Intelligent early warning system for abnormal state in fermentation process

Abstract

The invention relates to the technical field of biological fermentation detection, in particular to an intelligent early warning system for an abnormal state in a fermentation process. Comprising a fermentation environment parameter acquisition module, a fermentation liquid temperature and viscosity acquisition module, a dissolved oxygen concentration acquisition module, a metabolic characteristic acquisition module, a metabolic flux distribution calibration module, a weight coefficient adjustment module, a similarity matching module and a compensation control module. Wherein the fermentation liquid temperature and viscosity acquisition module comprises a detection probe rod, a plurality of viscosity sensors and a temperature sensor; the dissolved oxygen concentration acquisition module comprises a plurality of dissolved oxygen sensors, a plurality of vertical sliding grooves are formed in the detection probe rod, bearing sliding blocks are arranged in the vertical sliding grooves in a matched mode, and a height adjusting assembly is arranged on one side of the detection probe rod. According to the invention, multiple depth acquisition points are arranged, the characterization weight of the dissolved oxygen is calculated based on the viscosity data after temperature compensation, and the design of the detection unit capable of dynamically adjusting the height is combined, so that the accurate characterization and dynamic adaptation of the spatial heterogeneity of the dissolved oxygen in the fermentation liquor are realized.

Description

Technical Field

[0001] This invention relates to the field of bio-fermentation detection technology, specifically to an intelligent early warning system for abnormal states during fermentation. Background Technology

[0002] During the fermentation process, the fermentation process is easily affected by fluctuations in factors such as microbial activity and environmental parameters, which can cause it to deviate from the normal state. Real-time detection of its abnormal state is a necessary prerequisite for ensuring stable product quality, reducing production risks, and improving fermentation efficiency.

[0003] Existing technologies for detecting abnormal states in fermentation processes, such as the Chinese patent publication CN120257129A, disclose a method, system, device, and medium for early warning of abnormal states in fermentation processes. This method involves real-time acquisition of physical parameters such as temperature, pH, dissolved oxygen concentration, and stirring speed, as well as metabolic feature vectors such as NADH concentration, ATP concentration, and the dynamic NADH / ATP ratio obtained from Raman spectroscopy. These data are then fused to construct a multimodal data cube. Based on the metabolic feature vectors, a metabolic network model is built. Reaction rate parameters are optimized using physical parameters as boundary conditions to generate calibrated metabolic flux distribution data. The parameter monitoring weight coefficients are dynamically adjusted based on the deviation from preset benchmark data to generate a sensitivity weight vector. The multimodal data cube and weight vector are input into a spatiotemporal convolutional network to extract spatiotemporal features. These features are then matched with a preset abnormality pattern library (covering types such as dissolved oxygen anomalies and temperature gradient runaway). Based on the anomaly matching results, the warning type is determined, and compensation control commands are generated to adjust stirring speed, feeding rate, or aeration rate. This achieves accurate detection and real-time early warning control of abnormal states in the fermentation process.

[0004] However, in actual fermentation processes, due to differences in operating parameters such as stirring modes, rotation speed control precision, and aeration fluctuations during different fermentation stages, the dissolved oxygen in the fermentation broth is prone to spatial heterogeneity. Currently used single-point or multi-point fixed sensor monitoring methods are limited by insufficient spatial sampling representativeness, making it difficult to accurately characterize the spatiotemporal dynamics of the overall dissolved oxygen concentration level in the fermentation broth. This leads to quantitative deviations in key factors influencing microbial metabolic activity, ultimately affecting the accuracy and timeliness of abnormal state early warning during the fermentation process. Therefore, it is necessary to propose an intelligent early warning system for abnormal states during fermentation to address these technical shortcomings. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides an intelligent early warning system for abnormal states during fermentation. By setting multiple depth acquisition points and calculating dissolved oxygen characterization weights based on temperature-compensated viscosity data, combined with a detection unit design that can dynamically adjust height, the system achieves accurate characterization and dynamic adaptation of the spatial heterogeneity of dissolved oxygen in the fermentation broth.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows: an intelligent early warning system for abnormal states during fermentation, comprising: The fermentation environment parameter acquisition module is used to collect the pH value and stirring speed of the fermentation broth in real time. The fermentation broth temperature and viscosity acquisition module is used to set up several collection points corresponding to different liquid levels of the fermentation broth, acquire and collect the viscosity data and temperature data corresponding to each collection point, and compensate for each viscosity data based on the temperature data. The dissolved oxygen concentration acquisition module is used to acquire the dissolved oxygen concentration value corresponding to each collection point, and calculate the characterization weight of the dissolved oxygen concentration data of each collection point based on the viscosity data of the corresponding collection point and after temperature compensation optimization. Then, the characterization dissolved oxygen concentration of the whole fermentation broth is generated based on the characterization weight and dissolved oxygen concentration value of each collection point. The metabolic feature acquisition module is used to extract metabolic feature vectors using Raman spectroscopy. The metabolic flux distribution calibration module is used to predict and generate calibrated metabolic flux distribution data by taking temperature, pH, stirring speed and dissolved oxygen concentration as inputs, based on a genome-scale metabolic network model and combined with the calculation of reaction rate parameters. The weighting coefficient adjustment module is used to dynamically adjust the sensitivity weight vectors of temperature, pH value, stirring speed and dissolved oxygen concentration based on the deviation between real-time metabolic flux distribution data and preset metabolic baseline values. The similarity matching module is used to extract time series features of temperature, pH value, stirring speed, dissolved oxygen concentration and metabolic flux distribution data using a spatiotemporal convolutional network, perform similarity matching with a preset abnormal pattern library, output anomaly probability value, and output early warning results based on the anomaly probability value; The compensation control module is used to generate compensation control commands based on the early warning results.

[0007] The technical principle of the above scheme is as follows: First, the pH value of the fermentation broth and the stirring speed of the fermentation environment are collected. Then, the temperature and viscosity are collected, and the viscosity data is corrected by combining the temperature data. Based on the corrected viscosity data, the characterization weight of dissolved oxygen concentration is calculated, and the characterization weight and dissolved oxygen concentration are combined to generate the characterization dissolved oxygen concentration. Next, the metabolic feature module collects metabolic feature vectors. Then, the metabolic flux distribution calibration module integrates the above environmental and biochemical parameters based on the genome-scale metabolic network model to generate calibrated metabolic flux data. Then, the weight coefficient adjustment module dynamically optimizes the sensitivity weight of each parameter according to the metabolic flux deviation. Finally, the similarity matching module uses a spatiotemporal convolutional network to extract multi-parameter spatiotemporal features and match them with the abnormal pattern library, outputting early warning results and triggering compensation control commands.

[0008] The above approach has the following beneficial effects: 1. By deploying multiple depth collection points and calculating characterization weights, the overall dissolved oxygen concentration is generated by integrating the temperature-viscosity coupling relationship. This overcomes the shortcomings of insufficient representativeness in traditional single-point / fixed monitoring, improves the accuracy of spatiotemporal dynamic characterization of dissolved oxygen, and solves the problem of monitoring spatial heterogeneity of dissolved oxygen.

[0009] 2. Based on the genome-scale metabolic network model, metabolic flux is calibrated, and dynamic weight adjustment is combined to achieve a deep correlation between environmental parameters and metabolic state. This solves the problem of early warning lag caused by neglecting parameter correlation in traditional methods and improves the ability to couple and analyze metabolic and environmental parameters.

[0010] 3. Spatiotemporal convolutional networks are used to extract multi-parameter time series features, which are then matched with a pre-set abnormal pattern library to improve the recognition accuracy of complex abnormal states such as insufficient dissolved oxygen and uncontrolled temperature gradients, reduce the false alarm rate, and enhance the accuracy of abnormal identification.

[0011] Furthermore, in the fermentation liquid temperature and viscosity acquisition module, the specific method for temperature compensation of each viscosity data based on temperature data is as follows: By using the Arrhenius equation, the viscosity data at each sampling point at the corresponding temperature are normalized to the viscosity data at the standard temperature, eliminating the interference of temperature factors on viscosity measurement and allowing each viscosity data to reflect the differences in the physical properties of the fermentation layer at the corresponding sampling point.

[0012] Beneficial effects: Normalizing viscosity data from each collection point to a standard temperature eliminates the interference of temperature fluctuations on viscosity measurement, allowing viscosity data to truly reflect the rheological properties of the fermentation broth (such as changes in cell concentration and culture broth composition), improving the accuracy of viscosity-based dissolved oxygen correction models, and avoiding temperature dependence from masking the differences in true physical properties.

[0013] Furthermore, in the dissolved oxygen concentration acquisition module, based on the viscosity data corresponding to each collection point, the characterization weight of the dissolved oxygen concentration data at each collection point is calculated. The specific method for calculating the characterization of the dissolved oxygen concentration by combining the characterization weight of each collection point and the dissolved oxygen concentration data is as follows: Based on the principle that viscosity data is directly proportional to oxygen transport resistance and inversely proportional to dissolved oxygen representativeness, the reciprocal normalization method is used to calculate the characterization weight of dissolved oxygen concentration at each sampling point. The product of the characterization weight of each sampling point and the dissolved oxygen concentration is summed, and the summation result is used as the characterization of dissolved oxygen concentration.

[0014] Beneficial effects: Based on the principle that viscosity is proportional to oxygen transfer resistance, the reciprocal normalization method is used to calculate the dissolved oxygen characterization weights at each sampling point. The overall dissolved oxygen concentration is generated by weighted summation, which reduces the weight of high viscosity regions (low oxygen transfer efficiency) and increases the weight of low viscosity regions (high representativeness), thus solving the problem that the traditional arithmetic mean method cannot reflect spatial heterogeneity.

[0015] Furthermore, in the dissolved oxygen concentration acquisition module, after calculating and generating a characterizing dissolved oxygen concentration, the relative deviation of the dissolved oxygen concentration values ​​at each collection point is compared based on the characterizing dissolved oxygen concentration. When the relative deviation is greater than 30%, an abnormal state warning is triggered.

[0016] Beneficial effects: By comparing the relative deviations between the dissolved oxygen concentration at each collection point and the characterized dissolved oxygen concentration, an early warning is triggered when the deviation exceeds 30%. By utilizing the characteristic of the characterized value to fuse the temperature-viscosity correction relationship, dual monitoring of spatial distribution anomalies of dissolved oxygen and temperature / viscosity coupling anomalies is achieved, improving the comprehensiveness of anomaly detection.

[0017] Furthermore, the fermentation liquid temperature and viscosity acquisition module includes a detection probe, a number of viscosity sensors and temperature sensors corresponding to a certain number of acquisition points; The dissolved oxygen concentration acquisition module includes a number of dissolved oxygen sensors corresponding to the number of collection points, and several dissolved oxygen sensors are integrated on the detection probe.

[0018] Beneficial effects: Integrating viscosity sensor, temperature sensor and dissolved oxygen sensor into the same detection probe enables spatial integration of multi-parameter sensors, reduces the interference of multiple probe insertions on the fermentation broth flow field, simplifies the installation process, and improves the system's structural compactness and ease of operation.

[0019] Furthermore, the detection probe has vertical grooves axially arranged to match the number of sampling points. Each vertical groove is fitted with a bearing slider. A set of detection units consists of a dissolved oxygen sensor, a viscosity sensor, and a temperature sensor. Several detection units are integrated and installed on different bearing sliders. The detection probe is equipped with a height adjustment component on one side for adjusting the sliding distance of each bearing slider in the vertical groove.

[0020] Beneficial effects: By sliding the supporting slider in the vertical groove, the integrated sensor group can adjust its depth along the axial direction of the detection probe. Combined with the height adjustment component, it can achieve independent positioning of multiple detection units, adapt to the dynamic spatial distribution requirements caused by changes in the fermentation broth surface, and improve the synchronous acquisition accuracy of microenvironment parameters at different depths.

[0021] Furthermore, the height adjustment assembly includes a fixed sleeve and an adjusting rod, with the adjusting rod rotatably connected inside the fixed sleeve; The fixed sleeve is fixedly connected to one side of the detection probe. The side wall of the fixed sleeve near the detection probe has a sliding groove with a corresponding vertical sliding groove. Several sliding cylinders corresponding to the load-bearing sliders are slidably fitted inside the fixed sleeve. The several sliding cylinders are fixedly connected to the corresponding load-bearing sliders, and the several sliding cylinders are threadedly engaged with the adjusting rod.

[0022] Beneficial effects: By using the threaded engagement between the adjusting rod inside the fixed sleeve and the sliding cylinder, the rotational motion is converted into the axial displacement of the sliding cylinder, which drives the load-bearing slider to adjust synchronously, realizing the mechanical linkage adjustment of multiple detection units, simplifying the operation process and ensuring the stability and positioning accuracy of the adjustment process.

[0023] Furthermore, several dissolved oxygen sensors are fixedly installed at a 45° angle to the axis of the detection probe.

[0024] Beneficial effects: Installing the dissolved oxygen sensor at a 45° tilt angle allows any bubbles that may have been generated during fermentation to slide along the tilt direction of the sensor surface under the action of buoyancy, reducing the bubble residence time and adhesion area, lowering the probability of gas film formation, ensuring direct contact between the sensor and the fermentation liquid, and maintaining the accuracy of dissolved oxygen detection.

[0025] Furthermore, the axial length of several vertical grooves decreases gradually along the installation height, and the number of spiral coils of several sliding cylinders and adjusting rods in threaded engagement is the same, while the pitch decreases gradually along the height direction.

[0026] Beneficial effects: By using a gradient decreasing design of the axial length of the vertical slide groove and the pitch of the sliding cylinder, the sliding cylinders produce differentiated displacements when the adjusting rod rotates, matching the adjustment range of slide grooves of different heights, avoiding mechanical interference between multiple detection units during the adjustment process, and improving the reliability of synchronous depth adjustment.

[0027] Furthermore, a handle is fixedly connected to the top of the adjusting lever.

[0028] Beneficial effects: The handle design at the top of the adjusting lever provides a convenient point of force for manual operation. By rotating the handle, multiple detection units can be adjusted synchronously without complicated tools, improving the ability to respond quickly to changes in the fermentation broth level and simplifying the on-site operation process.

[0029] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the operation of the intelligent early warning system for abnormal states during fermentation in this invention. Figure 2 This is an isometric view of the installation of the detection probe in an embodiment of the intelligent early warning system for abnormal states during fermentation of the present invention; Figure 3 This is an isometric sectional view of the detection probe in an embodiment of the intelligent early warning system for abnormal states during fermentation of the present invention.

[0031] The reference numerals in the accompanying drawings of the instruction manual include: 1. Detection probe; 2. Viscosity sensor; 3. Temperature sensor; 4. Dissolved oxygen sensor; 5. Vertical slide groove; 6. Bearing slider; 7. Fixed sleeve; 8. Adjusting rod; 9. Sliding groove; 10. Sliding cylinder; 11. Rotary handle. Detailed Implementation

[0032] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0034] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0035] The following detailed description illustrates the specific implementation method: Example 1:

[0036] This embodiment provides an intelligent early warning system for abnormal states during fermentation, specifically as follows: Figure 1 As shown, it includes a fermentation environment parameter acquisition module, a fermentation broth temperature and viscosity acquisition module, a dissolved oxygen concentration acquisition module, a metabolic characteristic acquisition module, a metabolic flux distribution calibration module, a weight coefficient adjustment module, a similarity matching module, and a compensation control module.

[0037] Fermentation environment parameter acquisition module: Real-time acquisition of pH value and stirring speed of fermentation broth; Specifically, pH value acquisition is achieved by inserting an online pH composite electrode into a designated monitoring point in the fermentation broth. The sensitive membrane of the pH composite electrode selectively responds to hydrogen ions in the fermentation broth, outputting a mV-level electrical signal in real time, which is then converted into pH value data by a transmitter; Stirring speed acquisition is achieved by recording and obtaining the output power of the motor used for fermentation stirring, which is then converted into stirring speed data.

[0038] Conventional fermentation process monitoring techniques also require the collection of dissolved oxygen concentration in the fermentation broth to reflect the oxygen consumption rate of microbial respiration metabolism and the dynamic balance of the oxygen supply system in real time (one of the assessment indicators of abnormal conditions). However, in actual fermentation processes, different fermentation processes, different stirring modes, and different ventilation fluctuations can all lead to heterogeneity in the spatial distribution of dissolved oxygen concentration. The single-point or multi-point fixed sensor monitoring methods commonly used in existing technologies are difficult to accurately characterize the spatiotemporal dynamic characteristics of the overall dissolved oxygen concentration level in the fermentation broth.

[0039] Therefore, the intelligent early warning system proposed in this embodiment is equipped with several collection points corresponding to different liquid levels of the fermentation broth. This embodiment uses three collection points as an example for illustration (corresponding to the upper, middle, and bottom heights of the fermentation broth, respectively). Fermentation broth temperature and viscosity acquisition module: By installing viscosity and temperature acquisition devices corresponding to three acquisition points respectively, viscosity and temperature data corresponding to each acquisition point are obtained and collected. Since the viscosity of the fermentation broth is significantly negatively correlated with the dissolved oxygen concentration (i.e., increased viscosity leads to increased mass transfer resistance and decreased diffusion coefficient of oxygen in the fermentation broth, resulting in a more significant gradient in the dissolved oxygen concentration distribution), this embodiment can establish a dissolved oxygen concentration correction model based on viscosity parameters by acquiring viscosity data, effectively compensating for the dissolved oxygen detection deviation caused by viscosity changes, thereby achieving an increase in the accuracy of dissolved oxygen concentration detection. Furthermore, the viscosity of the fermentation broth exhibits a significant temperature dependence (increased temperature typically leads to decreased viscosity), and slight temperature differences may exist between different regions (upper, middle, and bottom) of the broth during fermentation, or the overall temperature may fluctuate due to metabolic heat production and temperature control system variations. Directly using the original viscosity value would mask the changes in the rheological properties of the fermentation broth itself (such as viscosity changes caused by variations in cell concentration and culture composition). Therefore, based on the collected temperature and viscosity data from each sampling point, the viscosity data at the corresponding temperature at each sampling point were normalized to the viscosity data μ1, μ2, and μ3 at the standard temperature using the Arrhenius equation. This eliminates the interference of temperature factors on viscosity measurement, ensuring that each viscosity data reflects the differences in the physical properties of the fermentation layer at the corresponding sampling point. The specific calculation process is as follows: The Arrhenius equation is:

[0040] Applied to the viscosity data normalization process described in this embodiment:

[0041] The standard viscosity at temperature T. Pre-exponential factor, R is the activation energy, R is the gas constant, and T is the absolute temperature.

[0042] The requirement is to first determine the standard viscosity of the fermentation broth at different temperatures through experiments, and then fit the result. and Then, the measured viscosity at each sampling point was analyzed. Substitute into the equation:

[0043] Calculate standard temperature Normalized viscosity at (e.g., 25°C).

[0044] Dissolved oxygen concentration acquisition module: By deploying three viscosity sensors at corresponding collection points, the dissolved oxygen concentration value at each collection point is obtained. Based on the temperature-compensated and optimized viscosity data μ1, μ2, and μ3 at the corresponding collection points, the characterization weight of the dissolved oxygen concentration data at each collection point is calculated. Then, based on the characterization weights and dissolved oxygen concentration values ​​at each collection point, the characterization dissolved oxygen concentration of the overall fermentation broth is calculated and generated. The specific process is as follows: Based on the principle that viscosity data is directly proportional to oxygen transport resistance and inversely proportional to dissolved oxygen representativeness, the reciprocal normalization method is used to calculate the characterization weight of dissolved oxygen concentration at each sampling point, as shown in the following formula:

[0045] In the formula, Let be the characterization weight of dissolved oxygen at point i. The viscosity value at point i after calibration; the mechanism for the change in the characterization weight is as follows: Viscosity value The higher the value, the greater the mass transfer resistance of dissolved oxygen, and the lower the representativeness of the data from that sampling point for overall dissolved oxygen. Therefore, the weighting is... The smaller the viscosity, the higher the viscosity; conversely, the higher the viscosity, the greater the weight given to low-viscosity regions. For example, if the viscosity at a certain sampling point... Increase (e.g., increase bacterial concentration), 1 / Reduced, leading to The concentration of dissolved oxygen at this point decreases, thus reducing its proportion in the overall calculation and avoiding misleading the overall assessment by data from high viscosity regions (while the specific curve representing the weight change with viscosity needs to be determined through specific experiments based on the characteristics of different fermentation solutions).

[0046] The product of the characterization weights at each collection point and the dissolved oxygen concentration is summed, and the summation result is used as the characterization of the dissolved oxygen concentration, as shown in the following formula:

[0047] In the formula, To characterize dissolved oxygen concentration, Dissolved oxygen concentrations collected at each sampling point.

[0048] In addition, in the dissolved oxygen concentration acquisition module, after calculating and generating a characterization of the dissolved oxygen concentration, the relative deviation of the dissolved oxygen concentration values ​​at each collection point is compared based on the characterization of the dissolved oxygen concentration. The specific risk level classification is as follows: Low risk: <30% (normal fluctuations); Medium risk: 30%-70% (requires monitoring, such as slight oxygen deficiency); High risk: >70% (emergency intervention, such as severe deviation in metabolic flux); When the relative deviation is greater than 30%, an abnormal state warning is triggered. The advantage of this design is that the generated dissolved oxygen concentration is a comprehensive value that integrates the relationship between temperature and viscosity correction. The comparison of its relative deviation can not only directly reflect whether the spatial distribution heterogeneity of dissolved oxygen concentration is abnormal, but also indirectly reflect whether the temperature or viscosity parameters deviate from the normal range through this comprehensive value. This enables direct assessment of the abnormal state of multi-parameter coupling in the fermentation process, avoids missing potential abnormal risks due to single parameter monitoring, and improves the efficiency and comprehensiveness of abnormal warning.

[0049] Metabolic feature acquisition module: Molecular vibrational spectral data of fermentation broth are acquired by Raman spectroscopy, and key metabolic feature vectors, including NADH concentration, ATP concentration and NADH / ATP dynamic ratio, are extracted using chemometric methods (such as partial least squares regression) to capture the core indicators of microbial metabolic state (metabolic rate) and reflect cellular energy metabolism and physiological activity.

[0050] Metabolic flux distribution calibration module: Taking temperature, pH, stirring speed and dissolved oxygen concentration as inputs, based on the genome-scale metabolic network model and combined with the calculation of reaction rate parameters, it predicts and generates calibrated metabolic flux distribution data; through the coupling of the genome-scale metabolic model and physical field constraints, it quantifies the spatial distribution and dynamic changes of metabolic reactions.

[0051] Weighting coefficient adjustment module: Based on the deviation between real-time metabolic flux distribution data and preset metabolic baseline values, dynamically adjust the sensitivity weight vectors of temperature, pH value, stirring speed and dissolved oxygen concentration; realize dynamic focus of monitoring each parameter, and improve sensitivity to key influencing factors in the fermentation process.

[0052] Similarity matching module: Utilizes a spatiotemporal convolutional network (STCN) to extract time-series features (such as dynamic temperature trends) of temperature, pH, stirring speed, dissolved oxygen concentration, and metabolic flux distribution data. These features are then matched against a pre-defined anomaly pattern library (containing six typical anomalies: excessive spatial heterogeneity of dissolved oxygen (relative deviation of oxygen concentration in different regions >30%), temperature gradient >2℃ / m, sudden pH changes (±0.5 / h), NADH / ATP ratio deviation from the baseline value ±30%, abnormal fluctuations in key metabolic fluxes (such as TCA cycle flux), and viscosity mutations (>50% / h). The module outputs an anomaly probability value (0-100%) and provides early warning results based on this probability. A deep learning model is used to accurately identify abnormal states, and spatiotemporal features are combined to improve matching accuracy.

[0053] The compensation control module generates graded early warnings based on the abnormal matching results (e.g., Level 1 warning: slightly insufficient dissolved oxygen; Level 2 warning: severely deviated metabolic flux), and calls the preset compensation rule library to generate an initial set of compensation parameters (e.g., adjusting ventilation, feeding speed, and stirring speed). After virtually verifying the compensation effect through a digital twin model, it outputs the optimal compensation control command.

[0054] The digital twin model is a virtualized model with pre-input feasible and calibrated data. It is a physical model built upon fermenter geometric parameters (volume, impeller type), heat / mass transfer equations, etc. Its main operation involves using real-time collected temperature, dissolved oxygen, and other data to correct model parameters and improve simulation accuracy. For example, when an alert is triggered indicating "dissolved oxygen spatial heterogeneity exceeds the standard (bottom dissolved oxygen 1.2 mg / L, top 4.5 mg / L)," the system generates a compensating control command to "increase the stirring speed from 300 rpm to 350 rpm," which is simulated through the digital twin model. Input: Current viscosity distribution ( =0.8cP, ==1.2cP), ventilation rate (1.0vvm); Output: After the simulated stirring speed is increased, the dissolved oxygen at the bottom rises to 2.5 mg / L and at the top drops to 3.8 mg / L. If the relative deviation is <30% (meets the normal threshold), the command will take effect.

[0055] Example 2:

[0056] The difference between this embodiment and Embodiment 1 is that, specifically as follows: Figure 2 As shown, the fermentation liquid temperature and viscosity acquisition module includes a detection probe 1, several viscosity sensors 2 (preferably coaxial cylindrical rheometers) with corresponding collection points, and a temperature sensor 3; while the dissolved oxygen concentration acquisition module includes dissolved oxygen sensors 4 with corresponding collection points, and several dissolved oxygen sensors 4 are integrated on the detection probe 1.

[0057] Among them, several dissolved oxygen sensors 4 are fixedly installed at a 45° angle to the axis of the detection probe 1. The 45° angled installation allows the bubbles generated during fermentation to slide along the inclined direction of the sensor surface under the action of buoyancy rather than to adhere vertically, thereby reducing the residence time and adhesion area of ​​the bubbles on the detection surface of the dissolved oxygen sensor 4, effectively reducing the probability of gas film formation, and ensuring direct contact between the dissolved oxygen sensor 4 and the fermentation liquid to maintain detection accuracy.

[0058] What's special is that, because the liquid level of the fermentation broth changes as fermentation progresses, specifically as follows: Figure 2 As shown in the illustration (this embodiment uses a conventional arc-shaped fermenter as an example of the fermentation environment), the detection probe 1 penetrates the top wall of the fermenter and its shape fits snugly against the inner wall of the fermenter. The upper part of the detection probe 1 is detachably connected to the fermenter. In this embodiment, a vertical groove 5 matching the number of sampling points is provided along the axial direction of the detection probe 1. Each vertical groove 5 is fitted with a bearing slider 6 (the bearing slider 6 is vertically adjustable within the vertical groove 5). A set of detection units is composed of a dissolved oxygen sensor 4, a viscosity sensor 2, and a temperature sensor 3. Several detection units are integrated and installed on different bearing sliders 6. With this design, during the fermentation process, the system can dynamically adjust the axial depth position of each detection unit according to the changes in the fermentation liquid level, so that the dissolved oxygen concentration, viscosity, and temperature are always located at the target monitoring points in the upper, middle, or bottom of the fermentation liquid. This effectively adapts to the dynamic spatial distribution requirements caused by changes in the liquid level during fermentation, improving the synchronous monitoring accuracy of microenvironmental parameters at different depths of the fermentation liquid.

[0059] Based on the aforementioned adjustable-height detection unit design, this embodiment includes a height adjustment component on one side of the detection probe 1 for adjusting the sliding distance of each bearing slider 6 within the vertical groove 5. Specifically, in conjunction with... Figure 2 and Figure 3 As shown, the height adjustment assembly includes a fixed sleeve 7 and an adjusting rod 8. The adjusting rod 8 is rotatably connected to the fixed sleeve 7 via a bearing. The fixed sleeve 7 is integrally formed on one side of the detection probe 1. The fixed sleeve 7 has several sliding grooves 9 with corresponding vertical sliding grooves 5 in number and position on the side near the detection probe 1. Several sliding cylinders 10 corresponding to the bearing sliders 6 are vertically slidingly adapted inside the fixed sleeve 7. The several sliding cylinders 10 are integrally formed with the corresponding bearing sliders 6. The several sliding cylinders 10 are threadedly engaged with the adjusting rod 8 (the axial length of the several vertical sliding grooves 5 decreases gradually along the arrangement height, and the number of spiral coils of the several sliding cylinders 10 and the adjusting rod 8 are the same, while the pitch decreases gradually along the height direction). The top of the adjusting rod 8 has an integrally formed handle 11. The specific operation and adjustment process is as follows: During the fermentation process, the user can rotate the top handle 11 of the adjusting rod 8 to drive the adjusting rod 8 to rotate around the bearing in the fixed sleeve 7. Since several sliding cylinders 10 are threadedly engaged with the adjusting rod 8, and the pitch of each sliding cylinder 10 decreases gradually along the height direction, while the axial length of the vertical groove 5 decreases gradually along the layout height, the sliding cylinders 10 at different heights will have different axial displacements when the adjusting rod 8 is rotated. This will drive the integrally formed bearing slider 6 to slide synchronously along the corresponding vertical groove 5, and finally achieve the precise placement of each detection unit at different depths in the fermentation liquid. The design is as follows: (1) The pitch of the inner wall of the sliding cylinder 10 at different heights and the vertical groove 5 are threadedly engaged with the adjusting rod 8 to drive the adjusting rod 8 to rotate around the bearing in the fixed sleeve 7. (1) The gradient matching of the length of the slide 5 ensures that the moving distance of each sliding cylinder 10 is matched with the axial length of the corresponding vertical slide 5 when the adjusting rod 8 is rotated, thus avoiding mechanical interference between different height detection units during the adjustment process; (2) The synchronous depth adjustment of multiple detection units can be achieved by driving with a single handle 11, which simplifies the operation process and improves the rapid response capability to the dynamic changes of the fermentation liquid surface; (3) Combined with the sliding adaptation characteristics of the vertical slide 5, each detection unit can flexibly cover the full depth range of the fermentation liquid from top to bottom, meet the spatial adaptation requirements of multiple depths and multiple parameters synchronous acquisition at different fermentation stages, and further improve the comprehensiveness and accuracy of fermentation process monitoring.

[0060] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. An intelligent early warning system for abnormal states during fermentation, characterized in that, include: The fermentation environment parameter acquisition module is used to collect the pH value and stirring speed of the fermentation broth in real time. The fermentation broth temperature and viscosity acquisition module is used to set up several collection points corresponding to different liquid levels of the fermentation broth, acquire and collect the viscosity data and temperature data corresponding to each collection point, and compensate for each viscosity data based on the temperature data. The dissolved oxygen concentration acquisition module is used to acquire the dissolved oxygen concentration value corresponding to each collection point, and calculate the characterization weight of the dissolved oxygen concentration data of each collection point based on the viscosity data of the corresponding collection point and after temperature compensation optimization. Then, the characterization dissolved oxygen concentration of the whole fermentation broth is generated based on the characterization weight and dissolved oxygen concentration value of each collection point. The metabolic feature acquisition module is used to extract metabolic feature vectors using Raman spectroscopy. The metabolic flux distribution calibration module is used to predict and generate calibrated metabolic flux distribution data by taking temperature, pH, stirring speed and dissolved oxygen concentration as inputs, based on a genome-scale metabolic network model and combined with the calculation of reaction rate parameters. The weighting coefficient adjustment module is used to dynamically adjust the sensitivity weight vectors of temperature, pH value, stirring speed and dissolved oxygen concentration based on the deviation between real-time metabolic flux distribution data and preset metabolic baseline values. The similarity matching module is used to extract time series features of temperature, pH value, stirring speed, dissolved oxygen concentration and metabolic flux distribution data using a spatiotemporal convolutional network, perform similarity matching with a preset abnormal pattern library, output anomaly probability value, and output early warning results based on the anomaly probability value; The compensation control module is used to generate compensation control commands based on the early warning results.

2. The intelligent early warning system for abnormal states during fermentation according to claim 1, characterized in that, In the module for acquiring fermentation liquid temperature and viscosity, the specific method for temperature compensation of various viscosity data based on temperature data is as follows: By using the Arrhenius equation, the viscosity data at each sampling point at the corresponding temperature are normalized to the viscosity data at the standard temperature, eliminating the interference of temperature factors on viscosity measurement and allowing each viscosity data to reflect the differences in the physical properties of the fermentation layer at the corresponding sampling point.

3. The intelligent early warning system for abnormal states during fermentation according to claim 2, characterized in that, In the dissolved oxygen concentration acquisition module, based on the viscosity data corresponding to each collection point, the characterization weight of the dissolved oxygen concentration data at each collection point is calculated. The specific method for calculating the characterization of dissolved oxygen concentration by combining the characterization weight of each collection point and the dissolved oxygen concentration data is as follows: Based on the principle that viscosity data is directly proportional to oxygen transport resistance and inversely proportional to dissolved oxygen representativeness, the reciprocal normalization method is used to calculate the characterization weight of dissolved oxygen concentration at each sampling point. The product of the characterization weight of each sampling point and the dissolved oxygen concentration is summed, and the summation result is used as the characterization of dissolved oxygen concentration.

4. The intelligent early warning system for abnormal states during fermentation according to claim 3, characterized in that, In the dissolved oxygen concentration acquisition module, after calculating and generating a characterizing dissolved oxygen concentration, the relative deviation of the dissolved oxygen concentration values ​​at each collection point is compared based on the characterizing dissolved oxygen concentration. When the relative deviation is greater than 30%, an abnormal state warning is triggered.

5. The intelligent early warning system for abnormal states during fermentation according to claim 4, characterized in that, The fermentation liquid temperature and viscosity acquisition module includes a detection probe (1), several viscosity sensors (2) with corresponding collection points, and a temperature sensor (3). The dissolved oxygen concentration acquisition module includes a number of dissolved oxygen sensors (4) corresponding to the number of acquisition points, and several dissolved oxygen sensors (4) are integrated on the detection probe (1).

6. The intelligent early warning system for abnormal states during fermentation according to claim 5, characterized in that, The detection probe (1) has a vertical groove (5) axially opened to match the number of collection points. Each vertical groove (5) is fitted with a bearing slider (6). A dissolved oxygen sensor (4), a viscosity sensor (2), and a temperature sensor (3) form a detection unit. Several detection units are integrated and installed on different bearing sliders (6). The detection probe (1) is provided with a height adjustment component on one side for adjusting the sliding distance of each bearing slider (6) in the vertical groove (5).

7. The intelligent early warning system for abnormal states during fermentation according to claim 6, characterized in that, The height adjustment assembly includes a fixed sleeve (7) and an adjusting rod (8), the adjusting rod (8) being rotatably connected inside the fixed sleeve (7); The fixed sleeve (7) is fixedly connected to one side of the detection probe (1). The side wall of the fixed sleeve (7) near the detection probe (1) is provided with a sliding through groove (9) corresponding to the vertical sliding groove (5). Several sliding cylinders (10) corresponding to the bearing sliders (6) are slidably fitted inside the fixed sleeve (7). Several sliding cylinders (10) are fixedly connected to the corresponding bearing sliders (6) respectively. Several sliding cylinders (10) are threadedly engaged with the adjusting rod (8).

8. The intelligent early warning system for abnormal states during fermentation according to claim 7, characterized in that, Several dissolved oxygen sensors (4) are fixedly installed at a 45° angle to the axis of the detection probe (1).

9. The intelligent early warning system for abnormal states during fermentation according to claim 8, characterized in that, The axial length of several vertical grooves (5) decreases gradually along the height of the arrangement. Several sliding cylinders (10) and adjusting rods (8) have the same number of spiral coils, while the pitch decreases gradually along the height direction.

10. The intelligent early warning system for abnormal states during fermentation according to claim 9, characterized in that, The top of the adjusting rod (8) is fixedly connected to a handle (11).

Images