model using six PLS components is shown. The average error measured as RMSECV is 1.8 BU and this is the error to be expected when predicting the BU of new samples. The model is developed on a relatively small sample set and thus further improvement in the prediction are possible by analyzing a larger sample set including wider variation among the analyzed samples.

Example 4 - PLS model of data from beer fluorescence spectra and BU reference data

PLS regression models were developed between the obtained fluorescence spectra and the measured BU values. Models were validated by full cross validation. The PLS model was developed on fluorescence data obtained by sample excitation at 400 nm and measuring emission from 420-600 nm, according to the conditions described in WO 95/21242. In Figure 3, a Predicted versus Measured plot for the optimal model using four PLS components is shown. The average error measured as RMSECV is 5.4 BU and this is the error to be expected when predicting the BU of new samples.

In conclusion, the measurement of bitterness of a brewing sample using the method of the present invention provides a more accurate measure of bitterness than a method based on fluorescence spectra. The expected error of the method of the invention, according to example 3, was found to be 1.8 BU for the 21 beer test samples, while the expected error for the method of used in Example 4 for the same samples, was 5.4 BU.

Example 5 - Off-line apparatus for bitterness analysis

A diagram of an offline apparatus is shown in Figure 4. The apparatus contains a UV-VIS light source (1 ), a detector (2) and a unit for data acquisition, spectrometer control and the quantitative measurement of bitterness content (5). A sample container can be placed in the apparatus in a receiving compartment (3) positioned in the light path between the light source (1 ) and light detector (2). The sample container can be a quartz cuvette or vial, or a stop-flow or flow-through cell and has a width that is equal to the optical path length. In a preferred embodiment the unit for data acquisition (5) is an integrated part of the apparatus. In another preferred embodiment the apparatus has an inlet and/or outlet (4) for the sample.

Example 6 - On-line apparatus for bitterness analysis

Figure 5 shows the components of an on-line apparatus. The apparatus contains a UV-VIS light source (17), means for transferring light (16), a measuring probe (15), a detector (18) and a unit for data acquisition, spectrometer control and the prediction of bitterness (19). The brewing sample flow (process stream) (13) is illuminated from the measuring probe (15). Transfer of light to and from the measuring probe (15) can be conducted using optical fibres (16). The measuring probe (15) can be designed as a transmission, transflectance or reflectance probe. The spectrometer and data processing unit can be situated in an area (12) that is safe and remote from the area (11 ) where the measuring probe is located in the process stream.

Claims

Patent Claims:

1. A method for performing a quantitative determination of the bitterness content of a brewing sample comprising the steps of:

2. A method according to claim ^characterized in that the brewing sample is subjected to sample preparation before said measuring.

3. A method according to claim ^characterized in that the brewing sample is used directly in said measuring.

4. A method according to claims 1 or 2, characterized in that the brewing sample is undiluted before said measuring.

5. A method according to any one of claims 1-4, characterized in that said measuring is performed directly in the brewing process stream, beer pipeline, or beer tank.

6. A method according to any of the preceding claims, characterized in that a full spectrum (about 190 nm to about 800 nm) is used in said measuring.

7. A method according to claim 6, characterized in that said full spectrum is measured as a discrete spectrum.

8. A method according to any of the preceding claims, characterized in that absorbance is measured in one or more predetermined subsection in the region 190 to 800 nm.

9. A method according to claim 8, characterized in that one or more of said subsections are selected from the group consisting of: 250 to 320 nm,

400 to 430 nm, and 650 to 700 nm.

10. A method according to any one of the claims 7-9, characterized in that said discrete spectrum is measured in steps of variable size or equal size.

11. A method according to claim 10, characterized in that the size of said steps are within the range 0.001 nm to 20 nm, preferably 0.01 nm to 5 nm, more preferably 0.02 nm to 2 nm and most preferably 0.1 nm to 1 nm.

12. A method according to any one of the claims 7-11, characterized in that said measured discrete spectrum has a bandwidth of variable size or equal size.

13. A method according to claim 12, characterized in that said bandwidth is within the range 0.001 nm to 20 nm, preferably 0.01 nm to 5 nm, more preferably 0.02 nm to 2 nm and most preferably 0.1 nm to 1 nm.

14. A method according to any one of the preceding claims, characteri z e d in that said absorbance is measured at a number of discrete wavelengths.

15. A method according to claim 14, characterized in that the number of wavelengths is uniformly distributed within said full spectrum and/or said predetermined subsections.

16. A method according to claim 14 or 15, characterized in that said number of wavelengths is unevenly distributed amongst said subsections and uniformly within each said subsections or visa versa.

17. A method according to claim 14 or 15, characterized in that said number of wavelengths in unevenly distributed amongst said subsections and non-uniformly distributed within each of said subsections.

18. A method according to claim 14 to 17, characterized in that the number of wavelengths used are from 2 to 100, preferably from 2 to 20, more preferably from 2 to 10, most preferably 5.

19. A method according to any of the preceding claims, characterized in that the regression analysis is a chemometric method, preferably a multivariate chemometric method.

20. A method according to any of the preceding claims, characterized in that the brewing sample is beer.

21. An apparatus for quantitative determination of the bitterness content of a brewing sample, comprising a data processing unit (5, 19) with a program code means adapted to: - receive data comprising absorbance of said brewing sample measured at two or more wavelengths in the range of about 190 nm to about 800 nm, and - compare by means of regression analysis said received data with stored data comprising stored multivariate absorbance measurements of brewing samples with known bitterness content, and - make a quantitative determination of the bitterness content of said brewing sample, when the program code means are executed on the data processing unit.

22. An apparatus according to claim 21 , further comprising a measuring unit adapted to: - irradiate the brewing sample at two or more wavelengths in the range of about 190 nm to about 800 nm (1 ,17), and - measure absorbance at said two or more wavelengths by said brewing sample (2,18), and - transfer data comprising said measured absorbance to said data processing unit (5, 19).

23. An apparatus according to claim 22, wherein said measuring unit is further adapted to: - transfer light to and from said brewing sample (16).

24. An apparatus according to claim 22 or 23, c h a r a c t e r i z e d in that said measuring unit measures a full spectrum (about 190 nm to about 800 nm).

25. An apparatus according to claim 24, c h a r a c t e r i z e d in that said measuring unit measures said spectrum as a discrete spectrum with varying step size and/or bandwidth.

26. An apparatus according to claim 25, c h a r a c t e r i z e d in that said step size and/or bandwidth is within the range 0.001 nm to 20 nm, preferably

0.01 nm to 5 nm, more preferably 0.02 nm to 2 nm and most preferably 0.1 nm to 1 nm.

27. An apparatus according to claim 22 or 23, c h a r a c t e r i z e d in that said measuring unit measures absorbance at two or more wavelengths optionally with a different bandwidth.

28. An apparatus according to any one of claims 22 to 27, c h a r a c t e r i z e d in that said measuring unit comprises a light source for irradiating the sample selected from the group consisting of a diode, laser, Xenon lamp, deuterium lamp, halogen lamp, tungsten lamp, Nernst glower and hollow cathode lamp.

29. An apparatus according to any one of claims 22 to 28, c h a r a c t e r i z e d in that said measuring unit comprises a wavelength selection means selected from the group consisting of a prism, grating and filter.

30. An apparatus according to any one of claims 22 to 29, c h a r a c t e r i z e d in that said measuring unit comprises a light detector to measure absorbance selected from the group consisting of a photodiode, photomultiplier, phototube, photocell and photoconductor.

31. An apparatus according to any one of claims 23 to 30, c h a r a c t e r i z e d in that said measuring unit comprises an optical fibre or light pipe to transfer light to and from said brewing sample.

32. An apparatus according to any one of claims 22 to 31, c h a r a c t e r i z e d in that said apparatus comprises a chip or a computer for recording said measured absorbance.

33. An apparatus according to any one of claims 21 to 32, c h a r a c t e r i z e d in that said chip or computer is an integrated part of the measuring unit.

34. Use of an apparatus according to any one of claims 21 to 33 for in-line, on-line, at-line or off-line quantitative measurement of the bitterness content of a brewing sample.