BIOCHEMICAL PROCESS MONITORING SYSTEM

Abstract

The invention relates to a method and system for in-situ monitoring of a biochemical process in a reactor comprising a tank (5) for receiving a liquid (7), said system comprising: - a measuring device (9) intended to be introduced by floating into said tank (5), said measuring device (9) being instrumented with sensors configured to perform measurements relating to the biochemical process at successive times and to transmit observation data representative of said measurements at said successive times, and - a control device (11) configured to control the regulation of the biochemical reactor (3) at said successive times according to said observation data received from the measuring device (9). Fig. 1

Description

Title of the invention: SYSTEM FOR MONITORING A BIOCHEMICAL PROCESS Technical field

[0001] The present invention relates to the field of biochemical process monitoring. State of the prior art

[0002] The regulation of a biochemical process can address numerous industrial applications in various fields such as, for example, agri-food (wine, cider, beer, spirits, liquid foods, etc.), pharmaceuticals (medicines or complementary nutrients, etc.), environmental (water analysis in treatment plants, methanization, etc.), petrochemicals (biogas, etc.), electrochemicals, etc.

[0003] Biochemical processes involve the action of microorganisms on complex substrates (fruit juices, leachates, etc.). Their regulation in industrial environments such as in the food industry is generally carried out by manual off-line analyses. Regulation in bioreactors by monitoring on-line control parameters is reserved for high added value products such as medical products.

[0004] The control parameters generally include temperature measurements, and possibly other additional measurements which may be specific to the application, such as acidity, turbidity, gas release, density, dissolved oxygen, etc.

[0005] In the case of alcoholic beverages, it is necessary to measure at least the temperature and density of a liquid to control food quality such as the sugar and alcohol content. To produce medicines or to analyze water in treatment plants, the measurements more generally concern the pH, the Redox potential, the conductivity and the presence of dissolved oxygen and carbon dioxide, etc.

[0006] Regarding density measurement, there are manual densimeters that are commonly used in various fields of activity. Their dimensions and weight define the measurement range and accuracy. However, they require samples to perform offline analyses and the operations are completely manual.

[0007] There are electronic densimeters that measure the oscillation frequency of U-shaped tubes that vary according to the mass of the liquid sampled. These devices can be used online but are expensive and sensitive to certain disturbances (fouling, bubbles).

[0008] There are also devices that determine the density of a liquid by measuring the refractive index. However, the measurement can be disturbed by the presence of gas (for example, CO2) or fouling of the sensor.

[0009] All these devices allow in some cases online measurements but not in situ. However, there are autonomous modules in the shape of a ball intended to measure pH, conductivity, pressure and movement in a liquid. These modules can be freely circulated in bioreactors and thus allow in situ measurements. However, these modules are limited to very specific measurements and are not suitable for measuring density. In addition, the transmission of data is not easy from the liquid medium and, moreover, their location in the liquid is unknown, resulting in a lack of spatial reference for the measurements.

[0010] Furthermore, bioprocesses generally involve the regulation of temperature, agitation, oxygenation and nutrient supply. The purpose of determining these control parameters is to improve both the quality and yield of the product while optimizing energy and nutrient inputs. This control generally relies on the know-how of the user (oenologist in the case of wine, microbiologist for pharmaceutical products, etc.) and can therefore be very tedious and time-consuming.

[0011] The object of the present invention is to overcome the aforementioned drawbacks by proposing automatic and precise adjustment of the control parameters of the desired product using real-time and in situ monitoring of the evolution of the biochemical process while allowing precise multi-parameter measurements using a single measuring device. Presentation of the invention

[0012] The subject of the invention is a system for in-situ monitoring of a biochemical process in a reactor comprising a tank intended to receive a liquid, said system comprising: - a measuring device intended to be introduced into said tank, said measuring device being equipped with sensors configured to carry out measurements relating to the biochemical process at successive times and to transmit at said successive times observation data representative of at least the temperature and the density, and - a control device configured to control at successive times the regulation of the biochemical reactor as a function of said observation data received from the measuring device.

[0013] Thus, the measuring device makes it possible to monitor the evolution of the physical parameters and to transmit them to the control device so that the latter can control by feedback and in real time the reaction in the tank. This system thus makes it possible to automatically control the biochemical process while reacting directly in real time in the event of detection of an anomaly in the process.

[0014] Advantageously, said measurements comprise measurements of temperature and density of the liquid and at least one other type of measurements from among the following measurements: mechanical measurements of pressures and / or gas flow rate and / or accelerations and / or flotation level, electrical measurements of voltages and / or currents and / or resonance frequencies, and optical measurements.

[0015] This allows for multiple indicators for more efficient and precise feedback and control of the biochemical process.

[0016] According to one embodiment, the measuring device comprises a microprocessor configured to determine at said successive times vectors of physical variables relating to the biochemical process as a function of the corresponding measurements carried out at said successive times, the observation data transmitted by the measuring device to said control device comprising said vectors of physical variables and said corresponding measurements.

[0017] Thus, the measuring device processes the measurements to determine over time vectors of physical values ​​of interest including the density before transmitting them to the control device.

[0018] According to another embodiment, the observation data transmitted by the measuring device to said control device comprise said measurements. In this case, the control device is configured to determine at said successive times vectors of physical variables relating to the biochemical process as a function of the corresponding measurements.

[0019] According to this embodiment, the measuring device only transmits the measurements and it is the control device which processes them to determine the physical parameters of interest. This makes it possible to minimize the processing and the electrical consumption at the level of the measuring device.

[0020] Each of said vectors of physical variables comprises a temperature variable, a liquid density variable determined from the measurements of buoyancy levels and / or pressures, and at least one other variable from among the following variables: gas release determined from the pressure measurements, electrical conductivity and / or permittivity of the liquid determined from the electrical measurements, movement of the liquid determined from the acceleration measurements, PH and / or redox potential determined from the electrical measurements, dissolved oxygen and / or CO 2 determined from the electrical measurements and / or optical measurements, optical absorption spectrum and / or optical rotation determined from the optical measurements.

[0021] These physical indicators make it possible to detect with great precision the slightest anomaly in the biochemical process.

[0022] Advantageously, the control device is configured to control the regulation of the biochemical reactor by at least one action among the following actions: modification of the stirring speed, modification of the temperature, modification of the oxygen supply flow rate, supply of nutrients or other elements for activating or stabilizing the biochemical process, supply of yeasts or bacterial strains.

[0023] Advantageously, the control device is configured to: - predicting at said successive instants predictive vectors of physical variables as a function of said previous vectors of physical variables from the measurements, - predicting at said successive instants anticipatory measurements as a function of said predictive vectors of physical variables, - calculating at said successive instants measurement deviations between the anticipatory measurements and the corresponding actual measurements, -correct at said successive instants the vectors of physical variables according to said measurement deviations, -determine at said successive instants vectors of regulation actions as a function of said vectors of corresponding physical variables, and -control the regulation of the biochemical reactor by triggering at said successive times regulation actions based on said regulation action vectors.

[0024] This makes it possible to determine with precision and for each instant of time, a series of actions with knowledge of the physical variables observable thanks to the sensors of the measuring device.

[0025] Advantageously, the control device is configured to determine the regulation action vectors as a function of the predictive vectors of physical variables in the event of a fault in the measuring device.

[0026] This allows the process to be regulated further in the event of a temporary malfunction of the measuring device.

[0027] Advantageously, the control device is configured to: - predicting each predictive vector of physical variables using a first temporal function f defining values ​​of physical variables with knowledge of their values ​​at a previous instant of time, said first function f being predetermined by a standard Markov model of order of at least 1, - predicting each anticipatory measurement using a second function g associating the measurements made by the measuring device with the physical variables at a given instant of time, said second function g being predetermined by equations linking the measurements and the physical variables, and - determination of each vector of regulation actions using a third function h defining a correspondence between the actions to be carried out and the values ​​of the physical variables at a given instant of time, said third function h being predetermined by a prior selection of action trigger thresholds.

[0028] These functions can be pre-adapted to be used in a simple manner for the control of any biochemical reaction.

[0029] Advantageously, the monitoring system comprises a database constructed during a regulation learning phase comprising correspondence data between measurements m carried out by the measuring device and actions aiT carried out in the tank, said correspondence data being generated automatically by a learning process, and at least any one of said first, second and third functions is determined from said correspondence data acquired from said database.

[0030] Thus, in the case where at least one of the three functions is unknown, it can be learned automatically with knowledge of a complete time series of biochemical reaction previously carried out by experts in the biochemical process.

[0031] Advantageously, the control device is further configured to construct a property learning model defining links associating properties P of the final product as a function of the corresponding physical variables and characteristics c. Knowledge of the property learning model makes it possible to obtain a product with properties P, knowing its characteristics c' by regulating its physical variables during the biochemical reaction process. Indeed, by knowing on the one hand the model which links the physical variables and the characteristics to the properties of the product, and on the other hand the input characteristics c (e.g. in the case of wine this could be the grape variety), the control device can during the biochemical reaction constrain the physical variables to follow a certain profile to obtain desired properties p at the output.In other words, the control device makes it possible to adapt the characteristics of the elements placed in the tank and to regulate the evolution of the physical variables of biochemical reactions according to the characteristics of the elements placed in the tank in order to obtain specific desired properties of the product leaving the reactor.

[0032] Advantageously, the control device is configured to determine said links by estimating a learning function using a statistical model of the polynomial regression type, or kernel model, or neural network.

[0033] Thus, the learning function makes it possible to understand the influence of the characteristics of the elements reacted or of the evolution of the biochemical process on the properties obtained on the product.

[0034] Advantageously, the measuring device comprises at least one temperature probe, and a differential pressure sensor comprising two connected tubes of different lengths.

[0035] Advantageously, the measuring device is integrated into a float comprising: - at least one temperature probe, - a differential pressure sensor comprising two tubes of different lengths partly immersed in the liquid, - at least two electromagnetic, electrochemical or optical sensors comprising sensitive parts fixed on the walls of the float near the liquid, - a movement sensor configured to measure the movements of the float, - a wireless communication module comprising an antenna arranged on the upper part of the float, the communication module being intended to transmit the data relating to the measurements made by the sensors, - an energy management module comprising an electrical supply means arranged in the lower part of the float, and - an electronic circuit connected to all the sensors and electronic elements integrated into the float, the electronic circuit comprising a microprocessor intended to manage the acquisition of all the sensors, a first level of processing and the transmission of data.

[0036] The float is very compact and can be installed very simply in a wide variety of tanks regardless of their filling level. In addition, it allows density measurement very precisely both in a transient regime at the beginning of the process as well as in a permanent regime. In addition, the antenna being arranged in the emerging part of the float allows data to be transmitted in a simple, regular and precise manner.

[0037] Advantageously, the float comprises an inductive calibration and recharging module intended to calibrate the float and to remotely recharge the power supply means integrated in the float, said inductive calibration and recharging module being included in a protective case for the float.

[0038] Thus, the float is completely autonomous while allowing a fusion and synthesis of various measurements in a single on-board device.

[0039] The invention also relates to a method for in-situ monitoring of a biochemical process in a reactor comprising a tank intended to receive a liquid, said method comprising the following steps: - carry out measurements relating to the biochemical process at successive times, - transmit at said successive times observation data representative of at least the temperature and the density, and - control at said successive times the regulation of the biochemical reactor according to said observation data.

[0040] Other advantages and characteristics of the invention will appear in the detailed non-limiting description below. Brief description of the figures

[0041] Embodiments of the invention will now be described, by way of non-limiting examples, with reference to the accompanying drawings, in which:

[0042] [Fig-1] illustrates very schematically an in-situ monitoring system of a biochemical process in a biochemical reactor, according to one embodiment of the invention;

[0043] [Fig.2A] and

[0044] [Fig.2B] illustrate a control algorithm of the biochemical reactor, according to a preferred embodiment of the invention;

[0045] [Fig.3] is a directed graph representing a property learning algorithm of the compound at the outlet of the biochemical reactor, according to a preferred embodiment of the invention;

[0046] [Fig.4] illustrates very schematically an in-situ monitoring system of a biochemical fermentation process in a reactor, according to one embodiment of the invention;

[0047] [Fig.5] illustrates very schematically a method of in-situ monitoring of the process biochemical fermentation in relation to the system of [Fig.4];

[0048] [Fig.6] illustrates very schematically an in-situ monitoring system of a biochemical process in a biochemical reactor, according to a preferred embodiment of the invention;

[0049] [Fig.7] illustrates very schematically an instrumented float, according to a mode of preferred embodiment of the invention; and

[0050] [Fig.8] illustrates very schematically the arrangement of a sensor capacitive electromagnetic on the lateral surface of the float, according to the embodiment of [Fig.7]. Description of the embodiments

[0051] The principle of the invention is to automatically control a biochemical process taking place in a tank using measurements carried out in situ.

[0052] Thus, the invention proposes to adjust the control parameters according to the characteristics of the raw material and the properties of the desired product throughout the biochemical process according to its evolution. In addition, this control is carried out thanks to real-time and in situ monitoring of the evolution of the process. biochemical while allowing precise multi-parameter measurements using a single measuring device.

[0053] [Fig.l] very schematically illustrates a system 1 for in-situ monitoring of a biochemical process in a biochemical reactor 3, according to one embodiment of the invention.

[0054] The biochemical reactor 3 comprises a tank 5 intended to receive a liquid substrate 7 on which a biochemical process involving the action of microorganisms is to be carried out. Furthermore, the monitoring system 1 comprises a measuring device 8 and a control device 11.

[0055] According to one embodiment, the measuring device 8 can be fixed, for example, by the top of the tank 5 while being partially immersed in the liquid. This embodiment is advantageous in the case of small reactors.

[0056] According to another embodiment, the measuring device 8 can be embedded or integrated into a floating device intended to be introduced while floating in a partially submerged manner in the liquid 7 contained in the tank 5 (see Figs. 6 and 7). This embodiment is advantageous for any type of reactor and in particular, in the case of large tanks with a variable filling level.

[0057] The measuring device 8 is equipped with measurement sensors configured to carry out measurements relating to the biochemical process directly on the liquid 7 and at successive times in order to transmit observation data representative of these measurements to the control device 11 at successive times.

[0058] The control device 11 (for example a computer or a microcomputer) is configured to receive the observation data transmitted by the measuring device 8 and to control at successive times the regulation of the biochemical reactor 3 as a function of the observation data received from the measuring device 8.

[0059] The monitoring system 1 thus makes it possible to automatically control the biochemical process while reacting directly in real time in the event of detection of an anomaly in the process.

[0060] Advantageously, the measuring device 8 integrates at least one temperature probe 13 for measuring the temperature of the liquid 7, and at least one other measurement sensor 15 for measuring the density of the liquid and / or the gas flow rate.

[0061] In addition to the temperature measurement sensors 3 and density 15 of the liquid, the measuring device 8 comprises at least one other measurement sensor 151 for carrying out mechanical, electrical, optical or chemical measurements. The mechanical measurements comprise measurements of pressures in the liquid and / or gas flow rate and / or acceleration movement within the liquid medium. The electrical measurements comprise measurements of voltages and / or currents at one or more frequencies. associated with specific electrodes arranged in the measuring device 8, and / or measurements of resonance frequencies of oscillators integrated in the measuring device 8. The optical measurements include measurements of light intensities at one or more wavelength(s). The chemical measurements include measurements of pH, gases dissolved in the liquid, Redox potential, etc.

[0062] Furthermore, the measuring device 8 comprises an electronic circuit 17 connected to all the sensors and electronic elements and comprising a microprocessor 171 (or microcontroller) configured to manage the acquisition of all the sensors and to carry out pre-processing operations: filtering, averaging, counting, corrections, etc.

[0063] The measuring device 8 also comprises a transmission or communication module 19 for transmitting the observation data to the control device 11.

[0064] In the case of a measuring device 8 embedded in a float 9 (Figs. 6, 7) the measuring device can advantageously comprise an energy management module 21 comprising a battery to power all the electronic elements of the measuring device 8. In addition, the measuring device 8 can also comprise means which can be the same electrical measurement sensors for measuring the buoyancy level of the float 9.

[0065] According to a first embodiment, the microprocessor 171 is configured to determine at successive times vectors of physical variables of interest relating to the biochemical process as a function of the corresponding measurements carried out at successive times. Thus, the transmission or communication module 19 is configured to transmit to the control device 11 the observation data comprising the vectors of physical variables determined by the microprocessor 171 as well as the corresponding measurements.

[0066] According to another embodiment, the observation data transmitted by the measuring device 8 to the control device 11 only comprise the measurements made by the sensors 13, 15, 151. In this case, the control device 11 is configured to receive these measurements and to determine at successive times the vectors of physical variables of interest relating to the biochemical process as a function of the corresponding measurements.

[0067] Each vector of physical variables of interest comprises a first temperature or heat transfer variable determined from the temperature difference between two points. The vector of physical variables of interest comprises at least one other physical variable corresponding to the measurement(s) other than temperature. This at least one other physical variable may be one of: the second to twelfth variables of interest defined above (see also Figs. 6, 7).

[0068] The second variable concerns the release of gas. This can be determined by a local count of the number of bubbles using a differential pressure sensor (as described in patent application FR039275) which could advantageously be integrated into the float 9 or by gas flow measurements measuring the total quantity of gas leaving the tank.

[0069] The third variable concerns the density of the liquid determined by measurements of differential pressure between two depths using a pressure sensor (as described in application FR039275) which could advantageously be integrated into the float 9 and / or from measurements of the buoyancy levels of the float 9.

[0070] The fourth variable concerns the electrical conductivity of the liquid determined from electrical measurements of voltages and / or currents at one or more frequencies and / or by resonance frequency measurements using an electromagnetic sensor which could advantageously be integrated into the float 9.

[0071] The fifth variable concerns the permittivity of the liquid determined from electrical measurements preferably at several frequencies with a specific electrode or by resonance frequency measurements which could advantageously be integrated into the float 9.

[0072] The sixth variable concerns the movement of the liquid determined from the acceleration measurements via an inertial unit or accelerometer which could advantageously be integrated into the float 9.

[0073] The seventh variable concerns the measurement of the pH of the liquid, determined from electrical measurements with a specific electrode which could advantageously be integrated into the float 9.

[0074] The eighth variable concerns the redox potential determined from electrical measurements with a specific electrode which could advantageously be integrated into the float 9.

[0075] The ninth variable concerns dissolved oxygen determined from electrical measurements and / or optical measurements.

[0076] The tenth variable concerns the dissolved CO2 determined from electrical measurements with a specific electrode and / or from optical measurements which could advantageously be integrated into the float 9.

[0077] The eleventh variable concerns the optical absorption spectrum (turbidity, color) determined from optical measurements of at least two wavelengths.

[0078] The twelfth variable concerns the optical rotation determined from optical measurements using polarized light which varies by the presence of chiral molecules.

[0079] The control device 11 is configured to record the initial conditions of the biochemical process, for example, the composition of the liquid substrate 7 in the tank 5, such as the concentration of all the ingredients, such as sugar, the characteristics of the juice, the yeast strains, or by chromatographic analyses, etc. The control device 11 is also configured to recover the results of the treatment, record them and compare them to thresholds established by the user according to his professional expertise and the particularities of the product he wishes to obtain.

[0080] Furthermore, the control device 11 is configured to control actuators 23 for regulating the biochemical process. Indeed, the control device 11 is configured to control the regulation of the biochemical reactor 3 by at least one action At to be carried out on the tank 5. For example, the actions may include a modification of the stirring speed (circulation, rotation, etc.) or the activation of mixing if the gas flow rate is too low; a modification of the temperature (increasing or lowering the heating or cooling circuit setpoint), for example, lowering the temperature setpoint if the density variation is too high; a modification of the oxygen supply flow rate; a supply of nutrients (mineral salts, sugar, nitrogen, etc.)) or other elements for activating or stabilizing the biochemical process, such as the addition of oxygen or nitrogen if the release of gas slows down; the addition of yeasts or bacterial strains, etc.

[0081] Figs. 2A and 2B illustrate a control algorithm for the biochemical reactor, according to a preferred embodiment of the invention. More particularly, [Fig.2A] represents the control algorithm in graphical form and [Fig.2B] represents it in the form of a flowchart.

[0082] This algorithm is based on the definition of three time functions defined over a period T corresponding to the biochemical process. The first function / concerns the time prediction at a time t, of the values ​​of the physical variables 0 / knowing their values ​​at a previous time instant:

[0083] [Math.l] (1)

[0084] The second function g associates the measurements made by the sensors with the physical variables at a given time instant t:

[0085] [Math.2] g^t^mt (2)

[0086] The third function h determines the action at to be carried out in the tank according to the values ​​of the physical variables at a given instant of time t:

[0087] [Math.3] h: 9t^at (3)

[0088] The control algorithm is configured to determine for each time step t, a series of actions at with knowledge of the physical variables Qt obtained thanks to the measurements made by the sensors.

[0089] Fig. 2A is a directed graph that is organized in a matrix form with three rows and T columns and where each column represents a time step t, going from 1 to T. The first row corresponds to a first time sequence of nodes representative of the measurements mt, the second row corresponds to a second time sequence of nodes representative of vectors of the physical variables and the third row corresponds to a third time sequence of nodes representative of vectors of actions At of regulation. The nodes on each column are connected by arrows oriented from the first row to the second row and then from the second row to the third row. In addition, the nodes of the second row (i.e. the vectors of physical variables 0j) are also connected by arrows oriented sequentially in the increasing direction of time from 1 to T.

[0090] More particularly, on the sequential arrows of the second line, the control device 11 is configured to predict at successive times: on the one hand, predictive vectors of physical variables as a function of the previous vectors of physical variables resulting from the measurements and on the other hand, anticipatory measurements as a function of the predictive vectors of physical variables.

[0091] Furthermore, on the arrows oriented from the first line to the second line, the control device 11 is configured to calculate at successive times measurement deviations between the anticipatory measurements and the corresponding actual measurements. The control device 11 is then configured to correct at successive times the vectors of physical variables as a function of the corresponding measurement deviations.

[0092] On the third line, the control device 11 is configured to determine at successive times vectors of regulation actions as a function of the corresponding vectors of physical variables. This allows the control device 11 to control the regulation of the biochemical reactor 3 by triggering at successive times regulation actions based on the vectors of regulation actions.

[0093] Fig. 2B is a flowchart illustrating the control steps of the biochemical reactor for any instant g [1, T] during the biochemical process according to the directed graph of [Fig.2A].

[0094] In step E1, the control device is configured to make a prediction of each predictive vector of physical variables of the biochemical process using the first time function / defining values ​​of physical variables at time t knowing their values ​​at the previous time instant t-1:

[0095] [Math.4] (4)

[0096] This first function / is predetermined for example, by a standard Markov model of order at least 1.

[0097] Step E2 is a test to check whether the sensors are functional at the instant, that is, whether measurements made by the sensors are present at the instant. If so, then we move on to step E3, otherwise, we move on to step E5.

[0098] Step E3 concerns the prediction of each anticipatory measurement using the second function £ associating the measurements carried out by the measuring device 8 at the instant of time t, with the vector of physical variables calculated in step El:

[0099] [Math.5] (5)

[0100] The second function g can be predetermined by equations linking the measurements made by the measuring device 8 and the physical variables.

[0101] Step E4 concerns a correction of the vector of physical variables 0f by inversion (or pseudo inversion) of the second function and calculation of the difference A mt between the true measurements mt made by the float and those predicted:

[0102] [Math.6] A mt = m? - mt (6)

[0103] Step E5 concerns the determination of each regulation action vector using the third function k defining a correspondence at each instant of time t, between the actions to be carried out and the values ​​of the physical variables:

[0104] [Math.7] has t = h(0 t )

[0105] The third function can be predetermined, for example, by user expertise by pre-selecting thresholds on the physical variables for triggering actions.

[0106] In step E6, the control device 11 is configured to control the regulation of the biochemical reactor 3 by triggering at said successive times regulation actions based on said regulation action vectors.

[0107] Advantageously, it should be noted that in the event of a measurement fault, the control device is configured to continue to execute steps E1, E5 and E6 independently of the measurements. Thus, the regulation action vectors can be determined as a function of the predictive vectors of physical variables in the event of a fault in the measuring device 8.

[0108] This control operation can take the form of Kalman filtering if the / and % functions are linear [Welch95], or of particle-type stochastic filtering in any other case [Doucet09]. Advantageously, it does not require a prior database.

[0109] It will be noted that in the case where at least one of the three functions is unknown, it can be learned automatically with knowledge of at least one complete time series (i.e. t= 1 to T) of biochemical reaction associating mkT measurements and aiT actions carried out in the tank 5.

[0110] Indeed, a database can be constructed during a regulation learning phase comprising several sets of data from correspondence between measurements made by the measuring device 8 and the actions atr carried out in the tank 5. The correspondence data are generated automatically by a learning process. Indeed, the physical variables become hidden variables which are automatically estimated by the learning process, it is not necessary to know them as long as mw and aw are. The actions during the learning phase of regulating the biochemical process are carried out by an expert operator in the field.

[0111] Thus, at least any one of the first, second, and third functions can be determined from the acquired correspondence data. from the database. This determination of functions can be carried out using an algorithm of the 'Expectation Maximization (EM)' type [Roweis99] if it is assumed that they are linear.

[0112] In the case where the functions are not linear, they can be determined from the correspondence data using a neural network recurrent network (RNR) [Goodfellowlô], this same network itself serving as a control algorithm at the end of its training.

[0113] Once determined, the functions can be used for the control of any biochemical reaction according to the algorithmic scheme previously explained.

[0114] [Fig.3] is a directed graph representing a property learning algorithm of the compound at the outlet of the biochemical reactor, according to a preferred embodiment of the invention.

[0115] This graph illustrates the construction by the control device 11 of a property learning model of the compound at the outlet of the biochemical reactor 3. The graph comprises a first node NI representing a vector of physical variables, a second node N2 representing a vector of characteristics c of the elements reacted in the tank 5, and a third node N3 representing a vector of properties p of the product at the outlet. The property learning model makes it possible to connect each of the first O}? and second c nodes to the third node p by a directed arrow.

[0116]

[0117]

[0118]

[0119]

[0120] Indeed, at the end of the biochemical process, the compound obtained has different properties which depend on the evolution of the physical variable vectors during the treatment in tank 5 as well as on certain characteristics c of the elements reacted in tank 5. These characteristics form a vector of characteristics c comprising, for example, the nature of microorganisms, characteristics of the raw material, and possibly added nutrients. The nature of microorganisms comprises, for example, the type of yeasts, bacteria, microalgae and their quantity. The characteristics of the raw material comprise, for example, in the case of wine, the concentration of sugar, grape variety, terroir, meteorological, harvest and pressing data. In other applications, these raw material characteristics may comprise the nature of the wastewater for treatment or the bacterial culture substrate for bioreactors. The added nutrients comprise, for example, yeast activators, mineral salts, or other elements such as activated carbon, glucose, etc. Thus, the control device 11 constructs a property learning database comprising a collection of sets between the vectors of physical variables 0^ and the vectors of corresponding characteristics c. Furthermore, the properties at the end of the biochemical process form a vector of properties P revealing the quality of the compound. This can be evaluated by different devices, such as those of the type that cannot be integrated for in situ measurement. For example, the concentration of the different metabolites can be determined in the laboratory with off-line equipment (chromatography, IR spectroscopy, etc.). The presence and concentration of microorganisms can be carried out by culture and biological analyses (PCR). Other analyses concerning, for example, organoleptic observations (tastes, color, aroma, etc.) of the final product can be carried out by the expertise of specialists. The control device 11 thus records all the observations and results of the online or offline analyses forming the final product in the form of p property vectors of the output compound. Thus, after a set of biochemical processes, the property learning model has a collection of sets of p property vectors of the output compound.

[0121] Furthermore, the control device 11 is configured to construct links between the physical variable vectors 0^ recorded at the end of the biochemical reaction process and the corresponding characteristic vectors c making it possible to obtain the property vectors p of the output compound. These links between the physical variable vectors and the corresponding characteristic vectors are determined by the estimation of a learning function z:

[0122] [Math.8] (8)

[0123] Knowing at least one set of the collection of sets included in the database, the learning function z can be estimated using different statistical models such as polynomial regression, or kernel model, or neural network, etc. [BishopOô]. The estimation of the learning function z will be all the better as the database is large.

[0124] The learning function z makes it possible to understand the influence of the characteristics of the elements reacted or of the evolution of the biochemical process in the properties obtained on the compound. Thus, it becomes possible to adapt the characteristics of the elements placed in the tank 5 and to regulate the evolution of the physical variables of biochemical reactions by the control algorithm in order to obtain specific properties of the compound at the outlet of the reactor 3. Indeed, by knowing the link between the physical variables and the characteristics c which give the properties p, the control device can specify during the biochemical process a series of actions a^T to "constrain" the physical variables to follow a time profile which at the outlet would make the product have the desired properties.

[0125] [Fig.4] very schematically illustrates a system and a method for in-situ monitoring of a biochemical fermentation process in a reactor, according to one embodiment of the invention.

[0126] More particularly, [Fig.4] is an example of application of the monitoring method to a fermentation process in the wine field, according to the invention.

[0127] The biochemical reactor 3 comprises a tank 5 intended to receive a juice 7 on which the fermentation process must be carried out to produce the wine. As in [Fig.l], the monitoring system comprises a measuring device 8 (which can be integrated into a float) and a control device 11.

[0128] The measuring device 8 is configured to remain partially immersed in the juice 7 contained in the tank 5 while carrying out, at successive times, measurements of physical quantities and more particularly, measurements of temperatures, pressures, accelerations, and frequencies.

[0129] Advantageously, the measuring device 8 is configured to process the measurement data in order to calculate physical variables of interest such as, for example, the density, the CO2 flow rate, the conductivity, and the movement of the liquid, before transmitting to the control device 11 observation data comprising the physical variables and the corresponding measurements.

[0130] Alternatively, the measuring device 8 may be configured to transmit to the control device 11 observation data comprising only the corresponding measurements. In this case, it is the control device 11 which processes the measurement data to calculate the physical variables of interest. In both cases, the physical variables form indicators of interest which will enable the control device 11 to automatically provide direct feedback on the fermentation process.

[0131] The control device 11 makes it possible to enter, via an HMI 111, the recipe 25 to be applied to the process. The recipe may include the temperature profile 251, the conditions for activating the mixing 252, the oxygenation conditions 253, the conditions for activating the supply of nutrients 254, etc.

[0132] Recipe 25 comprises a set of threshold parameters and coefficients combined in a logical manner to carry out instructions or actions, as in the non-limiting examples below.

[0133] According to a first example: If (Density < Threshold x) or if (Gas flow rate < Threshold y) then: Stirring action according to a stirring speed = Coef b * (Threshold y - Gas flow rate).

[0134] According to a second example: If (Density > Threshold w) and if (gas flow rate < Threshold z) then: Addition of oxygen according to an Oxygen flow rate = Coef o * (Threshold z - Gas flow rate).

[0135] According to a third example: Temperature setpoint = Density * Coef k + threshold 1.

[0136] Furthermore, the control device 11 makes it possible to capture input characteristics 27 of the juice 7 which can come from additional analyses (online or offline) or from the winemaker's expertise. These input characteristics include, for example, the grape variety 271, the terroir 272, the oenological recipe 273, the vintage 274, the yeast strain 275, etc. It should be noted that the characteristics are input data for determining the recipe 25 to be applied to the process based on professional expertise.

[0137] Furthermore, if necessary, certain measurements can be carried out offline on an ad hoc basis and entered by the user or transmitted directly to the control device.

[0138] [Fig.5] very schematically illustrates a method of in-situ monitoring of the biochemical fermentation process in relation to the system of [Fig.4].

[0139] Block B1 concerns the initialization during which the control device 11 acquires the recipe to be applied (arrow Fl) to the fermentation process. This recipe can be entered by the operator or determined automatically from the input characteristics 27 (arrow F2) concerning the grape variety 271, the terroir 272, the oenological recipe 273, the vintage 274, and the yeast strain 275.

[0140] In block B2, the measuring device 8 carries out, at each time step t, measurements of temperatures, pressures, accelerations, frequencies, etc.

[0141] In block B3, the measuring device 8 transmits at each time step t, to the control device 11 (arrow F3) observation data comprising the measurements mt carried out. Optionally, the control device 11 is also configured to acquire other measurements carried out for example offline on an ad hoc basis.

[0142] In block B4, the control device 11 calculates physical variables St from the corresponding measurement data mt, thus forming vectors of physical variables. A vector of physical variables may include temperature (or heat transfer), density, agitation, CO2 flow rate, acidity, sugar concentration, conductivity, etc. These parameters of interest allow the control device 11 to determine immediate feedback by controlling (arrow F4) the actuators 23 for regulating the biochemical process.

[0143] In blocks B5-B8, the control device 11 carries out at each time step t, the steps described in relation to [Fig.2B].

[0144] In block B5, the control device 11 predicts a predictive vector of physical variables using the first function f according to equation (4).

[0145] In block B6, the control device 11 predicts anticipatory measurements using the second function g according to equation (5), calculates the deviations between the true measurements made by the measuring device 8 and those predicted according to equation (6) and uses them to correct the vector of physical variables 6t.

[0146] In block B7, the control device 11 calculates the regulation action vector using the third function h, according to equation (7) and displays the values ​​of the parameters useful for regulation.

[0147] In block B8, the control device 11 controls the regulation of the fermentation process by triggering regulation actions based on the regulation action vector and on the recipe and instructions specific to the process.

[0148] These actions make it possible to act immediately on the fermentation process to regulate parameters according to the recipe and / or to prevent and avoid anomalies of the 'fermentation stoppage' type when the yeasts or bacteria no longer act.

[0149] A first example of a regulation action may be an increase or a decrease in the temperature of the juice 7 according to the temperature profile 251 specified at the start. Indeed, the development of bacteria depends mainly on the temperature and this acts among other things on the aromas of the wine. Thus, the temperature profile is defined at the start by the oenologist or wine expert according to the type of wine and the desired aroma.

[0150] A second regulating action may be the mixing 252 and mixing speed according to the density and / or the gas flow rate. A third regulating action may be the oxygenation 253 and the amount of oxygenation according to the gas flow rate. A fourth action may be the addition of yeast 254 according to the temperature, and / or the density, and / or the gas flow rate, etc.

[0151] Furthermore, the control device 11 is advantageously configured to create output databases 31 which will allow algorithmic learning on the basis of organoleptic assessments such as the evaluation of the quality of the product with regard to the desired aromatic profile.

[0152] Indeed, in block B9, at the end of the fermentation process, the control device 11 records (arrow F5) the properties 33 of the final product (for example, the evaluation and annotation of the final product made by the oenologist).

[0153] In block B10, the control device 11 constitutes a database 31 for learning the properties of the final product, grouping together the characteristics of the liquid at the start of the process, the recipe parameters, the recording of the corresponding physical measurements and variables, the complementary analyses (on-line or off-line), and evaluations and properties P of the quality of the final product.

[0154] Thus, by making recordings at the end of each new fermentation process, the property learning database 31 then comprises a collection of physical variables of the, corresponding characteristics c, and evaluations and properties P of the final product at the output of each process. This makes it possible to automate the feedback by determining, using different statistical models, a correlation or a learning function z associating the properties P of the final product as a function of the physical variables 6^ and the corresponding characteristics c, according to equation (8). This makes it possible to optimize all the process control parameters, for example, the thresholds and coefficients of the recipe in order to obtain the best results.

[0155] [Fig.6] very schematically illustrates a system 1 for in-situ monitoring of a biochemical process in a biochemical reactor 3, according to a preferred embodiment of the invention.

[0156] This embodiment is similar to that of [Fig.l] except that the measuring device 8 is advantageously integrated into a float 9 intended to be introduced by floating in a partially submerged manner in the liquid 7 contained in the tank 5. The float 9 makes it possible to compensate for any variability in the filling levels of the tank 5. This embodiment is advantageous in the case of large tanks 5, such as for example for the production of fermented drinks.

[0157] [Fig.7] illustrates very schematically an instrumented float, according to a mode of preferred embodiment of the invention.

[0158] The float 9 of a predetermined mass according to the desired measurement range is provided with a rod 91 extending upwards and a plunger 92 intended to make it partially immerse and to ensure its verticality in the liquid 7 when it is introduced into the tank 5.

[0159] The measuring device is integrated into the float 9 so that the latter comprises an electronic circuit 17, one or more temperature probes 13a, 13b, mechanical, electrical, and possibly optical measurement sensors 15 as well as a communication module 19, and an energy management module 21.

[0160] The mechanical measurement sensors comprise a differential pressure sensor 151 and a movement sensor 152. The electrical measurement sensors comprise at least two electromagnetic sensors 153a, 153b and possibly sensors 154 for measuring pH, for measuring gases dissolved in the liquid, for measuring the Redox potential, etc. Possible optical measurement sensors 155 may comprise sensors for measuring light intensity.

[0161] The electronic circuit 17 is connected to all the sensors 13, 15 and electronic elements 17, 21 integrated in the float 9 and comprises an analog-digital converter, a memory, and a microcontroller or microprocessor 171 intended to manage the acquisition of all the sensors.

[0162] The temperature probes 13a, 13b integrated in the float 9 are advantageously offset at the top and bottom of the wide part of the float 9 to measure a possible temperature gradient.

[0163] The differential pressure sensor 151 is arranged on the lower part of the float 9 and comprises two tubes 151a, 151b connected to each other of different lengths as well as a piezoelectric membrane (not shown) which deforms under the effect of a pressure difference generating a differential voltage. The free ends of the two tubes 151a, 151b are configured to be immersed in the liquid 7 at two distinct depths, thus creating in the circuit of the pressure sensor 151, a differential voltage representative of the differential pressure between the two distinct depths. The microprocessor 171 of the electronic circuit 17 is configured to determine the density of the liquid 7 as a function of the differential pressure.

[0164] Furthermore, the generation of gas bubbles at the ends of the tubes 151a, 151b influences the measurement of the differential pressure. More particularly, when the bubbles are detached, the sensor 151 generates voltage peaks representative of the appearance of bubbles so that each peak coincides with the appearance of a gas bubble at the end of one of the two tubes 151a, 151b. The microprocessor 171 of the electronic circuit 17 is configured to count the number of bubbles in order to determine the flow rate of the gas (generally CO2). This is explained in detail in the applicant's patent application FR3039275.

[0165] Thus, the differential pressure sensor 151 makes it possible to quantify the gas release resulting from the reaction in the tank 5 and to determine the density of the liquid 7. However, at the beginning, when the float 9 is placed in the tank 5, a portion of the liquid rises in the tubes 151a, 151b before they fill with gas and consequently, during this transient regime, it is not possible to have a precise measurement of the density. On the other hand, in steady state, the tubes 151a, 151b are filled with gas and consequently, the differential pressure sensor 151 delivers a very precise measurement of the density of the liquid 7.

[0166] Advantageously, the electromagnetic sensors 153a, 153b are not affected by this phenomenon and deliver an accurate measurement of the density both in transient and steady-state conditions. Thus, the float 9 comprises at least first 153a and second 153b electromagnetic sensors coupled to the electronic circuit 17. Each electromagnetic sensor 153a, 153b comprises an inductive-capacitive LC circuit establishing a resonator having a natural resonance frequency which depends on the fluidic environment of the resonator.

[0167] More particularly, each of the sensors 153a, 153b comprises a resonator which comprises an active reactance in the form of electrode(s), called reactance electrode(s) associated with a corresponding passive coupling element. The reactance electrode may be of a capacitive or inductive nature. In particular, if this electrode is of a capacitive nature, then the corresponding coupling element is of an inductive nature and vice versa, thus forming a resonator whose resonance frequency varies according to the buoyancy level of the float 9 and the nature of the liquid 7 contained in the tank 5.

[0168] [Fig.8] illustrates very schematically the arrangement of a capacitive electromagnetic sensor on the lateral surface of the float, according to the embodiment of [Fig.7].

[0169] As an example, we consider the case of an electromagnetic sensor 153 equipped with a capacitive reactance electrode. The reactance electrode 161 then comprises two conductive plates 161a and 161b arranged opposite each other on the internal surface of the float 9, matching the cylindrical shape of this surface. Thus, a tubular capacitor is formed by the two plates 161a and 161b. Given that the geometry of the capacitor is invariant, its capacitance C depends essentially on the nature and level of the liquid 7 contained in the tank 5. It will be noted that alternatively, the conductive armatures 161a and 161b of the sensor can be interdigitated.

[0170] Furthermore, the armatures 161a and 161b of the capacitor of capacitance C are connected to an inductance which can be a coil 162 of predetermined inductance L, thus forming a resonance circuit whose frequency depends on the capacitance C which can be variable and on the inductance L which is constant. The electromagnetic sensor 153 measures the resonance frequency thus allowing the microprocessor 171 to deduce the value of the capacitance C and consequently, the permittivity of the liquid at the level of the armatures 161a and 161b. Thus, the resonance frequency varies according to the level (i.e. liquid / air interface) of the liquid 7 in the tank 5 and its nature. The measurement by an electromagnetic sensor 153 of this resonance frequency thus provides information on the nature and configuration of this environment.

[0171] The first electromagnetic sensor 153a is integrated into the plunger so that when the float 9 is introduced into the liquid 7, the active electrode of the first electromagnetic sensor 153a is located in the fully submerged part. Thus, there is no air-liquid interface at the level of the armatures 161 and consequently, the measured resonance frequency is indicative only of the nature of the liquid 7. The first electromagnetic sensor 153a is thus configured to carry out a reference measurement relating to the physical properties of the liquid 7 and more particularly, to the electrical properties such as the conductivity and the permittivity of the liquid 7 contained in the tank 5.

[0172] The second electromagnetic sensor 153b is integrated into the rod 91 of the float 9. In particular, the active electrode of the second sensor 153b is arranged along the entire length of the rod 91. Thus, when the float 9 is introduced into the liquid 7, the active electrode always comprises a part partially immersed in the liquid. The second sensor 153b then measures the immersion level of the rod 91 and therefore of the volume of the displaced liquid indicative of the density of the liquid 7. Indeed, the variation in the resonant frequencies is indicative of the variation in the permittivity of the insulator between the armatures due to the displacement of the air-liquid interface. It will be noted that the relative permittivity of a liquid with respect to air is quite significant. In particular, the introduction of the float 9 into the liquid 7 increases the level of the liquid which consequently increases the capacitance generating a drop in the resonant frequency.

[0173] The microprocessor 171 is configured to calculate the density of the liquid 7 on the basis of the measurement of the volume of the displaced liquid carried out by the second sensor 153b and the measurement of the electrical properties of the liquid carried out by the first sensor 153a.

[0174] Indeed, by knowing the mass of the float 9, the microprocessor 171 can estimate the value of the density of the liquid 7 from the variation in the level of the liquid which is proportional to the volume of the liquid displaced by the float 9. Furthermore, the measurement of the electrical properties of the liquid 7 carried out by the first electromagnetic sensor 153a is a reference measurement which allows the microprocessor 171 to correct the first estimate and to determine the density of the liquid 7 with greater precision.

[0175] Thus, the electromagnetic sensors 153a, 153b give a precise measurement of the density of the liquid 7 which makes it possible to circumvent the measurement imprecision of the differential pressure sensor 151 at the start of the reaction. In addition, having two independent density measurements makes it possible to reinforce the precision of the measurement in steady state and possibly to detect any malfunction of the sensors.

[0176] The motion sensor 152 is, for example, an inertial unit or a three-axis accelerometer which can be integrated into the submerged part 92 of the float 9. This motion sensor 152 is configured to measure the movement of the float 9 and indirectly that of the liquid 7 and to provide this indirect information on the movements of the liquid to the microprocessor 171. Indeed, the liquid 7 can be subjected to a controlled stirring movement. The microprocessor 171 recovers the measurements provided by the motion sensor 152 to calculate the movement of the liquid 7. The measurement of the movements of the liquid 7 when it is subjected to controlled stirring makes it possible to estimate its viscosity. Estimating the viscosity could also be envisaged either by applying a rotation (for example with a stirrer) or by moving the float 9 away from its equilibrium position and measuring the time required to return there.

[0177] Advantageously, the motion sensor 152 is further configured to validate or correct the acquisition depending on whether the liquid 7 is at rest or in motion. It also makes it possible to eliminate artifacts linked to external disturbances.

[0178] The float 9 may also comprise other sensors 154 adapted to measure chemical properties of the liquid. For example, the float 9 may comprise a pH measurement sensor, sensors for measuring gases (O2, CO2) dissolved in the liquid, a sensor for measuring the Redox potential, a sensor for measuring the concentration of specific molecules using the ion selective electrode (ISE) technique (ion selective electrode), etc. Advantageously, the sensors 154 for measuring chemical properties may be integrated into the float 9 using DLC ​​(diamond like carbon) electrodes which make it possible to miniaturize these sensors. Their selection and use will depend on the needs of the application and the aggressiveness of the environment.

[0179] It should be noted that the acquisition frequency of each sensor can be defined according to the specificities of each measurement (for example: greater than 1Hz for differential pressure measurement, less than 0.1Hz for density measurement by the electromagnetic method, etc.).

[0180] The wireless communication module 19 comprises an antenna 20 integrated in the rod 91 of the float 9 and connected to the various sensors 13, 15 via the electronic circuit 17. The antenna 20 is configured to transmit the data relating to the measurements made by the sensors to the control device 11 according to a low-energy wireless communication system (for example: NFC, RFID, Blue tough, Wifi, etc.). Thus, the data are communicated by the emerging part of the float 9 to the control device 11 so that it can control the various actions to be carried out for the regulation of the biochemical process.

[0181] The energy management module 21 comprises a battery or other power supply means advantageously integrated into the plunger 92 of the float 9. Depending on the use, the autonomy required for the minimum duration of an acquisition will make it possible to size the power supply means. In the case where the power supply is a battery, the latter is preferably arranged in the lower part to serve as ballast. In contrast, a remote power supply would preferably be located in the non-submerged part.

[0182] Advantageously, the monitoring system 1 comprises an inductive calibration and recharging module 31 intended to calibrate the float 9 and to remotely recharge the battery integrated in the float 9. It will be noted that depending on the nature of the tank 5 and the liquid 7, the float 9 can be remotely powered continuously. The inductive calibration and recharging module 31 can advantageously be included in a case 33 for protecting the float 9.

[0183] The microprocessor 171 of the electronic circuit 17 in relation with the analog-digital converter and the memory is configured to manage the acquisition of all the sensors 13, 15 and to carry out pre-processing: filtering, averaging, counting and corrections thus making it possible to transmit only the useful part of this information. For example, the microprocessor 171 can apply corrective factors for measurements, for example, of temperature and pressure, by not taking into account the measurements in the event of agitation, or by applying corrective factors to the measured values ​​if the sensors are sensitive to temperature and pressure conditions.

[0184] Furthermore, the microprocessor 171 is configured to merge all the measurement data in order to detect any anomaly originating, for example, from deterioration or fouling of a sensor. It is configured to issue an alert for preventive maintenance. The microprocessor 171 is also adapted to measure the response of the system when the liquid is subjected to an external action: agitation, modification of the temperature setpoint, supply of nutrients, yeasts, oxygenation, etc.

[0185] The microprocessor 171 is also configured to carry out processing of the measurements in order to extract the physical variables characterizing the liquid 7 which are necessary for controlling the biochemical process in the tank 5.

[0186] Thus, the electronic circuit 7 makes it possible to manage the acquisition of all the sensors, to extract the useful data and to format them before transmitting them to the control device 11.

[0187] Of course, various modifications can be made by those skilled in the art to the invention which has just been described, solely by way of non-limiting examples. Bibliographies

[0188] [Welch95]: An introduction to the Kalman filter, G. Welch, G. Bishop, technical report 2006.

[0189] [Doucet09]: A Tutorial on Particle Filtering and Smoothing: Fifteen years later, A. Doucet, AM Johansen, technical report 2012.

[0190] [Roweis99]: A unifying review of linear Gaussian models, S. Roweis, Z. Ghahramani, Neural Computation 1999.

[0191] [Goodfellowlô]: Deep learning, I. Goodfellow, Y. Bengio, A. Courville, Y Bengio, MIT press 2016.

[0192] [BishopOô]: Pattern recognition and machine learning, CM Bishop, Springer 2006.

Claims

Claims

1. System for in-situ monitoring of a biochemical process in a reactor comprising a tank (5) intended to receive a liquid (7), characterized in that it comprises: - a measuring device (8) intended to be introduced into said tank (5), said measuring device (8) being instrumented with sensors configured to carry out at successive times measurements relating to the biochemical process and to transmit at said successive times observation data representative of at least the temperature and the density of the liquid, the measuring device being integrated in a float (9) comprising at least one temperature probe (13a, 13b) and a differential pressure sensor (151) determining the density of the liquid; and - a control device (11) configured to control at said successive times the regulation of the biochemical reactor (3) as a function of said observation data received from the measuring device (8).

2. System according to claim 1, characterized in that said measurements comprise measurements of temperature and density of the liquid and at least one other type of measurements among the following measurements: mechanical measurements of pressures and / or gas flow and / or accelerations, and / or buoyancy level, electrical measurements of voltages and / or currents and / or resonance frequencies, and optical measurements.

3. System according to claim 1 or 2, characterized in that the measuring device comprises a microprocessor (171) configured to determine at said successive times vectors of physical variables relating to the biochemical process as a function of the corresponding measurements carried out at said successive times, the observation data transmitted by the measuring device (8) to said control device (11) comprising said vectors of physical variables and said corresponding measurements.

4. System according to claim 1 or 2, characterized in that the observation data transmitted by the measuring device to said control device (11) comprise said measurements and in that the control device is configured to determine at said times successive vectors of physical variables relating to the biochemical process as a function of the corresponding measurements.

5. System according to claim 3 or 4, characterized in that each of said vectors of physical variables comprises a temperature variable, a liquid density variable determined from the measurements of buoyancy levels and / or pressures and at least one other variable among the following variables: gas release, determined from the pressure measurements, electrical conductivity and / or permittivity of the liquid determined from the electrical measurements, movement of the liquid determined from the acceleration measurements, PH and / or redox potential determined from the electrical measurements, dissolved oxygen and / or CO2 determined from the electrical measurements and / or optical measurements, optical absorption spectrum and / or optical rotation determined from the optical measurements.

6. System according to any one of the preceding claims, characterized in that the control device (11) is configured to control the regulation of the biochemical reactor by at least one action among the following actions: modification of the stirring speed, modification of the temperature, modification of the oxygen supply flow rate, supply of nutrients or other elements for activating or stabilizing the biochemical process, supply of yeasts or bacterial strains.

7. System according to any one of claims 4 to 6, characterized in that the control device (11) is configured to: - predict at said successive instants predictive vectors of physical variables as a function of said previous vectors of physical variables from the measurements, - predict at said successive instants anticipatory measurements as a function of said predictive vectors of physical variables, - calculate at said successive instants measurement deviations between the anticipatory measurements and the corresponding actual measurements, - correct at said successive instants the vectors of physical variables as a function of said measurement deviations, - determine at said successive instants vectors of regulation actions as a function of said corresponding vectors of physical variables, and -control the regulation of the biochemical reactor by triggering at said successive times regulation actions based on said regulation action vectors.

8. System according to claim 7, characterized in that the control device (11) is configured to determine the regulation action vectors as a function of the predictive vectors of physical variables in the event of a measurement failure by the measuring device (8).

9. System according to claim 7 or 8, characterized in that the control device (11) is configured to - predict each predictive vector of physical variables using a first temporal function f defining values ​​of physical variables with knowledge of their values ​​at a previous instant of time, said first function f being predetermined by a standard Markov model of order of at least 1, - predict each anticipatory measurement using a second function g associating the measurements made by the float with the physical variables at a given instant of time, said second function g being predetermined by equations linking the measurements made by the measuring device and the physical variables, and - determination of each vector of regulation actions using a third function h defining a correspondence between the actions to be carried out and the values ​​of the physical variables at a given instant of time,said third function h being predetermined by a prior selection of action trigger thresholds.,

10. System according to claim 9, characterized in that it comprises a database constructed during a regulation learning phase comprising correspondence data between measurements carried out by the measuring device and actions carried out in the tank, said correspondence data being generated automatically by a learning process, and in that at least any one of the first, second and third functions is determined from said correspondence data acquired from said database.

11. System according to any one of the preceding claims, characterized in that the control device (11) is further configured to construct a property learning model defining links associating properties P of the final product depending on the corresponding physical variables and characteristics c.

12. System according to claim 11, characterized in that the control device (11) is configured to determine said links by estimating a learning function by means of a statistical model of the polynomial regression type, or kernel model, or neural network.

13. System according to any one of the preceding claims, characterized in that the differential pressure sensor (151) comprises two connected tubes (151a, 151b) of different lengths.

14. System according to any one of the preceding claims, characterized in that the measuring device is integrated in a float (9) comprising: - at least one temperature probe (13a, 13b), - a differential pressure sensor (151) arranged on the lower part of the float (9) and comprising two tubes (151a, 151b) connected of different lengths, - at least two electromagnetic sensors (153a, 153b) comprising armatures fixed on the walls of the float, - a movement sensor (152) configured to measure the movements of the liquid, - a wireless communication module (19) comprising an antenna (20) arranged on the upper part of the float, the communication module (19) being intended to transmit the data relating to the measurements made by the sensors, - an energy management module (21) comprising an electrical supply means arranged in the lower part of the float,and - an electronic circuit (17) connected to all the sensors and electronic elements integrated in the float, the electronic circuit comprising a microprocessor (171) intended to manage the acquisition of all the sensors and the transmission of the data.,

15. System according to claim 14, characterized in that it comprises an inductive calibration and recharging module (31) intended to calibrate the float (9) and to remotely recharge the power supply means integrated in the float, said inductive calibration and recharging module (31) being included in a case (33) for protecting the float.

16. Method for in-situ monitoring of a biochemical process in a reactor comprising a tank (5) intended to receive a liquid (7), characterized in that it comprises the following steps: - carrying out at successive times measurements relating to the biochemical process, said measurements being carried out by means of a measuring device integrated in a float (9) comprising at least one temperature probe (13a, 13b) and a differential pressure sensor (151) determining the density of the liquid, - transmit at said successive times observation data representative of at least the temperature and density of the liquid, and - control at said successive times the regulation of the biochemical reactor as a function of said observation data.

Images