Method for determining the quality of molasses used in the production of yeast

Abstract

The present disclosure relates to a method for qualifying molasses from an optical measurement, the method may comprise: - from reference optical spectra of samples of distinct molasses, associated with respective known molasses qualities, constructing 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, from the statistical model, identifying the at least one spectral characteristic of the current optical spectrum and determining a quality of the current molasses, the quality of current molasses being relative to at least one yeast performance obtained when the yeast is fed with said current molasses. Abstract figure: Figure 2

Description

Title of the invention: Method for determining the quality of a molasses used in the production of a yeast Technical field

[0001] The present disclosure relates to the field of yeast production, and more particularly, it relates to a method for determining a quality of a molasses used as a source of nutrients in the production of a yeast, this quality of molasses 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 causing fermentation of organic matter. Yeasts are commonly used, whether in private homes or in the food or pharmaceutical industries. For example, they can be used in the production of wine, beer, or even leavened dough (e.g., bread). Among the yeasts, we can cite, but are not limited to, the genera Saccharomyces, Candida, Pichia, Kluyveromyces. The expression "yeast strain" designates 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 obtained from a single yeast cell.

[0003] Yeast is generally produced from the multiplication of a strain cultivated in a medium comprising a set of raw materials. In an industrial environment, from a given starting quantity of yeast, it is possible to produce several tons of it at the output. The raw materials generally used to feed yeast production are a set of nutrients and a source of sugar(s), which can be a sugar syrup, molasses, a mixture of sugar syrups and / or molasses.

[0004] Molasses is a raw material from the sugar industry and can be divided into two types of molasses, beet molasses or cane molasses. Nutrients can be sources of nitrogen, phosphorus, minerals, trace elements (eg, Fe, Zn, etc.), vitamins (eg, B, C, etc.), water, or even air.

[0005] When producing yeast, several factors can impact the yield or quality of the yeast produced. These factors can be the quality of the molasses used as well as the different proportions of nutrients provided to the yeast. Depending on a given quality of molasses or other factors or a set of factors including the quality of the molasses, it is possible to adjust the production schemes. production such as the respective quantities of each nutrient in order to optimize the production yield and / or the quality of the yeast output.

[0006] Currently, the quality of a molasses that has not yet been used in production as well as the quality of the yeast produced from this molasses can be determined only after either an initial production of yeast followed by qualification protocols for the yeast produced, or a series of miniaturized laboratory tests allowing the quality to be roughly estimated and atypical qualities of molasses to be ruled out. Indeed, the quality of an unknown molasses or the compensation of a known but poor quality molasses by adjusting the production schemes can only be done after qualification of the yeast obtained from this quality of molasses.

[0007] Knowing that a first production can correspond to several hundreds of tons of yeast produced and may have required the operation of several production sites, obtaining a poor production yield or poor quality yeast can represent a significant loss for a manufacturer.

[0008] At the same time, the organization of molasses production in "sugar campaigns" means that the quality of the molasses received over time is subject to significant variations, and that it is not possible to anticipate the impact of these variations.

[0009] These variations in quality and their impact on the quality of the yeast are all the more difficult to anticipate since the molasses are stored in tanks that can contain quantities that can be used for several weeks of production, which makes it even more difficult to detect a batch of molasses with an atypical quality that has been transferred to the storage tank.

[0010] There is therefore a need to determine the quality of a molasses before its use in yeast production so as to avoid the production of yeast having a quality and / or performance that does not comply (e.g. with a specification).

[0011] The present disclosure improves the situation.

[0012] A method for qualifying molasses from an optical measurement is thus proposed, the method being able to comprise:

[0013] - from reference optical spectra of distinct molasses samples, associated with respective known molasses qualities, construct by machine learning a statistical model of molasses qualities based on at least one spectral characteristic 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 characteristic of the current optical spectrum and determine a quality of the current molasses, said quality of common molasses being related to at least one yeast performance obtained when said yeast is fed with said common molasses.

[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, a molasses not having the quality required to enable the production of a yeast having one or more target performances can be discarded before any use in production. According to another advantage, from the determined molasses quality, it is possible to adapt the production schemes, for example by adjusting the proportions of nutrients necessary in order to correct a poor quality molasses so as to enable the target performance(s). Furthermore, depending on the degree of poor quality of a molasses, it may be possible to correct this quality by mixing the molasses with molasses of different qualities.

[0016] The features set forth in the following paragraphs may, optionally, be implemented. They may be implemented independently of one another or in combination with one another.

[0017] According to one or more embodiments, said at least one performance is based on at least one criterion chosen from: - a capacity for multiplying yeast fed by said common molasses, - a capacity for raising bread, - a capacity for preservation, - a capacity for being dried, - an ability to fix structural elements, - an ability to assimilate nutrients, - an ability to adapt to stress.

[0018] By multiplication capacity of the yeast fed by said common molasses, it can be understood a production yield per unit of mass of common molasses used or a ratio of quantity of yeast produced per quantity of sugar used in the production process.

[0019] By structural elements, we can understand elements such as nitrogen, phosphorus, carbon, hydrogen, oxygen or even sulfur.

[0020] In one or more embodiments, the quality of 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.

[0021] In one or more embodiments, rather than using a score or in combination with a score, the quality of the common molasses determined by said statistical model may be defined by a class such as good, bad, average, very good, green light to use this common molasses, red light to reject, or modify this common molasses, etc., the class being relative to at least one performance of the yeast obtained when fed with this common molasses.

[0022] In one or more embodiments, the method may further comprise: - 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.

[0023] In one or more embodiments, the established optical spectrum and the reference optical spectra may be measured under similar illumination and light signal measurement conditions.

[0024] Thus, for example, in one embodiment, the statistical model may be determined by implementing machine learning, for example from a random forest to arrive at a set of decision trees defining the statistical model.

[0025] In one or more embodiments, the optical spectrum may comprise a set of wavenumbers, wherein one or more wavenumbers of the set of wavenumbers are representative of a chemical compound of molasses.

[0026] According to one or more embodiments, the molasses may comprise compounds from among at least one sugar, one mineral, one vitamin, and the optical spectrum comprises a plurality of wave numbers, and in which at least one wave number characterizes the presence of one of said compounds.

[0027] According to one or more embodiments, said at least one spectral characteristic can be determined according to at least one criterion chosen from a presence or an absence of a predefined wave number, a position of a predefined number, an intensity value of a predefined wave number, an exceeding of a threshold of an amplitude of a predefined wave number, a maximum or minimum intensity value of a predefined wave number, an intensity difference between two predefined wave numbers, an intensity of a predefined wave number between two predefined intensity values, or even simple or complex combinations of the aforementioned spectral characteristics.

[0028] In one or more embodiments, said at least one spectral characteristic can be determined according to at least one criterion chosen from a presence or an absence of a predefined line, a position of a predefined line, a predefined line amplitude value, an exceeding of a threshold of a predefined line amplitude, a maximum or minimum value of a predefined line amplitude, an amplitude difference between two predefined lines, an amplitude of a line between two predetermined amplitude values, or even simple or complex combinations of the aforementioned spectral characteristics.

[0029] According to one or more embodiments, the optical measurement can be carried out by at least one technique among RAMAN spectroscopy, IR spectroscopy, or THZ spectroscopy.

[0030] According to one or more embodiments, the range of values ​​of the reference optical spectra may be between 1 cm1 and 2500 cm'1.

[0031] According to one or more embodiments, the range of values ​​of the reference optical spectra may be between 436 cm1 and 1700 cm'1.

[0032] According to one or more embodiments, the molasses is a cane molasses or a beet molasses or a mixture of the two, or a mixture of sugar products such as a molasses and a sugar syrup, or a mixture of products of which one of the components is molasses.

[0033] The present disclosure also relates to a computer program comprising program code instructions for executing the method of the present disclosure when the program is executable on a computer.

[0034] The present disclosure also relates to a non-transitory recording medium readable by a computer on which is recorded a program for implementing the method of the present disclosure when this program is executed by a processor.

[0035] The present disclosure also relates to a method for producing yeasts by fermentation which may comprise: - determining a respective molasses quality for each molasses of a plurality of molasses supplied as input to the fermentation reaction according to the molasses qualification method described in the present disclosure, and - feeding the fermentation reaction with each molasses 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, and / or rejecting the molasses.

[0036] By modifying the molasses, in one or more embodiments, at least one modification chosen from: - mixing 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, - adding one or more adjuvants so that the molasses improved by the adjuvants has a molasses quality higher than the predetermined quality.

[0037] 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 (eg temperature and pressure of the fermenter, type of fermenter, etc.) and for the same yeast strain. Brief description of the drawings

[0038] Other characteristics, details and advantages will appear on reading the detailed description below, and on analyzing the attached drawings, in which: Fig.l

[0039] [Fig.l] schematically describes, in one or more embodiments, an example of an optical spectrum of a molasses. Fig. 2

[0040] [Fig.2] illustrates a method for determining a quality of a molasses from a optical measurement in one or more embodiments. Fig. 3

[0041] [Fig.3] illustrates an example of application of the statistical model obtained according to the method of the present disclosure on spectra of unknown molasses. Fig.#4a

[0042] [Fig.4a] illustrates the principle of random forest for a statistical model in one or several embodiments. Fig. 4b

[0043] [Fig.4b] illustrates the principle of random forest for a statistical model in one or several embodiments. Fig. 5

[0044] [Fig.5] shows different spectra of cane molasses of known quality obtained by RAMAN spectroscopy. Fig. 6

[0045] [Fig.6] shows a device for implementing the method of the present disclosure according to one or more embodiments. Description of the embodiments

[0046] The terms "peak" and "line" may be interpreted in the same way and are interchangeable in the present disclosure. Similarly, the terms "intensity" and " amplitude” may be interpreted in the same manner and are interchangeable in this disclosure.

[0047] [Fig.l] schematically describes, in one or more embodiments, an example of an optical spectrum of a molasses.

[0048] There are many optical measurement techniques for obtaining an optical spectrum, such as infrared, microwave, terahertz spectroscopy, or RAMAN spectroscopy. All these techniques are based on the same principle of sending an electromagnetic wave (i.e. photons) onto the material to be probed by varying the frequency of the wave. A wave resulting from the wave-matter interaction can then be collected, either in transmission or in reflection, for each electromagnetic frequency. From the resulting wave, it is then possible to determine an optical spectrum representative of the constituents of an organic material (e.g. molasses), i.e. a chemical signature of the molasses.

[0049] An optical spectrum 100 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) in the spectrum. The abscissa 110 of an optical spectrum can be expressed either in frequency (Hz), or in wavelength or in wavenumber (cm-1). The ordinate 120 can be expressed in an arbitrary unit which can be a function of the signal processing applied to the measurement data. Indeed, in order to be correctly exploitable, it is common for one or more data processing operations to be applied to the raw data (i.e. data from optical measurements) in order to allow comparison between the spectra for example.

[0050] For example, in the case of RAMAN spectroscopy, a baseline, a derivation of this baseline or even a combination with the normalization of the data can be carried out in order to eliminate certain peaks presenting aberrations or whose presence results from the Laser used for the spectroscopy,

[0051] Each optical measurement technique may correspond to a specific range of frequencies (or wavelengths or wavelength numbers) of use in which the frequency of the incident wave varies. For example, the wavelength range (expressed in microns) may be between [Ip m-20pm], [0.3p m-0.7pm], or [1 cm 1 - 2500cm *]. In one or more embodiments, this frequency range may be between [436 cm1 - 1700 cm4].

[0052] Depending on the optical measurement technique and the wavelength range used to obtain a spectrum of an organic material, the position of a peak may correspond, for example, either to an atomic or chemical bond, or to 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 an organic matter by reflecting, for example, the chemical bonds or functional groups of this molecule

[0053] 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.

[0054] The optical measurement techniques for performing spectroscopy of 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 this medium.

[0055] Depending on the measurement techniques used, the exploitation of spectra can be more or less complicated, whether visually or according to conventional methods of exploitation of spectrum data. 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 set 105;107 making their exploitation difficult. In addition, depending on the resolution used (e.g. the step of the optical measurement), a peak, depending on its width at half-height, can spread more or less significantly over the optical spectrum, and thus can complicate the interpretation of the data a little more.

[0056] Furthermore, while a change in the optical measurement technique used may seem obvious when the optical spectrum obtained is not usable (e.g. 107 line clusters), the use of another optical measurement technique may lead to obtaining an optical spectrum that is completely or partially different from the spectrum initially obtained. Indeed, the compounds / molecules sought may not be sensitive to the wavelengths used by an alternative optical technique.

[0057] Thus, it may be crucial to be able to exploit an optical spectrum, independently of the configuration of the peaks of the spectrum, and more particularly when interpretations and comparisons between a large number of optical spectra obtained from samples relatively close in their chemical / molecular fingerprint are sought.

[0058] [Fig.2] illustrates a method for determining a quality of a molasses from an optical measurement in one or more embodiments.

[0059] From optical spectra (Spect_refl.. .Spect_refn, n being a positive integer) 201 of reference of distinct molasses samples which are known and which are respectively associated with known molasses qualities (Ql, 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 constructed by machine learning.

[0060] Machine learning can consist of determining, in the reference spectra of molasses, spectral configurations leading to a certain quality of molasses, in particular certain qualities of molasses suitable for obtaining one or more desired performances for a yeast for example.

[0061] 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 quality of molasses, this quality of molasses being relative to one or more performances of a yeast produced using this quality of molasses.

[0062] The reference spectra and the respective qualities of the molasses associated with the reference spectra may correspond to input data for feeding the machine learning, and each reference spectrum may be representative of a structure (e.g. chemical) of a molasses (e.g., cane or beet).

[0063] Furthermore, in one or more embodiments, the quality of molasses can be defined by at least one respective score Scn and / or according to a class of molasses (e.g. good, bad, average, red light, green light etc.) and / or a correlation between its spectral profile and a statistical group of industrial performances.

[0064] 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(s) of the resulting yeast.

[0065] For example, the score of each molasses associated with a reference spectrum may have a value between 0 and 1 (eg value 0 = bad and value 1 = good molasses) or 1 and 10. The value 1 may be representative of a poor quality molasses and the value 10 representative of a very good quality molasses. Thus, a score (or a score value) greater than 8 may correspond to a molasses making it possible to obtain a yeast with very good performance, for example a very good capacity for raising bakery dough such as bread and / or a very good multiplication capacity (ie a good production yield of quantity of yeast).

[0066] In one or more embodiments, a molasses quality may 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 average 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.

[0067] According to another example, in one or more embodiments, the molasses quality may be defined by a plurality of scores, each score relating to a respective performance. For example, a reference optical spectrum may be associated to a first score relating to a bread rising capacity, and to a second score relating to a multiplication capacity (i.e. yeast production yield).

[0068] The scores can be combined with each other in order to obtain information that can be used by personnel in charge of yeast production within an industry or a testing laboratory.

[0069] The score of each molasses associated with a reference spectrum may have been determined beforehand on the basis of one or more qualification protocols.

[0070] As an example, a protocol for qualifying a molasses may be the implementation of this molasses with a reference yeast in order to determine the performance of the reference yeast, in terms of cell multiplication and enzymatic activity.

[0071] For example, performance based on (or relative to) bread rising ability may be assessed by the ability of the prepared yeast to produce carbon dioxide, and / or may be assessed by the fermentative activity of the yeast.

[0072] The fermentative activity is determined by measuring the release of carbon dioxide gas (expressed absolutely in volume or relatively in percentage) from a given dough 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).

[0073] In one or more embodiments, a molasses quality Scn score greater than 5 (i.e. corresponding to a good quality molasses for example), may relate to a bread rising capacity (i.e. a yeast performance) in comparison with that generally observable on the yeasts produced in a reference process. Similarly, a molasses quality Scn score less than 3 (i.e. corresponding to a poor quality molasses), may relate to a bread rising capacity making it possible to achieve specific volume values ​​lower than the values ​​generally observable on the yeasts produced in a reference process.

[0074] The multiplication capacity of yeast fed with common molasses, also called production yield, can be expressed as the quantity of yeast produced per quantity of sugar used, or as the quantity of sugar used per quantity of yeast produced.

[0075] In one or more embodiments, a molasses quality Scn score greater than 5 (i.e. corresponding to a good quality molasses for example), may relate to a yeast multiplication capacity (i.e. a yeast performance) generally observable on the yeasts produced in a process of reference. Similarly, a molasses quality Scn score of less than 3 (i.e. corresponding to poor quality molasses), may be related to a yeast multiplication capacity allowing values ​​to be reached which are lower than the values ​​generally observable on yeasts produced in a reference process.

[0076] In one or more embodiments, the input data of the reference spectra for machine learning can be in the form of a set of wavelengths respectively associated with an intensity value (or amplitude). For example, the set of wavelengths can range from 0 to 2000 cm-1 and with a wavelength step of 0.1 cm-1.

[0077] There are many machine learning techniques. All machine learning methods can be used to learn the rules linking the characteristics of molasses and the performance of yeasts. In one or more embodiments, the machine learning may in particular be based on the use of random forests, on the use of one or more linear regression algorithms, on the use of K nearest neighbors, on the use of deep learning, or even the use of neural networks.

[0078] In addition, machine learning may include, as input data, one or more hyperparameters. A hyperparameter is a parameter used to parameterize the algorithm (e.g. random forest, deep learning, etc.) before any learning. Hyperparameters may be specific to a type of machine learning and may, therefore, vary depending on the type of machine learning used.

[0079] For example, a hyperparameter can be the number of decision trees making up a random forest, the number of nodes in a tree, the size of the final nodes, or even the number of observations chosen by “bootstrap” to construct each tree.

[0080] After determining the model, it can then be used to identify 205, for a current spectrum of a current molasses sample in input data 207, at least one spectral characteristic of the current spectrum, and determine a quality 209 of the current molasses.

[0081] From the determined quality of the common molasses, one or more performances of a yeast produced using the common molasses as a nutrient source can be determined. The quality of the common molasses can be represented according to a class (e.g. bad, average, good, very good, green light when it can be used or red light when it cannot be used or it needs to be modified, etc.) and / or according to one or more scores relating to yeast performance, and / or a correlation between its spectral profile and a statistical group of industrial performance.

[0082] Of course, the yeast performance obtained when the yeast is fed with common molasses depends on several contributions such as the yeast strain used to produce the yeast, production conditions (e.g. pressure, temperature in the fermenter, configuration of the fermenter, etc.), and inputs including the molasses used to feed the yeast production. Within the scope of the method of the present disclosure, it is understood that the yeast performance(s) determined as a function of molasses qualities are determined for an identical yeast strain, identical production conditions, and identical inputs (with the exception of molasses).

[0083] 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 according to one or more optical measurement techniques. For example, each reference spectrum may be representative of a combination of optical spectra of the same molasses and obtained by a respective optical measurement technique.

[0084] In one or more embodiments, the optical measurement technique on the molasses for obtaining the reference spectra is the same as that used on the common 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.

[0085] In one or more embodiments, data obtained after determining a molasses quality relative to a yeast performance according to the input spectrum 207 may be used to feed 210 the machine learning. Thus, advantageously, the statistical model may be continuously improved.

[0086] Of course, the statistical model may not be constructed every time an unknown molasses is required to be qualified. The statistical model may be determined once and then used with unknown molasses spectra.

[0087] The random forest methodology is described in more detail in the following publication: - Léo Breiman, Random Forests, Machine Leaming, vol. 45, no. 1, 2001, p. 5-32

[0088] As described above, the present disclosure is not limited to the use of random forest. Other types of machine learning may also be suitable, such as, for example, the machine learning methods described in the following publications: -The Elements of Statistical Leaming, 2nd edition, Trevor Hastie, Robert Tibshirani, Jerome Friedman, ISBN-13: 978-0387848570

[0089] - Deep Leaming, Lan Goodfellow, Yoshua Bengio and Aaron Courville, ISBN-13: 978-0262035613 .

[0090] [Fig.3] illustrates an example of application of the statistical model obtained according to the method of the present disclosure on spectra of unknown molasses.

[0091] The respective quality (Q1; Q2) of two molasses M1 and M2 which can potentially be used as a source of nutrients in yeast production (eg yeast production by fermentation) can be sought.

[0092] This quality can be sought in order to know if the molasses M1 and M2 can be used to feed a yeast production while making it possible to achieve one or more target performances of the yeast produced, or / and in order to determine the molasses most suitable for producing a yeast having one or more target performances, or / and according to a compromise of one or more target performances.

[0093] With reference to [Fig.3], by way of example, a discretization between the two molasses M1 and M2 can be sought in order to determine the molasses which can lead to obtaining a yeast production yield higher than a value generally observed with a reference yeast operated in a reference process and / or obtaining a yeast (produced) which makes it possible to achieve specific volume values ​​higher than a value generally observed with a reference yeast operated in a reference process. According to another example, the discretization between the two molasses M1 and M2 can be sought in order to determine the molasses which has the desired performance ratio, eg a production yield / a bread rising capacity.

[0094] Each molasses Ml; M2 can be associated respectively with an optical spectrum (Spect_Xl and Spect_X2), obtained for example by RAMAN spectroscopy.

[0095] Thus, the application of the statistical model (Model_statist) of molasses qualities on the optical spectra of molasses (Spect_Xl and Spect_X2), different from the reference spectra used for machine learning, can make it possible to identify particular spectral characteristics. For example, the statistical model can determine a quality Q1 (and / or a score and / or a class) of the spectrum Spect_Xl on the basis of the identification of the spectral characteristics corresponding to the presence of the wave numbers RI, R2 and R3 coupled with a respective amplitude threshold THRi, THr2jTH R3 for each of the wave numbers. Similarly, the prediction model can determine a quality Q2 of the spectrum Spect_X2 on the basis of the identification of the spectral characteristics corresponding to the presence of the wave numbers RI, R2 and R4 coupled with an amplitude range A to be respected for the wave numbers RI and R2.

[0096] From the determined qualities (eg according to a score), molasses M1 may seem to be a better candidate for feeding yeast production than molasses M2. Indeed, molasses M1 has a score of 8, representative of a very good molasses, that is to say allowing, for example, obtaining a yeast production yield higher than a value generally observed with a reference yeast operated in a reference process, and / or or obtaining a good bread rising capacity defined by achievable specific volume values ​​higher than a value generally observed with a reference yeast operated in a reference process.

[0097] In one or more embodiments, for a plurality of molasses that can potentially be used to feed a yeast production (by fermentation), the molasses having a molasses quality higher than a predetermined quality (e.g., one or more scores) can be used to feed a yeast production. In one or more embodiments, among the molasses having a quality higher than a predetermined quality, priority for feeding the yeast production can be given to the molasses having the best quality. For example, the best quality can 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 averages of the respective quality scores of each molasses.

[0098] Figures 4a and 4b illustrate the random forest principle for a statistical model in one or more embodiments.

[0099] The principle of the random forest algorithm is to determine, according to a majority vote for categorical data or according to an average of individual results for quantitative data, the most probable prediction made by a multitude of decision trees constituting the random forest.

[0100] All decision trees in a random forest may be independent of each other, and the prediction determined by each decision tree before majority voting or averaging may be uncorrelated from the predictions determined by the other decision trees in the random forest.

[0101] A first step of this algorithm can consist of constructing decision trees (or decision trees). [Fig.4a] presents such a decision tree which can, for example, be divided into three parts: - a root node NI corresponding to the first node of the tree (i.e. input node), - internal nodes N2; N3 which may have descendants (i.e. other nodes), - terminal nodes N4; N5; N6; N7 (also called leaves).

[0102] Each intermediate "node", i.e. 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 partition rule for example.

[0103] Nodes NI to N3 can be defined by explanatory variables (e.g. quantitative or qualitative type).

[0104] In order to construct the different decision trees such as the one presented in [Fig.4a], a selection of subsets of individuals from the training data set with replacements can be carried out using so-called “bootstrap” sampling, which consists of randomly drawing individuals (with replacement) from the training data set.

[0105] By discount, it can be understood that the candidates selected in a subset remain in the training set, that is to say that they can be selectable for another subset. The individuals used to construct a decision tree are generally called “in bag” and the individuals used to evaluate the predictiveness of the tree are generally called “out of bag” (called “OOB”).

[0106] This training data set can itself be obtained following the implementation of a cross-validation strategy consisting of separating the initial data set (eg 70 / 30) into two sub-data sets, one for training (eg 70%) and the other for testing (eg 30%) the prediction model.

[0107] Each decision tree can then be constructed (or trained) on a respective subset of individuals. That is, there can be as many subsets of individuals as there are decision trees, and the construction of the tree is only carried out on a single subset of individuals.

[0108] In the case of the present disclosure, the initial data set may correspond to a set of reference optical spectra (eg in the form of sets of wave numbers) associated with a respective quality of molasses (or / and one or more respective molasses scores), and the respective variable used in the test of each node NI to N3 may be, for example, an intensity (or amplitude) of a specific wave number.

[0109] During training (or construction of the tree), at each construction of a new node, the variable allowing the partition at the node level (into two descendant nodes) can be determined by the machine learning algorithm from the random subset from the training data set. For example, machine learning can review (eg following the discretization step of the optical spectra) all (or a subset) of the wave numbers of the reference optical spectra of the random subset in order to determine a wave number and / or an intensity to be used as a variable. More precisely, the algorithm can seek to determine if there exists, for example, an intensity value allowing the spectra to be separated in the purest way between the spectra corresponding to good or poor quality molasses. For example, the test associated with node N2 might be to determine whether the amplitude (or intensity) associated with a particular wavenumber is above an intensity threshold or within an intensity range.

[0110] The choice of one variable over another may be based on measuring the purity of the node using, for example, the Gini index or Shannon entropy.

[0111] Thus, none of the input data provided for machine learning influences it in its choice of variables for partitioning at each node. Random forest machine learning can here determine its own variables (e.g. wave number, amplitudes, etc.) in order to build a statistical model, without a priori on the presence of one or more wave numbers that should be detected in the reference spectra. For example, node N3 can include a test based on an intensity value of a wave number not characterizing any specific line or the presence of particular elements (e.g. vitamin, nutrients, etc.) but having been determined simply as the wave number with its intensity value allowing a purest partition.

[0112] The terminal nodes N4 to N7 correspond to respective qualities (or / and scores) of molasses and which can be relative to the possible performances of a yeast.

[0113] 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 in its prediction.

[0114] 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 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.

[0115] The determination by the statistical model of certain recurrences of intensity values ​​associated with a given wave number can, for example, possibly provide information on the specific presence of a molecule or group of molecules such as vitamins, nutrients or sugars according to a certain quantity (eg reflected by the amplitude / intensity associated with this wave number in RAMAN spectroscopy).

[0116] For example, a test based on an intensity value of a wave number 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 wave number that can characterize a line representative of a specific sugar can be implemented by a node of one or more decision trees, etc.

[0117] Advantageously, the determination of certain recurrences can make it possible to identify a trend on the particular important elements making it possible to discretize a good from a bad quality of molasses.

[0118] [Fig.4b] illustrates an example of determining a molasses quality by the statistical model for a reference optical spectrum as input data.

[0119] One or more reference optical spectra from the test data set [ens_test] previously presented can be provided as input to the statistical model based on a random forest trained as described previously in order to test the relevance of its predictions at the model output.

[0120] For each reference optical spectrum provided as input to the model, the trees Tr_l; Tr_2; Tr_n can then respectively determine (grayed nodes) a class, i.e. a score and / or a quality of the molasses of the input reference spectrum. The determination 405 of the molasses quality can then be carried out on the basis of a majority vote 403 or according to the average in order to determine the most probable quality and / or molasses score. The molasses quality of the reference spectrum (e.g. RAMAN, or IR or THz spectrum) provided as input being known, since it comes from the test data set, the accuracy of the statistical model based on the trained random forest can be determined by comparing the prediction of a quality obtained from the trained model with the actual quality of the molasses.

[0121] The statistical model (Model_predict) presenting an acceptable prediction accuracy can then be used to determine a quality and / or a score of an unknown molasses (eg Spect_Xl and spect_X2 of [Fig.3]) whose optical spectrum is provided as input to the model, and to deduce the performance(s) of a yeast produced using the unknown molasses as a nutrient source.

[0122] [Fig.5] shows, in one or more embodiments, different spectra of cane molasses of known quality obtained by RAMAN spectroscopy.

[0123] A Raman spectrum can be plotted as a function of the Raman shift in wavenumber 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 a Stokes Raman scattering. Indeed, in Raman spectroscopy, there are three types of scattering: Rayleigh scattering (elastic), Stokes Raman 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, Stokes or anti-Stokes. Stokes Raman scattering is the most commonly used to obtain the optical spectrum of a sample.

[0124] Thus, four optical spectra (Stokes) 503 of cane molasses characterizing a good quality of molasses (GD_QLT) and four optical spectra (Stokes) 505 of cane molasses characterizing a poor quality of molasses (BD_QLT) are presented. Each spectrum can be representative of a quality of molasses which makes it possible to obtain a yeast with at least one respective performance (eg a certain bread rising capacity and / or a certain yeast multiplication capacity). These optical spectra can be, for example, reference spectra used to feed the random forest machine learning described above.

[0125] With reference to [Fig.5], several areas of difference between a good and a poor quality of molasses (d1 to d7) may be notable. However, apart from area d4 which presents a marked difference between two respective qualities of molasses, most of the other areas identified do not make it possible to clearly determine the important criteria which can discretize a good quality of molasses from a poor quality of molasses, i.e. a quality of molasses making it possible to obtain a yeast having one or more acceptable performances. Indeed, from the areas of differences d2 and / or d3 (or even d6), certain molasses which are nevertheless of different quality may, however, present identical or very close spectra in certain areas 510; 513.

[0126] Furthermore, while zone d4 appears to exhibit marked differences when the molasses are of good or poor quality, the differences in this zone may ultimately be of little or no relevance to being representative of a certain quality of molasses (good or poor), and therefore representative of the performance of a yeast fed in production with these molasses of certain quality.

[0127] Thus, while a visual analysis of the optical spectra makes it very complicated to find a correlation between a configuration of one or more spectral characteristics and a molasses quality making it possible to obtain a yeast having one or more acceptable performances, the advantageous use of the method as described in the present disclosure, i.e. automatic 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 its prediction of molasses quality relative to yeast performances.

[0128] As mentioned previously, obtaining a trained statistical model can then make it possible, before any yeast production, to determine whether or not to discard this molasses due to poor quality (e.g. score less than 3 and / or poor class and / or red light), i.e. not making it possible to obtain a yeast having one or more acceptable performances (e.g. according to specifications). In addition, from the prediction of the quality or the quality score of the molasses, it may be possible to advantageously determine the nutrients and / or the proportions of nutrients to be used so as to compensate for the poor quality of a molasses.

[0129] [Fig.6] shows, in one or more embodiments, a device for implementing the method of the present disclosure according to one or more embodiments.

[0130] In this embodiment, the device 600 may comprise a computer 601, this computer comprising a memory 602 for storing program instructions loadable into a circuit, and adapted to cause the circuit 603 to execute the method of the present disclosure when the program instructions are managed by the circuit 603.

[0131] The memory 602 may also store data and information useful for carrying out the method of the present description as described above.

[0132] Circuit 603 may be for example: - a processor or processing unit capable of interpreting instructions in a computer language, the processor or processing unit may comprise, be associated with or be attached to a memory comprising the instructions, or - the combination of a processor / processing unit and a memory, the processor or processing unit being adapted to interpret instructions in a computer language, the memory comprising said instructions, or - an electronic card in which the process sequence is described in silicon, or - a programmable electronic chip such as an FPGA chip (for “Field-Programmable Gate Array”).

[0133] This computer may include an input interface 605 for receiving input data used for machine learning and an output interface for providing a set of useful data. For example, the input interface may receive input data 604 such as reference optical spectra associated with a respective molasses quality for machine learning or spectra whose quality relative to at least one performance of a yeast is to be determined. The input interface may also receive a set of configuration parameters 605 for configuring machine learning, for example hyperparameters. Furthermore, optionally, the input interface may be connected 611 to a device enabling optical measurement (eg RAMAN spectroscopy), or the device 600 can be directly integrated into an optical measurement device.

[0134] By way of example, the set of useful data provided by the output interface 607 may be a trained prediction model 609, for example a statistical model of molasses quality relating to at least one yeast performance obtained from this molasses. In addition, the output interface may provide a prediction (or determination) 613 of a quality of a molasses (eg class and / or score) provided 604 to the input interface 605, the molasses quality being relative to at least one yeast performance obtained from this molasses. Finally, on the basis of the trained prediction model, the output interface may also provide the nutrient(s) to be used and in what proportions to correct a molasses quality determined by the trained prediction model.

[0135] To facilitate interaction with the computer 601, a screen 611 and a keyboard 612 may be provided and connected to the computer circuit 603.

[0136] Expressions such as "comprise", "include", "incorporate", "contain", "be" and "have" are to be interpreted in a non-exclusive manner when construing the description and its associated claims.

[0137] The method is not limited to the examples of embodiments described above, only by way of example, but it encompasses all the variants that a person skilled in the art may envisage within the framework of the claims below.

[0138] Although described through a number of detailed exemplary embodiments, the proposed method and apparatus for implementing an embodiment of the method include various variations, modifications, and improvements that will be apparent to those skilled in the art, it being understood that these various variations, modifications, and improvements are within the scope of the present disclosure, as defined by the following claims. In addition, different aspects and features described above may be implemented together, or separately, or substituted for each other, and all different combinations and subcombinations of the aspects and features are within the scope of the present disclosure. Furthermore, some systems and equipment described above may not incorporate all of the modules and functions described for the preferred embodiments.

Claims

Claims

1. A method for qualifying molasses from an optical measurement, the method being implemented by computer and comprising: - from reference optical spectra of distinct molasses samples, associated with respective known molasses qualities, each molasses quality respectively associated 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, constructing 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, from the statistical model, identifying said at least one spectral characteristic of the current optical spectrum and determining a quality of the current molasses,said quality of common molasses being relative to at least one yeast performance obtained when said yeast is fed with said common molasses, and in which process the range of values of the reference optical spectra is between 1 cm 1 and 2500 cm '.,

2. Method according to claim 1, in which said at least one performance is based on at least one criterion chosen from: - a capacity for multiplication of the yeast fed by said common molasses, - a capacity for bread rising, - 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 stress.

3. A method according to claim 1 or claim 2, 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 being relative to said at least one performance.

4. A method according to any preceding claim, comprising: - illuminating a sample of molasses 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 sample of molasses.

5. A method according to any preceding claim, wherein the optical spectrum comprises a set of wavenumbers, one or more wavenumbers of the set of wavenumbers being representative of a chemical compound of molasses.

6. The method of claim 5, wherein the molasses comprises compounds from at least one sugar, one mineral, one vitamin, and the optical spectrum comprises a plurality of wavenumbers, and wherein at least one wavenumber characterizes the presence of one of said compounds.

7. Method according to one of the preceding claims, in which said at least one spectral characteristic is determined according to at least one criterion chosen from a presence or an absence of a predefined wave number, a position of a predefined number, an intensity value of a predefined wave number, an exceeding of a threshold of an amplitude of a predefined wave number, a maximum or minimum intensity value of a predefined wave number, an intensity difference between two predefined wave numbers, an intensity of a predefined wave number between two predefined intensity values, or simple or complex combinations of the aforementioned spectral characteristics.

8. Method according to one of the preceding claims, in which the optical measurement is carried out by at least one technique among RAMAN spectroscopy, IR spectroscopy, or THZ spectroscopy.

9. A method according to any preceding claim, wherein the range of values of the reference optical spectra is between 436 cm 1 and 1700 cm '.

10. A method according to any preceding claim, wherein the molasses is cane molasses or beet molasses or a mixture of both, or a mixture of sugar products such as molasses and sugar syrup, or a mixture of products in which one of the components is molasses.

11. A computer program comprising program code instructions for executing the method of any one of claims 1 to 10 when said program is executed on a computer.

12. Non-transitory recording medium readable by a computer on which is recorded a program for implementing the method according to one of claims 1 to 10 when this program is executed by a processor.

13. A method of producing yeast by fermentation comprising: - determining a respective molasses quality for each molasses of a plurality of molasses supplied as input to the fermentation reaction according to the qualification method of any one of claims 1-10, and - feeding the fermentation reaction with each molasses 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, and / or rejecting the molasses.

14. A method according to claim 13 wherein said modification of the molasses comprises at least one modification selected from: - mixing molasses having a molasses quality lower than the predetermined quality, with another molasses having a distinct molasses quality, and such that the mixture of molasses has a molasses quality higher than the predetermined quality, - adding one or more adjuvants such that the molasses improved by the adjuvants has a molasses quality higher than the predetermined quality.

Images