Method for determining the quality of molasses used in yeast production
Patent Information
- Application Number
- ES2022792855T
- Authority / Receiving Office
- ES · ES
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2022-09-23
- Publication Date
- 2026-11-13
- Estimated Expiration
- 2042-09-23
AI Technical Summary
Current methods for determining the quality of molasses used in yeast production are inadequate, leading to significant losses due to poor production yield or quality, as quality assessment is only possible after initial production, and variations in molasses quality are difficult to anticipate.
A method using machine learning to construct a statistical model from reference optical spectra of molasses samples, allowing for the identification of spectral features that correlate with yeast performance, enabling the determination of molasses quality and potential yeast performance before production.
Enables the pre-qualification of molasses based on its optical spectrum, allowing for the discard of non-compliant batches and adjustment of production processes to achieve target yeast performance levels, thereby reducing production losses.
Smart Images

Figure 00000015_0000 
Figure 00000015_0001 
Figure 00000016_0000
Abstract
Description
Domaine technique
[0001] This disclosure falls within the domain of yeast production, and more specifically, it relates to a method for determining the quality of molasses used as a nutrient source in yeast production, this molasses quality being relative to one or more performances of the yeast produced.
[0002] The term "yeast," in the singular or plural, is a generic term designating eukaryotic microorganisms capable of fermenting organic matter. Yeasts are commonly used, both by individuals and in the food and pharmaceutical industries. For example, they can be used in the production of wine, beer, or leavened doughs (e.g., bread). Among the yeasts, the following genera are a non-exhaustive list: Saccharomyces, Candida, Pichia, Kluyveromyces. The term "yeast strain" refers to a relatively homogeneous population of yeast cells. A yeast strain is obtained from the isolation of a clone. A clone gives rise to a population of cells derived from a single yeast cell.
[0003] Yeast is generally produced by multiplying a strain cultivated in a medium containing a set of raw materials. In an industrial setting, starting with a given quantity of yeast, it is possible to produce several tons. The raw materials typically used to feed yeast production are a set of nutrients and a source of sugar(s), which can be sugar syrup, molasses, or a mixture of sugar syrups and / or molasses.
[0004] Molasses is a raw material from the sugar industry and can be distinguished by two types: beet molasses and cane molasses. Its nutrients can include sources of nitrogen, phosphorus, minerals, trace elements (e.g., iron, zinc, etc.), vitamins (e.g., B, C, etc.), water, and even air.
[0005] During yeast production, several factors can impact the yield or quality of the yeast produced. These factors include the quality of the molasses used and the proportions of nutrients supplied to the yeast. Depending on the molasses quality, other factors, or a combination of factors including molasses quality, it is possible to adjust production methods, such as the respective quantities of each nutrient, to optimize production yield and / or the quality of the final yeast.
[0006] Currently, the quality of molasses not yet used in production, as well as the quality of the yeast produced from it, can only be determined after either an initial yeast production followed by yeast qualification protocols, or a series of miniaturized laboratory tests that allow for a rough estimation of quality and the elimination of atypical molasses grades. Indeed, assessing the quality of unknown molasses or compensating for known but poor-quality molasses by adjusting production processes can only be done after the yeast obtained from that molasses grade has been qualified.
[0007] Given that an initial production run can involve several hundred tons of yeast produced and may have required the commissioning of several production sites, obtaining a poor production yield or poor quality yeast can represent a significant loss for an industrialist.
[0008] In parallel, the organization of molasses production into "sugar campaigns" means that the quality of molasses received over time is subject to significant variations, and it is not possible to anticipate the impact of these variations.
[0009] These variations in quality and their impact on yeast quality are all the more difficult to anticipate as molasses is stored in tanks that can contain quantities usable for several weeks of production, which makes it even more difficult to detect a batch of molasses with atypical quality that may have been transferred into the storage tank.
[0010] Therefore, there is a need to determine the quality of molasses before its use in yeast production in order to avoid the production of yeast with non-compliant quality and / or performance (e.g., with specifications).
[0011] The document CHAPANYA PISITTINEE ET AL: "Temperature Compensation on Sugar Content Prediction of Molasses by Near-Infrared Spectroscopy (NIR)", SUGAR TECH, SPRINGER INDIA, NEW DELHI, vol. 21, no. 1, July 3, 2018 (2018-07-03), pages 162-169, describes a method for determining, using NIR spectroscopy and predictive models, the fermentable sugar content of molasses in order to optimize ethanol yield.
[0012] This disclosure improves the situation.
[0013] A method is thus proposed, according to claim 1, for qualifying molasses from an optical measurement, the method possibly comprising: from reference optical spectra of distinct molasses samples, associated with respective known molasses grades, construct by machine learning a statistical model of molasses grades as a function of at least one spectral feature of the reference spectra, and for a current optical spectrum of a sample of a current molasses, from the statistical model, identify said at least one spectral feature of the current optical spectrum and determine a grade of the current molasses, said current molasses grade being relative to at least one yeast performance obtained when said yeast is fed with said current molasses, in which each molasses grade associated respectively with a distinct molasses sample is relative to at least one reference yeast performance obtained when said reference yeast is fed with said distinct molasses sample.
[0014] Advantageously, it is thus possible to determine the quality of an unknown molasses from its optical spectrum, and to deduce at least one yeast performance of the yeast produced from this unknown molasses for given production conditions (e.g. temperature and pressure in the fermenter) and a given yeast strain.
[0015] Thus, advantageously, molasses that does not meet the quality requirements for producing yeast with one or more target performance levels can be discarded before any use in production. Another advantage is that, based on the determined molasses quality, production processes can be adapted, for example, by adjusting the proportions of necessary nutrients to correct poor-quality molasses and achieve the target performance level(s). Furthermore, depending on the degree of poor quality, it may be possible to improve this quality by blending it with molasses of varying qualities.
[0016] In one or more embodiments, said at least one performance relating to a molasses quality associated respectively with a distinct molasses sample is relative to a reference yeast obtained when said reference yeast is fed with said distinct molasses sample. In one or more embodiments, the molasses quality of a distinct molasses sample is defined by at least one molasses quality score, said at least one molasses quality score being relative to said at least one yeast performance.
[0017] Reference yeast may be understood to mean a reference yeast used in a qualification protocol which has determined at least one performance of said reference yeast, said reference yeast having been fed with said separate molasses sample.
[0018] The features described in the following paragraphs may optionally be implemented. They may be implemented independently or in combination with each other.
[0019] According to one or more embodiments, said at least one performance is based on at least one criterion chosen from: a capacity for multiplication of yeast fed by said common molasses, a capacity for raising bread, a capacity for preservation, a capacity for drying, a capacity for fixing structural elements, a capacity for assimilating nutrients, a capacity for adapting to stresses.
[0020] The capacity for multiplying yeast fed with said common molasses can be understood as a production yield per unit mass of common molasses used or a ratio of quantity of yeast produced to quantity of sugar used in the production process.
[0021] Structural elements can be understood to include elements such as nitrogen, phosphorus, carbon, hydrogen, oxygen, or sulfur.
[0022] In one or more embodiments, the quality of the current molasses determined by said statistical model is defined by at least one molasses quality score, said at least one molasses quality score being relative to said at least one performance.
[0023] In one or more embodiments, rather than using a score or in combination with a score, the quality of the current molasses determined by said statistical model may be defined by a class such as good, bad, average, very good, green light to use this current molasses, red light to reject, or modify this current molasses, etc., the class being relative to at least one performance of the yeast obtained when fed with this current molasses.
[0024] In one or more embodiments, the process may further include: illuminate a molasses sample with a first emitted light signal interacting with the sample, and collect a second light signal, resulting from the interaction between the first light signal and the current sample; and measure the second light signal, in chosen frequency bands, to establish an optical spectrum representative of a chemical signature of the current molasses sample.
[0025] In one or more embodiments, the established optical spectrum and reference optical spectra can be measured under similar conditions of illumination and measurement of light signals.
[0026] Thus, for example, in one embodiment, the statistical model can be determined by implementing machine learning, for example from a random forest to arrive at a set of decision trees defining the statistical model.
[0027] In one or more embodiments, the optical spectrum may include a set of wavenumbers, one or more wavenumbers from the set of wavenumbers being representative of a chemical compound of molasses.
[0028] According to one or more embodiments, molasses may comprise compounds including at least one sugar, one mineral, one vitamin, and the optical spectrum comprises a plurality of wavenumbers, and in which at least one wavenumber characterizes the presence of one of said compounds.
[0029] According to one or more embodiments, said at least one spectral characteristic can be determined according to at least one criterion chosen from among the presence or absence of a predefined wavenumber, a position of a predefined number, an intensity value of a predefined wavenumber, an exceedance of the threshold of an amplitude of a predefined wavenumber, a maximum or minimum intensity value of a predefined wavenumber, an intensity difference between two predefined wavenumbers, an intensity of a predefined wavenumber between two predetermined intensity values, or simple or complex combinations of the aforementioned spectral characteristics.
[0030] In one or more embodiments, said at least one spectral characteristic can be determined according to at least one criterion chosen from among the presence or absence of a predefined line, the position of a predefined line, the value of a predefined line amplitude, the exceeding of a threshold of a predefined line amplitude, the maximum or minimum value of a predefined line amplitude, the amplitude difference between two predefined lines, the amplitude of a line between two predefined amplitude values, or simple or complex combinations of the aforementioned spectral characteristics.
[0031] According to one or more embodiments, the optical measurement can be carried out by at least one technique from among RAMAN spectroscopy, IR spectroscopy, or THZ spectroscopy.
[0032] According to one or more embodiments, the range of values of the reference optical spectra can be between 1 cm⁻¹ and 2500 cm⁻¹.
[0033] According to one or more embodiments, the range of values of the reference optical spectra can be between 436 cm⁻¹ and 1700 cm⁻¹.
[0034] According to one or more embodiments, molasses is cane molasses or beet molasses or a mixture of the two, or a mixture of sugar products such as molasses and sugar syrup, or a mixture of products of which one of the components is molasses.
[0035] This disclosure also relates to a computer program, according to claim 12, comprising program code instructions for carrying out the process of this disclosure when the program can be run on a computer.
[0036] This disclosure also relates to a non-transient, computer-readable recording medium, according to claim 13, on which is recorded a program for implementing the method of this disclosure when that program is executed by a processor.
[0037] This disclosure also relates to a process, according to claim 14, for the production of yeast by fermentation, which may include: determine a respective molasses grade for each molasses from a plurality of molasses supplied as input to the fermentation reaction according to the molasses qualification process described in this disclosure, and feed the fermentation reaction with each molasses from the plurality of molasses whose respective molasses grade is above a predetermined grade, and for all or part of the molasses having a molasses grade below the predetermined grade, feed the fermentation reaction after modification of the molasses, and / or reject the molasses.
[0038] By modifying the molasses, in one or more embodiments, at least one modification chosen from the following may be considered: mix molasses of lower quality than predetermined, with another molasses of a distinct quality, and so that the mixture of molasses has a quality higher than predetermined, add one or more adjuvants so that the molasses improved by the adjuvants has a quality higher than predetermined.
[0039] The respective molasses quality of each molasses of the plurality of molasses and which is relative to at least one yeast performance obtained can be determined for identical fermentation conditions (e.g. fermenter temperature and pressure, fermenter type, etc.) and for the same yeast strain. Brève description des dessins
[0040] Other features, details, and advantages will become apparent upon reading the detailed description below and analyzing the attached drawings, on which: Fig. 1 [ Fig. 1 ] describes schematically, in one or more embodiments, an example of the optical spectrum of a molasses. Fig. 2 [ Fig. 2 ] illustrates a method for determining the quality of a molasses from an optical measurement in one or more embodiments. Fig. 3 [ Fig. 3 ] illustrates an example of the application of the statistical model obtained according to the method of this disclosure on spectra of unknown molasses. Fig. 4a [ Fig. 4a ] illustrates the random forest principle for a statistical model in one or more embodiments. Fig. 4b [ Fig. 4b ] illustrates the principle of random forest for a statistical model in one or more embodiments. Fig. 5 [ Fig. 5 ] presents different spectra of cane molasses of known quality obtained by RAMAN spectroscopy. Fig. 6 [ Fig. 6 ] presents a device for implementing the process of this disclosure according to one or more embodiments. Description des modes de réalisation
[0041] The terms "peak" and "line" may be interpreted in the same way and are interchangeable in this disclosure. Similarly, the terms "intensity" and "amplitude" may be interpreted in the same way and are interchangeable in this disclosure.
[0042] There figure 1 describes schematically, in one or more embodiments, an example of the optical spectrum of a molasses.
[0043] Numerous optical measurement techniques exist for obtaining an optical spectrum, such as infrared, microwave, terahertz, and Raman spectroscopy. All these techniques rely on the same principle: sending an electromagnetic wave (i.e., photons) onto the material being studied while varying the wave's frequency. A resulting wave from the wave-matter interaction can then be collected, either in transmission or reflection, for each frequency of the electromagnetic wave. From this resulting wave, it is possible to determine an optical spectrum representative of the constituents of an organic material (e.g., molasses), that is, a chemical signature of the molasses.
[0044] An optical spectrum can be represented as a set of peaks, where each peak can be associated with a position (e.g., abscissa) and an amplitude (e.g., ordinate) within the spectrum. The abscissa of an optical spectrum can be expressed in frequency (Hz), wavelength, or wavenumber (cm⁻¹). The ordinate can be expressed in an arbitrary unit, which may depend on the signal processing applied to the measurement data. Indeed, to ensure proper usability, it is common practice to apply one or more data processing steps to the raw data (i.e., data from optical measurements) to allow for comparisons between spectra, for example.
[0045] For example, in the case of RAMAN spectroscopy, a baseline, a derivation of this baseline, or a combination with data normalization can be performed to remove certain peaks exhibiting aberrations or whose presence results from the laser used for spectroscopy.
[0046] Each optical measurement technique can correspond to a specific range of frequencies (or wavelengths or wavenumbers) in which the frequency of the incident wave varies. For example, the wavelength range (expressed in microns) can be between [1 µm–20 µm], [0.3 µm–0.7 µm], or [1 cm⁻¹–2500 cm⁻¹]. In one or more embodiments, this frequency range can be between [436 cm⁻¹–1700 cm⁻¹].
[0047] Depending on the optical measurement technique and the wavelength range used to obtain a spectrum of organic matter, the position of a peak can correspond, for example, to an atomic or chemical bond, a functional group of an organic molecule, or even the vibrations of a molecule. Several peak positions can therefore correspond to the presence of the same molecule (or constituent, for example, a vitamin) in organic matter, reflecting, for example, the chemical bonds or functional groups of that molecule.
[0048] The amplitude of a peak can have different meanings depending on the optical measurement technique used and the data processing applied to the raw optical data.
[0049] Optical measurement techniques for spectroscopying a medium are generally complementary, and the same medium can be measured by different optical measurement techniques (e.g., different types of spectrometers) in order to obtain a relatively complete map of that medium.
[0050] Depending on the measurement techniques used, the interpretation of spectra can be more or less complicated, both visually and according to classical methods of spectral data processing. Indeed, an optical spectrum can comprise a multitude of peaks (or lines), some of which can be perfectly discretized 102;103, while others can be so close together that they form a cluster 105;107, making their interpretation difficult. Furthermore, depending on the resolution used (e.g., the optical measurement step size), a peak, based on its full width at half maximum (FWHM), can spread more or less extensively across the optical spectrum, thus further complicating data interpretation.
[0051] Furthermore, while a change in the optical measurement technique used may seem obvious when the obtained optical spectrum is unusable (e.g., a 107 line cluster), using a different optical measurement technique can lead to obtaining an optical spectrum that is completely or partially different from the initially obtained spectrum. This is because the compounds / molecules of interest may not be sensitive to the wavelengths used by an alternative optical technique.
[0052] Therefore, it can be crucial to be able to exploit an optical spectrum, regardless of the configuration of the peaks in the spectrum, and more particularly when interpretations and comparisons between a large number of optical spectra obtained from samples that are relatively close in their chemical / molecular fingerprint are sought.
[0053] There figure 2 illustrates a method for determining the quality of molasses from an optical measurement in one or more embodiments.
[0054] From optical spectra (Spect_ref1... Spect_refn, n being a positive integer) 201 of reference samples of distinct molasses which are known and which are associated respectively with known qualities of molasses (Q1, Q2, Qn), a statistical model of qualification 203 of molasses as a function of at least one spectral characteristic of the reference spectra can be built by machine learning.
[0055] Machine learning can consist of determining, in reference spectra of molasses, spectral configurations leading to a certain quality of molasses, in particular certain qualities of molasses specific to obtaining one or more desired performances for a yeast for example.
[0056] Based on the determined spectral configurations, machine learning can then construct one or more rules which, together, constitute a statistical model correlating one or more spectral characteristics of the spectrum to a molasses quality, this molasses quality being relative to one or more performances of a yeast produced using this molasses quality.
[0057] Reference spectra and the respective qualities of molasses associated with those spectra can serve as input data for machine learning, and each reference spectrum can be representative of a specific (e.g., chemical) structure of a molasses (e.g., cane or beet). Furthermore, each molasses quality presented as input data and associated with a reference spectrum can be relative to at least one yeast performance (e.g., reference yeast) obtained when that yeast is fed with that molasses.
[0058] In addition, in one or more embodiments, the quality of molasses can be defined by at least one respective Sc n score and / or according to a molasses class (e.g. good, bad, average, red light, green light, etc.) and / or a correlation between its spectral profile and a statistical group of industrial performance.
[0059] In one or more embodiments, the molasses qualification score may relate to the impact of the quality of a molasses, when used as a nutrient source in yeast production, on the performance or performances of the yeast obtained.
[0060] For example, the score for each molasses associated with a reference spectrum can range from 0 to 1 (e.g., 0 = poor quality and 1 = good quality) or from 1 to 10. A value of 1 might represent poor quality molasses, and a value of 10 might represent very good quality molasses. Thus, a score (or score value) greater than 8 could correspond to a molasses that produces yeast with very good performance, for example, excellent leavening power for bread dough and / or excellent multiplication capacity (i.e., a high yield of yeast).
[0061] In one or more embodiments, molasses quality can be defined according to a range of values. For example, a score between 0 and 3 may correspond to poor quality molasses, a score between 3 and 5 may correspond to medium quality molasses, a score between 5 and 8 may correspond to good quality molasses, and a score greater than 8 on a scale of 10 may correspond to very good quality molasses.
[0062] In another example, in one or more embodiments, the quality of molasses can be defined by a plurality of scores, each score being relative to a respective performance. For example, a reference optical spectrum can be associated with a first score relating to a bread-raising capacity, and a second score relating to a multiplication capacity (i.e., yeast production yield).
[0063] The scores can be combined to provide actionable information for staff involved in yeast production within an industry or testing laboratory.
[0064] The score of each molasses associated with a reference spectrum may have been previously determined based on one or more qualification protocols.
[0065] As an example, a molasses qualification protocol may involve implementing this molasses with a reference yeast to determine the performance of the reference yeast in terms of cell multiplication and enzymatic activity.
[0066] For example, performance based on (or related to) bread-raising ability can be assessed by the ability of the prepared yeast (e.g., prepared reference yeast) to produce carbon dioxide, and / or can be assessed by the fermentative activity of the yeast.
[0067] Fermentation activity is determined by measuring the release of carbon dioxide gas (expressed in absolute terms by volume or in relative terms by percentage) from a given dough piece using, for example, a Burrows and Harrison fermentometer described in the "Journal of Institute of Brewing", vol. LXV, No. 1, January-February 1959 or a Risograph according to the protocol described by Rattin et al., 2009 (Cereal Foods World, 54 (6): 261-265).
[0068] In one or more embodiments, a molasses quality score (Scn) greater than 5 (i.e., corresponding to good quality molasses, for example) may relate to a bread-raising capacity (i.e., yeast performance) compared to that generally observed in yeasts produced in a reference process. Similarly, a molasses quality score (Scn) less than 3 (i.e., corresponding to poor quality molasses) may relate to a bread-raising capacity that allows for specific volume values lower than those generally observed in yeasts produced in a reference process.
[0069] The multiplication capacity of yeast fed on common molasses, also called production yield, can be expressed as the amount of yeast produced per amount of sugar committed, or as the amount of sugar committed per amount of yeast produced.
[0070] In one or more embodiments, a molasses quality score (Scn) greater than 5 (i.e., corresponding to good quality molasses, for example) may relate to a yeast multiplication capacity (i.e., yeast performance) generally observed in yeasts produced in a reference process. Similarly, a molasses quality score (Scn) less than 3 (i.e., corresponding to poor quality molasses) may relate to a yeast multiplication capacity that allows for values lower than those generally observed in yeasts produced in a reference process.
[0071] In one or more embodiments, the input data for the reference spectra for machine learning can be in the form of a set of wavenumbers respectively associated with an intensity (or amplitude) value. For example, the set of wavenumbers can be between 0 and 2000 cm⁻¹ and with a wavenumber step of 0.1 cm⁻¹.
[0072] There are many machine learning techniques. All machine learning methods can be used to learn the rules linking molasses characteristics and yeast performance. In one or more embodiments, machine learning can be based on the use of random forests, one or more linear regression algorithms, K-nearest neighbors, deep learning, or neural networks.
[0073] Furthermore, machine learning can include one or more hyperparameters as input data. A hyperparameter is a parameter used to configure the algorithm (e.g., random forest, deep learning, etc.) before any training. Hyperparameters can be specific to a type of machine learning and can therefore vary depending on the type of machine learning used.
[0074] As an example, a hyperparameter could be the number of decision trees composing a random forest, the number of nodes in a tree, the size of the final nodes, or the number of observations chosen by "bootstrap" to build each tree.
[0075] After the model has been determined, it can then be used to identify 205, for a current spectrum of a sample of current molasses in input data 207, at least one spectral feature of the current spectrum, and to determine a quality 209 of the current molasses.
[0076] Based on the determined quality of standard molasses, one or more performance characteristics of a yeast produced using standard molasses as a nutrient source can be determined. The quality of standard molasses can be represented according to a class (e.g., poor, average, good, very good, green light when usable, or red light when unusable or requiring modification, etc.) and / or according to one or more scores related to yeast performance, and / or a correlation between its spectral profile and a statistical group of industrial performance.
[0077] Of course, the yeast performance obtained when yeast is fed with standard molasses depends on several factors, such as the yeast strain used to produce the yeast, production conditions (e.g., pressure, temperature in the fermenter, fermenter configuration, etc.), and inputs, including the molasses used to feed the yeast. For the purposes of the process described in this disclosure, it is understood that the yeast performance(s) determined based on molasses qualities are determined for an identical yeast strain, identical production conditions, and identical inputs (except for the molasses).
[0078] In one or more embodiments, each reference spectrum of a respective molasses provided for machine learning may have been obtained by characterizing that respective molasses using one or more optical measurement techniques. For example, each reference spectrum may be representative of a combination of optical spectra of the same molasses obtained by a respective optical measurement technique.
[0079] In one or more embodiments, the optical measurement technique used on molasses to obtain the reference spectra is the same as that used on ordinary molasses to obtain the current spectrum. Furthermore, in one or more embodiments, the optical measurement technique used to obtain the spectra may be Raman spectroscopy.
[0080] In one or more embodiments, the data obtained after determining a molasses quality relative to a yeast performance according to the input spectrum 207 can be used to feed 210 into machine learning. Advantageously, this allows the statistical model to be continuously improved.
[0081] Of course, the statistical model may not need to be built every time an unknown molasses is required. The statistical model can be determined only once and then used with unknown molasses spectra.
[0082] The random forest methodology is described in more detail in the following publication: Leo Breiman, Random Forests, Machine Learning, vol. 45, no. 1, 2001, p. 5-32
[0083] As described previously, this disclosure is not limited to the use of random forests. Other types of machine learning may also be suitable, such as the machine learning methods described in the following publications: The Elements of Statistical Learning, 2nd edition, Trevor Hastie, Robert Tibshirani, Jerome Friedman, ISBN-13: 978-0387848570 Deep Learning, Lan Goodfellow, Yoshua Bengio and Aaron Courville, ISBN-13: 978-0262035613 .
[0084] There figure 3 illustrates an example of the application of the statistical model obtained according to the method of this disclosure on spectra of unknown molasses.
[0085] The respective quality (Q1; Q2) of two molasses M1 and M2 that could potentially be used as a source of nutrients in yeast production (e.g., yeast production by fermentation) can be investigated.
[0086] This quality can be sought in order to know whether M1 and M2 molasses can be used to feed a yeast production while allowing one or more target performances of the yeast produced to be achieved, and / or in order to determine the most suitable molasses to produce a yeast with one or more target performances, and / or according to a compromise of one or more target performances.
[0087] With reference to the figure 3 For example, a discretization between two molasseses, M1 and M2, can be sought to determine which molasses will yield a higher yeast production yield than typically observed with a reference yeast produced in a standard process, and / or produce a yeast that achieves specific volume values higher than typically observed with a reference yeast produced in a standard process. As another example, the discretization between two molasseses, M1 and M2, can be sought to determine which molasses offers the desired performance ratio, e.g., production yield / bread-raising capacity.
[0088] Each molasses M1; M2 can be associated respectively with an optical spectrum (Spect_X1 and Spect_X2), obtained for example by RAMAN spectroscopy.
[0089] Thus, applying the statistical model (Model_statist) of molasses quality to the optical spectra of molasses (Spect_X1 and Spect_X2), which differ from the reference spectra used for machine learning, can identify specific spectral characteristics. For example, the statistical model can determine a quality Q1 (and / or a score and / or a class) for the Spect_X1 spectrum based on the identification of spectral characteristics corresponding to the presence of wavenumbers R1, R2, and R3 coupled with respective amplitude thresholds THR1, THR2, and THR3 for each wavenumber. Similarly, the predictive model can determine a quality Q2 for the Spect_X2 spectrum based on the identification of spectral characteristics corresponding to the presence of wavenumbers R1, R2, and R4 coupled with a specific amplitude range Δ for wavenumbers R1 and R2.
[0090] Based on the determined qualities (e.g., according to a score), M1 molasses may appear to be a better candidate for supplying yeast production than M2 molasses. Indeed, M1 molasses has a score of 8, representative of a very good molasses, that is to say, allowing, for example, the obtaining of a yeast production yield higher than a value generally observed with a reference yeast operated in a reference process, and / or the obtaining of a good bread-raising capacity defined by specific volume values achievable higher than a value generally observed with a reference yeast operated in a reference process.
[0091] In one or more embodiments, for a plurality of molasses that could potentially be used to feed yeast production (by fermentation), molasses with a quality exceeding a predetermined standard (e.g., one or more scores) may be used to feed yeast production. In one or more embodiments, among the molasses with a quality exceeding a predetermined standard, priority for feeding yeast production may be given to the molasses with the highest quality. For example, the highest quality may be determined according to an ascending or descending order of the quality scores of the molasses or according to an ascending or descending order of the average of the respective quality scores of each molasses.
[0092] THE figures 4a et 4b illustrate the random forest principle for a statistical model in one or more embodiments.
[0093] The principle of the random forest algorithm is to determine, by majority vote for categorical data or by averaging individual results for quantitative data, the most probable prediction made by a multitude of decision trees constituting the random forest.
[0094] All decision trees in a random forest can be independent of each other, and the prediction determined by each decision tree before majority voting or averaging can be uncorrelated with the predictions determined by other decision trees in the random forest.
[0095] A first step in this algorithm might involve constructing decision trees. figure 4a presents such a decision tree which can, for example, be divided into three parts: a root node N1 corresponding to the first node of the tree (i.e. entry node), internal nodes N2; N3 which may have descendants (i.e. other nodes), terminal nodes N4; N5; N6; N7 (also called leaves).
[0096] Each intermediate "node," that is, nodes N2 and N3, can perform a test on a variable whose result indicates the branch to follow in the tree (i.e., N4, N5, N6, or N7). Thus, the test(s) performed at a node can define a partitioning rule, for example.
[0097] Nodes N1 to N3 can be defined by explanatory variables (e.g., of quantitative or qualitative type).
[0098] In order to construct the various decision trees such as the one presented in figure 4a , a selection of subsets of individuals from the training dataset with replacement can be carried out using a so-called "bootstrap" sampling method, which consists of randomly drawing (with replacement) individuals from the training dataset.
[0099] By "replacement," it can be understood that the candidates selected in one subset remain in the training set; that is, they can be selected for another subset. The individuals used to construct a decision tree are generally called "in bag," and the individuals used to evaluate the predictive power of the tree are generally called "out of bag" (OOB).
[0100] This training dataset can itself be obtained by implementing a cross-validation strategy consisting of separating the initial dataset (e.g. 70 / 30) into two sub-datasets, one for training (e.g. 70%) and the other for testing (e.g. 30%) the prediction model.
[0101] Each decision tree can then be built (or trained) on a respective subset of individuals. That is to say, there can be as many subsets of individuals as there are decision trees, and the tree construction is only carried out on a single subset of individuals.
[0102] In the case of this disclosure, the initial dataset may be a set of reference optical spectra (e.g., in the form of wavenumber sets) associated with a respective molasses quality (or / and one or more respective molasses scores) that may be relative to at least one yeast performance (e.g., reference yeast) obtained when the yeast is fed with that molasses, and the respective variable used in testing each node N1 to N3 may be, for example, an intensity (or amplitude) of a specific wavenumber.
[0103] During training (or tree construction), at each new node construction, the variable used for partitioning at the node level (into two descendant nodes) can be determined by the machine learning algorithm from the random subset of the training dataset. For example, the machine learning algorithm can review (e.g., according to the discretization step of the optical spectra) all (or a subset) of wavenumbers from the reference optical spectra in the random subset to determine a wavenumber and / or intensity to use as a variable. More precisely, the algorithm can try to determine if there is, for example, an intensity value that allows for the purest possible separation of spectra between those corresponding to high-quality and those corresponding to low-quality molasses.As an example, the test associated with node N2 may be to determine whether the amplitude (or intensity) associated with a particular wavenumber is above an intensity threshold or within an intensity range.
[0104] The choice of one variable over another can be based on the measurement of the purity of the node using, for example, the Gini index or Shannon entropy.
[0105] Thus, none of the input data provided for machine learning influences its choice of variables for partitioning at each node. Random forest machine learning can determine its own variables (e.g., wavenumber, amplitudes, etc.) to build a statistical model, without any prior assumptions about the presence of one or more wavenumbers that should be detected in the reference spectra. For example, node N3 might include a test based on the intensity value of a wavenumber that does not characterize any specific spectral line or the presence of any particular elements (e.g., vitamins, nutrients, etc.) but has simply been determined as the wavenumber with its intensity value that allows for the purest partitioning.
[0106] Terminal nodes N4 to N7 correspond to respective qualities (or / and scores) of molasses and may be related to the possible performance of a yeast.
[0107] For each decision tree, a prediction error (OOB error) can be calculated to determine the percentage of individuals misclassified by the decision tree. The lower this rate, the better the decision tree can be considered to be in its prediction quality.
[0108] As described previously, most machine learning methods can identify the important factors that most influence predictions or classifications. In the context of random forests, the algorithm can determine which specific intensities or wavenumbers are important for discriminating between molasses. Several evaluation methods have been described in the literature. For example, this importance can be determined by the number of times an intensity value associated with a wavenumber is used to partition the data at a node in the different decision trees of the random forest.
[0109] The determination by the statistical model of certain recurrences of intensity values associated with a given wavenumber can, for example, possibly provide information on the specific presence of a molecule or group of molecules such as vitamins, nutrients or sugars in a certain quantity (e.g. reflected by the amplitude / intensity associated with this wavenumber in RAMAN spectroscopy).
[0110] For example, a test based on an intensity value of a wavenumber that can characterize a line representative of a specific sugar can be implemented by a node of one or more decision trees (of the random forest), another test based on an intensity value of a wavenumber that can characterize a line representative of a specific sugar can be implemented by a node of one or more decision trees, etc.
[0111] Advantageously, determining certain recurrences can reveal a trend in particular important elements that allow us to distinguish good from poor quality molasses.
[0112] There figure 4b illustrates an example of determining molasses quality using a statistical model for a reference optical spectrum in input data.
[0113] One or more reference optical spectra from the test dataset [ens_test] previously presented can be provided as input to the statistical model based on a trained random forest as described previously in order to test the relevance of its predictions at model output.
[0114] For each reference optical spectrum provided as input to the model, the trees Tr_1, Tr_2, and Tr_n can then respectively determine (shaded nodes) a class, that is, a score and / or a quality of the molasses from the input reference spectrum. The determination of the molasses quality can then be carried out based on a majority vote or according to the mean in order to determine the most probable molasses quality and / or score. Since the molasses quality of the reference spectrum (e.g., Raman, IR, or THz spectrum) provided as input is known, as it comes from the test dataset, the accuracy of the statistical model based on the trained random forest can be determined by comparing the predicted quality obtained from the trained model with the actual molasses quality.
[0115] The statistical model (Model_predict) with acceptable prediction accuracy can then be used to determine a quality and / or score of an unknown molasses (e.g., Spect_X1 and Spect_X2 of the figure 3 ) whose optical spectrum is provided as input to the model, and to deduce the performance(s) of a yeast produced using unknown molasses as a nutrient source.
[0116] There figure 5 presents, in one or more embodiments, different spectra of cane molasses of known quality obtained by RAMAN spectroscopy.
[0117] A Raman spectrum can be plotted based on the Raman wavenumber shift and its intensity in arbitrary units. The Raman shift corresponds to the difference between the light from the monochromatic laser used and the scattered light. The Raman intensity is the number of photons detected from Raman-Stokes scattering. Indeed, in Raman spectroscopy, there are three types of scattering: Rayleigh scattering (elastic), Raman-Stokes scattering (inelastic), and anti-Stokes Raman scattering (inelastic). In inelastic scattering, the photon emitted by a laser and interacting with matter is not re-emitted at the same wavelength, but with a shift, either Stokes or anti-Stokes. Raman-Stokes scattering is the most commonly used to obtain the optical spectrum of a sample.
[0118] Thus, four optical spectra (Stokes) 503 of cane molasses characterizing good quality molasses (GD_QLT) and four optical spectra (Stokes) 505 of cane molasses characterizing poor quality molasses (BD_QLT) are presented. Each spectrum can be representative of a molasses quality that allows for the production of yeast with at least one respective performance characteristic (e.g., a certain bread-rising capacity and / or a certain yeast multiplication capacity). These optical spectra can be, for example, reference spectra used to train the random forest machine learning described earlier.
[0119] With reference to the figure 5 Several zones of difference between good and poor quality molasses (d1 to d7) can be observed. However, apart from zone d4, which shows a marked difference between two respective molasses qualities, most of the other identified zones do not allow for a clear determination of the important criteria that can distinguish good quality molasses from poor quality molasses—that is, molasses quality that allows for the production of yeast with one or more acceptable performance levels. Indeed, starting from the difference zones d2 and / or d3 (or even d6), some molasses that are of different quality can, nevertheless, exhibit identical or very similar spectra in certain zones 510; 513.
[0120] Furthermore, while the d4 zone appears to show marked differences when the molasses is of good or poor quality, the differences at the level of this zone may ultimately be of little or no relevance to be representative of a certain quality of molasses (good or bad), and therefore representative of the performance of a yeast fed in production with these molasses of a certain quality.
[0121] Thus, while a visual analysis of optical spectra makes it very difficult to find a correlation between a configuration of one or more spectral characteristics and a molasses quality that allows for yeast with one or more acceptable performances, the advantageous use of the process as described in this disclosure, i.e., machine learning coupled with appropriately chosen input data (e.g., reference spectra, hyperparameters, etc.), makes it possible to obtain a statistical model (e.g., based on a trained random forest) of molasses qualities that is effective in predicting molasses quality relative to yeast performance.
[0122] As mentioned previously, obtaining a trained statistical model can then allow us, before any yeast production, to determine whether or not the molasses should be discarded due to poor quality (e.g., a score below 3 and / or poor class and / or a red light), meaning it cannot produce yeast with one or more acceptable performance levels (e.g., according to specifications). Furthermore, based on the prediction of the molasses' quality or quality score, it may be possible to advantageously determine the nutrients and / or nutrient proportions to use in order to compensate for the poor quality of the molasses.
[0123] There figure 6 presents, in one or more embodiments, a device for implementing the process of this disclosure according to one or more embodiments.
[0124] In this embodiment, the device 600 may include a computer 601, this computer including a memory 602 for storing program instructions that can be loaded into a circuit, and capable of causing the circuit 603 to execute the process of this disclosure when the program instructions are handled by the circuit 603.
[0125] Memory 602 can also store data and information useful for carrying out the process of this description as described above.
[0126] Circuit 603 could be, for example: a processor or processing unit capable of interpreting instructions in a computer language, the processor or processing unit may include, be associated with or be attached to a memory comprising the instructions, or the association of a processor / processing unit and a memory, the processor or processing unit adapted to interpret instructions in a computer language, the memory comprising said instructions, or an electronic board in which the process sequence is described in silicon, or a programmable electronic chip such as an FPGA (for "Field-Programmable Gate Array") chip.
[0127] This computer may include an input interface 605 for receiving input data used for machine learning and an output interface for providing a useful dataset. For example, the input interface may receive input data 604 such as reference optical spectra associated with a respective molasses grade for machine learning, or spectra whose quality relative to at least one yeast performance is to be determined. The input interface may also receive a set of configuration parameters 605 for configuring the machine learning, for example, hyperparameters. In addition, optionally, the input interface may be connected 611 to a device for optical measurement (e.g., Raman spectroscopy), or the device 600 may be directly integrated into an optical measurement device.
[0128] For example, the useful dataset provided by the output interface 607 could be a trained prediction model 609, for instance, a statistical model of molasses quality relative to at least one yeast performance obtained from that molasses. Furthermore, the output interface could provide a prediction (or determination) 613 of a molasses quality (e.g., class and / or score) provided 604 to the input interface 605, the molasses quality being relative to at least one yeast performance obtained from that molasses. Finally, based on the trained prediction model, the output interface could also provide the nutrient(s) to be used and in what proportions to correct a molasses quality determined by the trained prediction model.
[0129] To facilitate interaction with the computer 601, a screen 611 and a keyboard 612 may be provided and connected to the computer circuit 603.
[0130] Expressions such as "understand", "include", "incorporate", "contain", "be" and "have" should be interpreted non-exclusively when interpreting the description and its associated claims.
[0131] The process is not limited to the examples of embodiments described above, which are only by way of example, but encompasses all the variants that a person skilled in the art may consider within the framework of the following claims.
[0132] Although described through a number of detailed embodiments, the proposed method and the device for implementing an embodiment of the method include various variants, modifications, and improvements that will be obvious to those skilled in the art, it being understood that these various variants, modifications, and improvements form part of the scope of this disclosure, as defined by the following claims. Furthermore, some of the systems and equipment described above may not incorporate all the modules and functions described for the preferred embodiments.
Claims
1. A method for qualifying molasses for yeast production based on an optical measurement, the method comprising: - based on reference optical spectra of samples of distinct molasses, associated with respective known molasses qualities, building by machine learning a statistical model of molasses qualities as a function of at least one spectral characteristic of the reference spectra, and - for a current optical spectrum of a sample of a current molasses, based on the statistical model, identifying said at least one spectral characteristic of the current optical spectrum and determining a quality of the current molasses, said current molasses quality relating to at least one yeast performance obtained when said yeast is fed with said current molasses, wherein each molasses quality associated respectively with a distinct molasses sample relates to at least one yeast performance obtained when said yeast is fed with said distinct molasses sample.
2. The method according to claim 1, wherein said at least one performance is based on at least one criterion selected from among: - a multiplication capacity of the yeast fed with said current molasses, - a bread leavening capacity, - a capacity to be preserved, - a capacity to be dried, - a capacity to fix structural elements, - a capacity to assimilate nutrients, - a capacity to adapt to stresses.
3. The method according to any one of the preceding claims, wherein the quality of the current molasses determined by said statistical model is defined by at least one molasses quality score, said at least one molasses quality score relating to said at least one performance.
4. The method according to any one of the preceding claims, comprising: - illuminating a molasses sample with a first emitted light signal interacting with the sample, and collecting a second light signal, resulting from the interaction between the first light signal and the current sample; and - measuring the second light signal, in selected frequency bands, to establish an optical spectrum representative of a chemical signature of the current molasses sample.
5. The method according to one of the preceding claims, wherein the optical spectrum comprises a set of wavenumbers, one or more wavenumber(s) of the set of wavenumbers being representative of a chemical compound of the molasses.
6. The method according to claim 5, wherein the molasses includes compounds at least one from among a sugar, a mineral, a vitamin, and the optical spectrum includes a plurality of wavenumbers, and wherein at least one wavenumber characterises the presence of one of said compounds.
7. The method according to one of the preceding claims, wherein said at least one spectral characteristic is determined according to at least one criterion selected from among a presence or an absence of a predefined wavenumber, a position of a predefined number, an intensity value of a predefined wavenumber, overpassing of a threshold of an amplitude of a predefined wavenumber, a maximum or minimum intensity value of a predefined wavenumber, an intensity deviation between two predefined wavenumbers, an intensity of a predefined wavenumber comprised between two predetermined intensity values, or simple or complex combinations of the aforementioned spectral characteristics.
8. The method according to one of the preceding claims, wherein the optical measurement is carried out by at least one technique from among a RAMAN spectroscopy, an IR spectroscopy, or a THZ spectroscopy.
9. The method according to one of the preceding claims, wherein the range of values of the reference optical spectra is comprised between 1 cm-1 and 2,500 cm-1.
10. The method according to claim 8, wherein the range of values of the reference optical spectra is comprised between 436 cm-1 and 1,700 cm-1.
11. The method according to one of the preceding claims, wherein the molasses is a cane molasses or a beet molasses or a mixture of both, or a mixture of sugar products such as a molasses and a sugar syrup, or a mixture of products one of the components of which is the molasses.
12. A computer program comprising program code instructions for executing the method according to any one of claims 1 to 11 when said program is executed on a computer.
13. A non-transitory computer-readable recording medium on which a program is recorded for implementing the method according to one of claims 1 to 11 when this program is executed by a processor.
14. A method for producing yeasts by fermentation comprising: - determining a respective molasses quality for each molasses of a plurality of molasses supplied at the input of the fermentation reaction according to the qualification method of any one of claims 1-11, and - feeding the fermentation reaction with each molasse of the plurality of molasses whose respective molasses quality is higher than a predetermined quality, and for all or part of the molasses having a molasses quality lower than the predetermined quality, feeding the fermentation reaction after modification of the molasses, or / and discarding the molasses.
15. The method according to claim 14 wherein said modification of the molasses comprises at least one modification selected from among: - mixing the molasses having a molasses quality lower than the predetermined quality, with another molasses having a distinct molasses quality, and so that the molasses mixture has a molasses quality higher than the predetermined quality, and - adding one or more additive(s) so that the molasses improved by the additives has a molasses quality higher than the predetermined quality.
