A fermentation monitoring system for the co-production of microbial proteins using hydrogen oxidizing bacteria

Abstract

This application discloses a fermentation monitoring system for the co-production of microbial proteins using hydroxide bacteria, relating to the field of fermentation monitoring technology. It includes a thermal data acquisition module for acquiring the tank wall thermal radiation matrix; a metabolic field reconstruction module for receiving the tank wall thermal radiation matrix and calculating a temperature field uniformity index based on the numerical dispersion of all pixels in the metabolic intensity distribution map; an abnormal state determination module for triggering regional feature recognition logic on the metabolic intensity distribution map if the temperature field uniformity index exceeds a preset uniformity threshold range, generating a metabolic anomaly partition mask based on the pixel temperature value distribution; a flow field regulation generation module for receiving the metabolic anomaly partition mask and generating a flow field correction vector according to the anomaly type corresponding to the spatial region coordinates; and an execution control interface module for sending the flow field correction vector to the actuators of the fermenter. This application improves the efficiency of fermentation monitoring.

Description

Technical Field

[0001] This application relates to the field of fermentation monitoring technology, and in particular to a fermentation monitoring system that utilizes hydroxide bacteria to co-produce microbial proteins. Background Technology

[0002] Gas fermentation using hydroxyl bacteria to produce single-cell proteins is a significant direction in the field of biomanufacturing. This process relies on efficient mass transfer of hydrogen, oxygen, and carbon dioxide in the liquid phase and the effective removal of heat generated by microbial metabolism. However, in industrial-scale fermentation, the reaction system is not ideally homogeneous. As the fermenter volume increases, the gas-liquid mixing flow field becomes extremely complex, leading to significant differences in the microenvironment in different regions within the fermenter. Most existing fermentation control systems are based on the assumption of macroscopic homogeneity, making it difficult to cope with this complex heterogeneity challenge, often resulting in low fermentation efficiency, unstable protein yield, and low energy utilization.

[0003] Traditional fermentation monitoring methods typically rely on contact sensors, such as temperature probes and dissolved oxygen electrodes, inserted at specific locations within the fermenter to acquire discrete data from a single point or a few points. This monitoring method has a significant "blind spot," only providing feedback on the local state around the sensor and failing to characterize the spatial distribution characteristics of the entire fermenter. In high-density fermentation involving hydroxide bacteria, due to the high cell concentration and large viscosity variations, fluid dead zones can easily form outside the agitator's operating range, or high-intensity metabolic centers can form in localized areas. In such cases, data acquired by single-point sensors often exhibits lag and incompleteness, failing to promptly detect metabolic stagnation (dead zones) caused by impaired mass transfer or heat accumulation (hot spots) due to intense localized reactions, leading to misjudgments of the fermentation status by the monitoring system.

[0004] Furthermore, existing control methods are too crude and lack specificity in addressing the heterogeneity of the fermentation environment. Conventional temperature control is mainly achieved by adjusting the flow rate of jacket cooling water. This external heat exchange method is difficult to effectively eliminate heat accumulation in the center of the tank or specific flow field areas, easily leading to the persistence of internal thermal gradients. In terms of dissolved oxygen or dissolved hydrogen control, existing technologies usually adopt a uniform adjustment of stirring speed or total aeration rate. This "one-size-fits-all" overall adjustment strategy often has unintended consequences: increasing the overall stirring speed to eliminate local oxygen deficiency will lead to excessive shear forces in other areas, thereby damaging the shear-sensitive cell structure; or excessive cooling to reduce local temperature will suppress the metabolic activity of cells in other areas. This control method, lacking spatial resolution, limits further improvement in the spacetime yield of microbial proteins.

[0005] To address the aforementioned issues, there is an urgent need in this field to develop a fermentation monitoring system that can non-invasively and in real-time sense the differences in metabolic heat distribution within the fermenter and can perform zoning and precise fluid morphology control based on local flow field characteristics, in order to eliminate the heterogeneity of the microenvironment within the fermenter and achieve efficient and coordinated metabolism of the entire cell population. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this application provides a fermentation monitoring system that utilizes hydroxide bacteria to co-produce microbial proteins.

[0007] This application provides a fermentation monitoring system utilizing hydroxide bacteria to co-produce microbial protein, comprising: The thermal data acquisition module is used to acquire the tank wall thermal radiation matrix at a preset sampling frequency by using an array of infrared thermal imaging sensors deployed on the outer wall of the fermenter. The tank wall thermal radiation matrix contains the real-time temperature values ​​of each coordinate point on the surface of the tank. The metabolic field reconstruction module is used to receive the tank wall thermal radiation matrix, use a heat conduction inversion algorithm to map the tank wall thermal radiation matrix into a metabolic intensity distribution map, and calculate and generate a temperature field uniformity index based on the numerical dispersion of all pixels in the metabolic intensity distribution map. An abnormal state determination module is used to compare the temperature field uniformity index with a preset uniformity threshold range; if the temperature field uniformity index exceeds the preset uniformity threshold range, the regional feature recognition logic of the metabolic intensity distribution map is triggered, and a metabolic abnormality partition mask is generated based on the pixel temperature value distribution. The metabolic abnormality partition mask marks the spatial coordinates of the fermenter where there is an abnormal temperature gradient. The flow field regulation generation module is used to receive the metabolic abnormality partition mask and generate a flow field correction vector according to the abnormality type corresponding to the spatial region coordinates. The execution control interface module is used to send the flow field correction vector to the actuator of the fermenter. The flow field correction vector includes the blade angle of attack command and the micro-air injection opening command.

[0008] Preferably, the thermal data acquisition module is further used for: Acquire background environmental data collected by an ambient temperature sensor; The tank wall thermal radiation matrix is ​​subjected to differential correction operation with the background environment data, and the corrected tank wall thermal radiation matrix is ​​then determined based on the result of the differential correction operation. The corrected tank wall thermal radiation matrix is ​​transmitted to the metabolic field reconstruction module.

[0009] Preferably, the corrected tank wall thermal radiation matrix is ​​determined based on the result of the differential correction calculation, specifically including: Extract the corresponding real-time ambient temperature value from the background environment data, and multiply the real-time ambient temperature data with the preset wall reflectivity parameter to construct an environmental reflection noise matrix with the same dimension as the tank wall thermal radiation matrix; Subtract the corresponding coordinate position value in the environmental reflection noise matrix from the value of each pixel in the tank wall thermal radiation matrix to obtain the net radiation difference matrix. The preset distance attenuation compensation coefficient is retrieved to perform linear gain calculation on the net radiation difference matrix, thereby determining the corrected tank wall thermal radiation matrix.

[0010] Preferably, the metabolic field reconstruction module performs the following steps when generating the metabolic intensity distribution map: Recall pre-stored tank wall thermal resistance model data, which includes the tank wall thickness and the thermal conductivity of the tank wall material; By combining the thermal radiation matrix of the tank wall and the thermal resistance model data of the tank wall, the temperature value corresponding to the liquid film adhering to the tank wall is calculated. Then, a two-dimensional array is constructed based on the temperature value corresponding to the liquid film adhering to the tank wall, and the two-dimensional array is set as a metabolic intensity distribution map.

[0011] Preferably, the temperature value corresponding to the liquid film adhering to the wall inside the tank is calculated, and then a two-dimensional array is constructed based on the temperature value corresponding to the liquid film adhering to the wall inside the tank, and the two-dimensional array is set as a metabolic intensity distribution map, specifically including: The tank wall thermal resistance model data is analyzed to extract the pre-stored tank wall thickness value and the thermal conductivity of the tank wall material. Divide the tank wall thickness by the thermal conductivity of the tank wall material to calculate the thermal resistance parameter that characterizes the resistance to heat penetration through the tank wall. Obtain a preset heat flux density estimation factor, and multiply the preset heat flux density estimation factor with the thermal resistance parameter to generate a temperature compensation increment; The temperature compensation increment is accumulated point by point into the value of each pixel in the thermal radiation matrix of the tank wall, thereby determining the temperature value corresponding to the liquid film adhering to the tank wall.

[0012] Preferably, the specific formula for calculating the temperature field uniformity index is as follows: Calculate the ratio of the temperature standard deviation to the temperature average value of all valid pixels in the metabolic intensity distribution map, and use the ratio as the temperature field uniformity index.

[0013] Preferably, when the abnormal state determination module executes the region feature recognition logic to generate a metabolic abnormality partition mask, it specifically performs the following steps: Traverse all pixels in the metabolic intensity distribution map, filter out the set of pixels with temperature values ​​higher than a preset hotspot threshold, and mark the connected region where the set of pixels is located as a metabolic hotspot region. Filter out the set of pixels in the metabolic intensity distribution map whose temperature values ​​are lower than a preset cold point threshold, and mark the connected region where the set of pixels is located as a metabolic dead zone region; The location information of the metabolic hotspots and metabolic dead zones in the metabolic intensity distribution map is extracted, and then the location information is mapped to the physical hierarchical coordinate system of the fermenter to generate a metabolic anomaly partition mask containing physical hierarchical location information.

[0014] Preferably, the specific logic for generating the flow field correction vector is as follows: The physical level at which the metabolic hotspot region is located is determined, and commands to increase the blade angle of attack and decrease the micro-injection opening are generated. The physical level of the metabolic dead zone is located, and commands to increase the micro-injection air opening and fine-tune the blade angle of attack are generated.

[0015] Preferably, it also includes a closed-loop feedback verification module, used for: After the execution control interface module sends the flow field correction vector and a preset response time has elapsed, the thermal data acquisition module is triggered to acquire a new tank wall thermal radiation matrix again. The metabolic field reconstruction module is triggered to calculate a new temperature field uniformity index; Determine whether the new temperature field uniformity index falls within the preset uniformity threshold range. If it does not, generate a secondary correction coefficient and feed it back to the flow field control generation module.

[0016] Preferably, the flow field control generation module further includes a global perturbation submodule, used for: When the temperature field uniformity index exceeds a preset emergency threshold, a global blade angle of attack command is generated.

[0017] In summary, this application includes at least one of the following beneficial technical effects: 1. This application provides a fermentation monitoring system for the co-production of microbial proteins by hydroxyl bacteria. By deploying an array of infrared thermal imaging sensors on the outer wall of the fermenter, the system can non-invasively acquire the thermal radiation matrix covering the entire tank wall. Combined with a heat conduction inversion algorithm, the outer wall temperature can be accurately mapped to the internal metabolic intensity distribution map. This allows the previously invisible internal flow field differences, metabolic hotspots, and dead zones to be presented intuitively in the form of thermal images, effectively solving the problem of "visual blind spots". This enables operators or control systems to comprehensively and in real time grasp the true spatial metabolic state inside the fermenter, providing a data foundation for precise regulation. 2. By setting the differential correction operation in the thermal data acquisition module, external interference factors such as ambient temperature changes and tank wall reflections can be separated. By constructing an environmental reflection noise matrix and performing difference calculations, pure temperature data that only reflects the fermentation process can be obtained. This ensures that the basis for subsequent metabolic field reconstruction and abnormal state judgment remains highly accurate under the complex environmental conditions of industrial production sites, avoiding system misjudgments caused by ambient temperature fluctuations or light changes. 3. This application proposes a temperature field uniformity index, which transforms complex image information into a quantifiable and comparable single indicator. Instead of relying on manual observation or a single threshold alarm, it automatically and objectively determines whether the fermentation system is in an ideal homogeneous state by comparing the temperature field uniformity index with a preset threshold. When the uniformity index exceeds the range, it can achieve early warning and intervention before it seriously affects the overall yield, thus preventing the fermentation process from getting out of control. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of a fermentation monitoring system for co-producing microbial proteins using hydroxide bacteria, as described in this application embodiment. Detailed Implementation

[0020] The following description, in conjunction with the implementation of the present invention, is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0021] Application Overview: In existing technologies, monitoring of hydroxide bacterial fermentation largely relies on single-point temperature or dissolved oxygen probes, making it difficult to consider both the overall spatial distribution within the fermenter and local microenvironmental differences. Traditional methods, when dealing with uneven flow fields within large fermenters, are susceptible to interference from local dead zones or hot spots, leading to increased errors in metabolic state retrieval. Existing systems cannot simultaneously perceive the impact of internal fluid morphology changes on metabolic heat production, especially under high-density fermentation conditions where local overheating intensifies. Single control models exhibit systematic biases, failing to meet the demands of high-yield protein synthesis.

[0022] To address the aforementioned issues, the inventors discovered a heat-fluid coupling relationship between the tank wall's thermal distribution characteristics and internal metabolic intensity. They optimized the control strategy by establishing a thermal imaging-flow field inversion model. During the research, it was found that the surface temperature gradient is highly sensitive to internal metabolic hotspots, while the uniformity index, although less sensitive, provides good global characterization. Therefore, they proposed a method to dynamically trigger biomass regulation based on a uniformity threshold. Further experimental verification incorporated the mapping relationship between hotspot / dead zone distribution and flow field correction vectors into the model parameter dynamic adjustment mechanism, forming a closed-loop feedback system.

[0023] Specifically, the monitoring system first synchronously acquires high-resolution thermal infrared matrix data of the tank wall and environmental background data. Using differential correction and thermal resistance inversion algorithms, it calculates the temperature spectrum of the adhering liquid film, characterizing the internal metabolic intensity. Simultaneously, it extracts the temperature field uniformity index from the spectrum data as a spatial consistency indicator. When the uniformity index exceeds a set threshold, the system automatically triggers anomaly identification logic, using hot / cold point thresholds to pinpoint the specific level of metabolic imbalance and generating targeted blade angle of attack and micro-gas injection commands. During continuous monitoring, the system verifies the control effect in real time and generates secondary correction coefficients based on the temporal variation characteristics of the uniformity index through a closed-loop feedback mechanism, forming a dynamic optimization closed loop. For cases where the uniformity threshold is not exceeded, the current flow field state is maintained or minor global optimization is performed.

[0024] Compared to existing technologies, traditional methods rely on discrete point monitoring and lack spatial compensation mechanisms, making them prone to localized deactivation or overheating in chaotic flow fields. This innovative approach integrates non-contact thermal imaging with fluid dynamics control, achieving dynamic elimination of microenvironmental differences by establishing a metabolic heat-flow field coupling model. Unlike existing "one-size-fits-all" overall control methods, this approach intelligently adjusts impellers and gas supply based on real-time heat distribution characteristics, and continuously optimizes the fermentation environment through closed-loop feedback, significantly improving the space-time yield of large-scale bioreactors.

[0025] Through the above technical solution, this application effectively overcomes the problem of local metabolic imbalance caused by traditional single-point monitoring, and improves the metabolic consistency of the entire fermentation tank while ensuring real-time monitoring. The zoned targeted regulation mechanism combines the advantages of high efficiency in eliminating hot spots and precision in activating dead zones, and the adaptive correction function of model parameters ensures the stability of long-cycle fermentation.

[0026] After introducing the basic concept of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0027] Example This application discloses a fermentation monitoring system that utilizes hydroxide bacteria to co-produce microbial proteins.

[0028] Reference Figure 1A fermentation monitoring system utilizing hydroxide bacteria for co-production of microbial protein, comprising: The thermal data acquisition module is used to acquire the thermal radiation matrix of the fermenter wall by means of an infrared thermal imaging sensor array deployed on the outer wall of the fermenter at a preset sampling frequency. The thermal radiation matrix of the fermenter wall contains the real-time temperature values ​​of each coordinate point on the surface of the fermenter. The thermal radiation matrix of the fermenter wall refers to the set of two-dimensional thermodynamic data acquired by the infrared thermal imaging sensor array deployed on the outer wall of the fermenter. Specifically, it can be realized by using a high-resolution industrial-grade infrared thermal imager to perform grating scanning at a frequency of 2Hz to 10Hz. It is used to record the heat distribution state of the fermenter surface in the spatiotemporal dimension in real time and is the basic data source for realizing non-contact full-field monitoring. The metabolic field reconstruction module is used to receive the tank wall thermal radiation matrix, use a heat conduction inversion algorithm to map the tank wall thermal radiation matrix into a metabolic intensity distribution map, and calculate and generate a temperature field uniformity index based on the numerical dispersion of all pixels in the metabolic intensity distribution map. An abnormal state determination module is used to compare the temperature field uniformity index with a preset uniformity threshold range; if the temperature field uniformity index exceeds the preset uniformity threshold range, the regional feature recognition logic of the metabolic intensity distribution map is triggered, and a metabolic abnormality partition mask is generated based on the pixel temperature value distribution. The metabolic abnormality partition mask marks the spatial coordinates of the fermenter where there is an abnormal temperature gradient. The flow field control generation module is used to receive the metabolic abnormality partition mask and generate a flow field correction vector according to the abnormality type corresponding to the spatial region coordinates. The flow field correction vector refers to a set of specific execution instructions generated for a specific spatial region, specifically including a combination of instructions to increase the blade angle of attack / reduce air injection for hot spot regions and to fine-tune the angle of attack / increase air injection for dead zone regions, which is used to accurately eliminate local metabolic imbalances through fluid dynamics. The execution control interface module is used to send the flow field correction vector to the actuator of the fermenter. The flow field correction vector includes the blade angle of attack command and the micro-air injection opening command.

[0029] The working process and principle of this application are as follows: First, the thermal data acquisition module uses an infrared sensor array to collect the thermal radiation matrix of the tank wall, and performs differential correction and distance compensation in combination with background environmental data to output a high-fidelity corrected matrix; then, the metabolic field reconstruction module calls the tank wall thermal resistance model to invert the outer wall temperature into a metabolic intensity distribution map characterizing the internal biological activity, and calculates the temperature field uniformity index accordingly; subsequently, the abnormal state judgment module compares the index with a preset threshold range. If an abnormality (non-uniformity) is found, the regional feature recognition logic is triggered, and the specific metabolic hot spots or dead zones are located using hot / cold point thresholds, and the inclusions are mapped and generated. The system employs a metabolic anomaly partitioning mask based on hierarchical information. The flow field control generation module receives this mask and generates flow field correction vectors, including blade angle of attack and micro-injection opening, for different anomaly types. The execution control interface module sends commands to the fermenter actuators for physical adjustment. Simultaneously, the closed-loop feedback verification module re-detects uniformity after a preset response time; if it still fails to meet the standard, it generates secondary correction coefficients for iterative optimization. In extreme cases, the global perturbation submodule forcibly resets the flow field. Through this method, the system achieves full-field perception and precise control of the hydrogen peroxide bacterial fermentation process, significantly improving the yield and stability of microbial protein synthesis.

[0030] Furthermore, the thermal data acquisition module is also used for: Acquire background environmental data collected by an ambient temperature sensor; The tank wall thermal radiation matrix is ​​subjected to differential correction operation with the background environment data, and the corrected tank wall thermal radiation matrix is ​​then determined based on the result of the differential correction operation. The corrected tank wall thermal radiation matrix is ​​transmitted to the metabolic field reconstruction module.

[0031] Furthermore, the corrected tank wall thermal radiation matrix is ​​determined based on the results of the differential correction calculation, specifically including: Extract the corresponding real-time ambient temperature value from the background environment data, and multiply the real-time ambient temperature data with the preset wall reflectivity parameter to construct an environmental reflection noise matrix with the same dimension as the tank wall thermal radiation matrix; Subtract the corresponding coordinate position value in the environmental reflection noise matrix from the value of each pixel in the tank wall thermal radiation matrix to obtain the net radiation difference matrix. The preset distance attenuation compensation coefficient is retrieved to perform linear gain calculation on the net radiation difference matrix, thereby determining the corrected tank wall thermal radiation matrix.

[0032] In one specific embodiment, the thermal data acquisition module first collects background environmental data through multiple temperature sensors distributed throughout the workshop environment. This background environmental data is not a single value, but rather includes time-series temperature data at different heights and orientations around the tank. The unit is Kelvin, used to dynamically capture diurnal temperature differences and interference from air conditioning ventilation on infrared thermometry. Subsequently, the system performs differential correction to eliminate environmental reflection noise. The specific process is as follows: The system extracts the real-time ambient temperature value at the current sampling time t. The system calls a preset wall reflectivity parameter ρ, which is a dimensionless coefficient typically ranging from 0.1 to 0.3. The specific value depends on the polishing degree and surface oxidation condition of the stainless steel material on the outer wall of the fermenter, and is preferably obtained through calibration using a standard blackbody radiation source during the installation and commissioning phase. The system utilizes the formula... Construct an environmental reflection noise matrix with the same dimensions as the tank wall thermal radiation matrix. , where σ is the Stefan-Boltzmann constant, and (i,j) represents the pixel coordinates. This formula is based on Kirchhoff's laws of thermal radiation, treating the ambient temperature as an equivalent blackbody radiation source, and calculating its reflection component on the surface of the tank wall.

[0033] Next, the system will use the original collected tank wall thermal radiation matrix The values ​​of each pixel in the matrix are subtracted one by one from the ambient reflection noise matrix. The net radiation difference matrix is ​​obtained by taking the values ​​at the corresponding coordinate positions in the matrix. The calculation formula is: This step effectively isolates thermal radiation signals not generated by the tank itself, ensuring the accuracy of subsequent inversion.

[0034] Finally, to compensate for the energy attenuation of infrared radiation during air transmission, the system retrieves a preset distance attenuation compensation coefficient. Linear gain calculation is performed on the net radiation difference matrix. This coefficient... It is a dimensionless correction factor derived from the Beer-Lambert law, and the calculation model is as follows: Where κ is the atmospheric extinction coefficient, and d is the Euclidean distance from the sensor lens to each pixel on the tank wall. The system uses the formula... Determine the corrected tank wall thermal radiation matrix The exponential distance compensation used here is to precisely correct the uneven signal attenuation caused by differences in depth of field among different pixels, ensuring that the temperature data in the edge and center regions of the matrix have equal physical confidence. For example, if a pixel is 5 meters away from the sensor and the atmospheric extinction coefficient is 0.005, the compensation coefficient is approximately 1.025, which translates to a signal enhancement of 2.5%. After the above processing, the corrected matrix... It accurately reflects the intrinsic thermal radiation intensity of the outer wall of the fermenter after eliminating environmental interference and transmission loss, providing a high-fidelity data foundation for subsequent metabolic field reconstruction.

[0035] Furthermore, when generating the metabolic intensity distribution map, the metabolic field reconstruction module specifically performs the following steps: Recall pre-stored tank wall thermal resistance model data, which includes the tank wall thickness and the thermal conductivity of the tank wall material; By combining the thermal radiation matrix of the tank wall and the thermal resistance model data of the tank wall, the temperature value corresponding to the liquid film adhering to the tank wall is calculated. Then, a two-dimensional array is constructed based on the temperature value corresponding to the liquid film adhering to the tank wall, and the two-dimensional array is set as a metabolic intensity distribution map.

[0036] Furthermore, the temperature value corresponding to the liquid film adhering to the wall inside the tank is calculated, and then a two-dimensional array is constructed based on the temperature value corresponding to the liquid film adhering to the wall inside the tank. This two-dimensional array is then set as a metabolic intensity distribution map, specifically including: The tank wall thermal resistance model data is analyzed to extract the pre-stored tank wall thickness value and the thermal conductivity of the tank wall material. Divide the tank wall thickness by the thermal conductivity of the tank wall material to calculate the thermal resistance parameter that characterizes the resistance to heat penetration through the tank wall. Obtain a preset heat flux density estimation factor, and multiply the preset heat flux density estimation factor with the thermal resistance parameter to generate a temperature compensation increment; The temperature compensation increment is accumulated point by point into the value of each pixel in the thermal radiation matrix of the tank wall, thereby determining the temperature value corresponding to the liquid film adhering to the tank wall.

[0037] In a specific embodiment, the metabolic field reconstruction module first calls the tank wall thermal resistance model data stored in the system database. This data is not a general parameter, but a specific calibration value for the physical properties of the current fermenter, including the tank wall thickness value δ and the thermal conductivity λ of the tank wall material. Next, a reverse deduction based on Fourier's law of heat conduction is performed to determine the thermal resistance parameters that characterize the resistance to heat penetration through the tank wall. The specific calculation logic is as follows: the system divides the extracted tank wall thickness value δ by the thermal conductivity λ of the tank wall material, i.e., the formula... Calculated The unit is Kelvin per watt per square meter. For example, when the wall thickness is 0.01 m and the thermal conductivity is 16 W / (m²), the thermal conductivity is 16 W / (m²). At K), the thermal resistance parameter Approximately 0.000625 (m2) K) / W. The necessity of this step lies in the fact that the fermenter wall, as a thermal resistance layer, causes the metabolic heat generated inside to decrease in temperature when conducted to the surface, and this degree of decrease must be quantified by thermal resistance.

[0038] Subsequently, the system obtains the preset heat flux density estimation factor. This factor represents the radial heat loss per unit area through the tank wall, expressed in watts per square meter (W / m²). This parameter... The value is not arbitrarily set, but rather is a baseline value pre-calculated based on the balance between the heat generation rate per unit volume of hydroxide bacteria during the exponential growth phase and the heat exchange efficiency of the cooling jacket, and is set within an adjustable range of 1000 to 3000 W / m². The system uses this heat flux density estimation factor... The thermal resistance parameters obtained from the above calculations Perform multiplication to generate temperature compensation increments. The calculation formula is: ; Finally, the system will calculate the temperature compensation increment. As a scalar constant, it is accumulated point by point to the corrected tank wall thermal radiation matrix output from the previous step. In the values ​​of each pixel, the formula is executed. The addition operation is performed here because fermentation is an exothermic reaction, with heat flowing from the inside out; therefore, the temperature of the liquid film adhering to the wall must be higher than the surface temperature of the outer wall. Through this calculation, the system determines the temperature matrix corresponding to the liquid film adhering to the wall inside the tank. The two-dimensional temperature matrix was directly defined as a metabolic intensity distribution map.

[0039] Furthermore, the specific formula for calculating the temperature field uniformity index is as follows: The ratio of the temperature standard deviation to the temperature average value of all valid pixels in the metabolic intensity distribution map is calculated, and the ratio is used as the temperature field uniformity index. The temperature field uniformity index is used to quantify the spatial consistency of the metabolic rate of hydroxide bacteria inside the fermenter at the current moment.

[0040] In one specific embodiment, after generating the metabolic intensity distribution map, the metabolic field reconstruction module does not directly perform discrete threshold alarms based on the original image. Instead, it generates a dimensionless index, namely the temperature field uniformity index, through statistical calculations to macroscopically characterize the mixing state of the entire tank. This calculation process first performs effective region screening and statistical extraction on the metabolic intensity distribution map. The specific logic is as follows: The system iterates through the data T(x,y) of each pixel in the metabolic intensity distribution map, and removes pixels belonging to the gas phase region of the fermenter headspace based on the preset liquid level height parameter, retaining only the effective pixel set of the liquid phase region. Let the total number of valid pixels be N. Next, the system calculates the arithmetic mean temperature of the set of valid pixels. The calculation formula is: ,in, This is the temperature value of the i-th valid pixel. The average temperature... Physically, it represents the baseline of overall thermal balance within the fermenter at the current moment.

[0041] Subsequently, the system is based on average temperature Calculate the standard deviation of temperature The calculation formula is: The sample standard deviation formula is used here instead of the population standard deviation in order to make an unbiased estimate of the volatility of uncovered areas in a limited number of infrared sampling points, thereby capturing small local temperature fluctuations more sensitively. The physical dimension of is Celsius, which quantifies the dispersion of temperature distribution inside the fermentation broth.

[0042] Finally, the system performs a normalization operation to generate the temperature field uniformity index UI, calculated using the following formula: By dividing the standard deviation by the mean, the system eliminates the influence of the absolute temperature magnitude on the evaluation index, making UI a pure, dimensionless percentage value.

[0043] use Instead of using temperature as a simple evaluation criterion, the UI index is chosen because a simple extreme difference is easily affected by individual noisy pixels (such as sensor dead pixels), leading to false alarms. The UI index, based on full pixel statistics, has an integral smoothing effect, resisting random noise and accurately reflecting the disorder of the "heat-fluid" coupling field within the entire fermenter. The closer the UI value is to 0, the more... A temperature approaching 0, meaning that the temperature of all pixels is highly similar, indicates that the fluid mixing inside the fermenter is extremely uniform, and the metabolic rate of the hydroxide bacteria remains highly consistent in space. Conversely, when the UI value increases, it mathematically proves that there must be a significant temperature gradient distribution inside the tank, which physically corresponds to the formation of metabolic hotspots or dead zones, thus providing a solid mathematical basis for subsequent triggering of zonal regulation.

[0044] Furthermore, when the abnormal state determination module executes the region feature recognition logic to generate a metabolic abnormality partition mask, it specifically performs the following steps: Traverse all pixels in the metabolic intensity distribution map, filter out the set of pixels with temperature values ​​higher than a preset hotspot threshold, and mark the connected region where the set of pixels is located as a metabolic hotspot region. Filter out the set of pixels in the metabolic intensity distribution map whose temperature values ​​are lower than a preset cold point threshold, and mark the connected region where the set of pixels is located as a metabolic dead zone region; The location information of the metabolic hotspots and metabolic dead zones in the metabolic intensity distribution map is extracted, and then the location information is mapped to the physical hierarchical coordinate system of the fermenter to generate a metabolic anomaly partition mask containing physical hierarchical location information.

[0045] In one specific embodiment, after receiving the metabolic intensity distribution map generated by the preceding steps, the abnormal state determination module does not simply perform pixel-level comparison, but instead executes region feature recognition logic based on connected component analysis to generate a metabolic abnormality partition mask with spatial semantics. This process first defines a preset hotspot threshold. and preset cold spot threshold These two thresholds are not fixed constants, but are based on the optimal growth temperature range of hydroxide granules. The calculation formula is based on the dynamically set allowable deviation coefficient δ. and δ is set to 0.05 to 0.10 (i.e., a deviation of 5% to 10%) to ensure that the alarm mechanism conforms to the biological limits of metabolic tolerance.

[0046] The system performs binarization segmentation by traversing the data d(x,y) of each pixel in the metabolic intensity distribution map: if d(x,y) > Mark this point as a "hotspot candidate pixel" (add +1); if d(x,y) < The point is marked as a "dead zone candidate pixel" (assigned a value of -1); otherwise, it is marked as "normal background" (assigned a value of 0). To eliminate discrete isolated point interference caused by sensor noise or instantaneous bubble bursts in the infrared image, the system then applies an 8-neighborhood connected component labeling algorithm to aggregate adjacent candidate pixels of the same type into independent connected regions. And calculate the pixel area of ​​each connected region. At this point, the system introduces a preset effective area threshold. Only retain The region is used as an effective metabolic hotspot or metabolic dead zone. This step effectively prevents false alarms through spatial filtering, ensuring that only metabolically abnormal clumps of a certain size are identified.

[0047] Finally, the system performs physical space mapping to generate a metabolic anomaly partitioning mask. The system extracts each valid connected region. Geometric centroid coordinates The inverse operation of the perspective transformation matrix is ​​used to transform the ordinate of the image coordinate system. Physical height coordinates mapped to the fermenter The specific mapping logic is based on a linear hierarchical formula: ,in, For physical hierarchy indexes (e.g., 1 represents the bottom layer, 2 represents the middle layer, 3 represents the top layer). The total pixel height covered by the image. The number of independent agitator blade layers configured for the fermenter (usually 3-4 layers). This indicates a floor operation. Using this formula, the system precisely pinpoints the temperature anomaly region on the two-dimensional image to a specific mechanical control level, ultimately generating a data structure containing a level index and anomaly type (hot / cold) key-value pairs—a metabolic anomaly partition mask. This provides clear target location information for the subsequent flow field control generation module, achieving spatial coordinate unification from "visual perception" to "mechanical execution."

[0048] Furthermore, the specific logic for generating the flow field correction vector is as follows: The physical level at which the metabolic hotspot region is located is identified, and commands to increase the blade angle of attack and decrease the micro-gas injection opening are generated. The command to increase the blade angle of attack is used to drive the variable pitch agitator at this physical level to increase the angle of attack, thereby enhancing the local axial fluid exchange and heat dissipation efficiency. The command to decrease the micro-gas injection opening is used to prevent metabolic bursts caused by excessive local substrate. The physical level of the metabolic dead zone is located, and an increased micro-gas injection opening command and a fine-tuned blade angle of attack command are generated. The increased micro-gas injection opening command is used to drive the side wall micro-gas injection valve of the physical level to open, so as to increase the local hydrogen and oxygen substrate concentration. The fine-tuned blade angle of attack command is used to cooperate with gas-liquid mixing to activate the metabolic activity of bacteria in the region.

[0049] Furthermore, it also includes a closed-loop feedback verification module, used for: After the execution control interface module sends the flow field correction vector and a preset response time has elapsed, the thermal data acquisition module is triggered to acquire a new tank wall thermal radiation matrix again. The metabolic field reconstruction module is triggered to calculate a new temperature field uniformity index; Determine whether the new temperature field uniformity index falls within the preset uniformity threshold range. If it does not, generate a secondary correction coefficient and feed it back to the flow field control generation module.

[0050] In one specific embodiment, the closed-loop feedback verification module does not immediately determine the effect of flow field control, but instead introduces timing control logic based on the hysteresis characteristics of fluid dynamics. This module first starts a countdown timer, the duration of which is set to a preset response time. This time is not a fixed value, but depends on the characteristic mixing time of the fermenter. (That is, the time required for the tracer to achieve 95% homogeneity in the container) is dynamically calculated, and the calculation formula is: Where k is the thermal inertia coefficient, usually taken as 2 to 3, to ensure that the fluid in the tank completes at least 2-3 complete circulation cycles, so that the thermal convection effect caused by the change of the angle of attack of the stirring blade is fully transferred to the surface of the tank wall, eliminating the misjudgment of "pseudo-steady state" caused by the lag of fluid heat transfer.

[0051] After the countdown ends, the system triggers the thermal data acquisition module and the metabolic field reconstruction module to execute the data acquisition and processing flow again, and outputs the verification temperature field uniformity index at the current moment. Subsequently, the system judged Does it fall within the preset uniformity threshold range? (For example, [0, 0.05]). If Still higher This indicates that the previous flow field correction was insufficient. At this point, the system uses an error proportional feedback algorithm to generate dimensionless secondary correction coefficients. The calculation formula is: ,in, Let α be the target value for ideal uniformity (usually the midpoint of the range, such as 0.025), and α be the feedback gain coefficient (ranging from 0.3 to 0.5), used to adjust the system's sensitivity to prevent control oscillations. This formula is a variation of the classic proportional control algorithm, and its physical meaning lies in quantifying the deviation of the current residual non-uniformity from the target value, and generating a gain factor greater than 1 accordingly.

[0052] Finally, the system will use this secondary correction coefficient. The feedback is sent to the flow field control generation module, which takes the blade angle of attack command value output in the previous round. and Multiply to obtain the corrected instructions. This allows the control intensity to be increased non-linearly and proportionally without changing the control direction (such as continuing to increase the angle of attack) until the UI index converges to the normal range in the next round of closed-loop verification. This process ensures that the system can adaptively respond to the changes in rheological properties caused by the increase in cell concentration, avoiding the common problems of "inadequate control" or "over-control" in open-loop control.

[0053] Furthermore, the flow field control generation module also includes a global perturbation submodule, used for: When the temperature field uniformity index exceeds a preset emergency threshold, a global blade angle of attack command is generated.

[0054] Furthermore, it also includes a historical data optimization module, used for: The metabolic intensity distribution map, the flow field correction vector, and the corresponding protein yield data are stored during the fermentation process. A flow field-yield correlation model is established, and the upper and lower limits of the preset uniformity threshold range are dynamically adjusted based on the flow field state corresponding to the historical best yield, so that the abnormal state determination module can call it.

[0055] In one specific embodiment, the historical data optimization module does not simply store data in bulk, but rather constructs a dynamic optimization closed loop based on time series. This module first synchronizes and aligns multi-source heterogeneous data during the fermentation process, reducing the dimensionality of the high-dimensional metabolic intensity distribution map to extract the eigenvalue temperature field uniformity index UI(t), and then compares it with the flow field correction vector at the same time. and instantaneous protein yield Perform timestamp association to construct a triple sequence Among them, instantaneous protein yield It is not measured offline, but rather by monitoring the hydrogen consumption rate using an online exhaust gas analyzer in conjunction with a preset conversion coefficient. The calculation formula is as follows: In the formula, Ventilation volume, The concentration of hydrogen gas entering and exiting the gas. The fermentation broth is expressed by volume. This is the preset conversion coefficient.

[0056] Next, this module establishes a flow field-yield correlation model, aiming to find the optimal flow field uniformity range that maximizes yield. The system employs a yield-weighted moment estimation algorithm to calculate the expected value of optimal uniformity corresponding to historically high-yield periods. and tolerable bandwidth The calculation formula is: The physical meaning of this formula is to find the "center of gravity of productivity," that is, the UI center value corresponding to the highest output; and This formula calculates the weighted standard deviation of the UI value under high-yield conditions. These two formulas are variations of the first-order raw moment and the second-order central moment in probability and statistics. Their advantage lies in their ability to automatically filter out UI data interference during low-yield periods, thereby accurately identifying the flow field characteristics under high-yield conditions.

[0057] Finally, the system calculates... and The preset uniformity threshold range for use by the abnormal state determination module is dynamically adjusted using the exponential moving average update law. The updated formula is as follows: ,as well as Where β is the learning rate factor, ranging from 0.05 to 0.1 (to avoid drastic threshold jumps due to fluctuations in a single batch), and k is the confidence interval coefficient, typically taking a value of 2 or 3.

[0058] For example, assuming that in historical batches, hydrogen utilization was highest when the UI remained around 0.03, the calculation yielded... =0.03, =0.005. If k=2 is set, the calculated optimal physical interval is [0.02, 0.04]. If the original interval is set to [0.01, 0.05], by introducing a learning rate β=0.1, the system will slowly converge the interval boundary towards [0.02, 0.04]. This mechanism ensures that as production batches accumulate, the monitoring system's understanding of "what constitutes a good flow field" will continuously evolve, shifting from an experience-based static threshold to a data-driven dynamic optimal interval, thereby achieving self-iterative optimization of the fermentation process.

[0059] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention, they should all fall within the protection scope of the present invention.

[0060] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0061] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A fermentation monitoring system utilizing hydroxide bacteria for co-production of microbial protein, characterized in that, include: The thermal data acquisition module is used to acquire the tank wall thermal radiation matrix at a preset sampling frequency by using an array of infrared thermal imaging sensors deployed on the outer wall of the fermenter. The tank wall thermal radiation matrix contains the real-time temperature values ​​of each coordinate point on the surface of the tank. The metabolic field reconstruction module is used to receive the tank wall thermal radiation matrix, use a heat conduction inversion algorithm to map the tank wall thermal radiation matrix into a metabolic intensity distribution map, and calculate and generate a temperature field uniformity index based on the numerical dispersion of all pixels in the metabolic intensity distribution map. An abnormal state determination module is used to compare the temperature field uniformity index with a preset uniformity threshold range; if the temperature field uniformity index exceeds the preset uniformity threshold range, the regional feature recognition logic of the metabolic intensity distribution map is triggered, and a metabolic abnormality partition mask is generated based on the pixel temperature value distribution. The metabolic abnormality partition mask marks the spatial coordinates of the fermenter where there is an abnormal temperature gradient. The flow field regulation generation module is used to receive the metabolic abnormality partition mask and generate a flow field correction vector according to the abnormality type corresponding to the spatial region coordinates. The execution control interface module is used to send the flow field correction vector to the actuator of the fermenter. The flow field correction vector includes the blade angle of attack command and the micro-air injection opening command.

2. The fermentation monitoring system for co-production of microbial protein using hydroxide bacteria according to claim 1, characterized in that, The thermal data acquisition module is also used for: Acquire background environmental data collected by an ambient temperature sensor; The tank wall thermal radiation matrix is ​​subjected to differential correction operation with the background environment data, and the corrected tank wall thermal radiation matrix is ​​then determined based on the result of the differential correction operation. The corrected tank wall thermal radiation matrix is ​​transmitted to the metabolic field reconstruction module.

3. The fermentation monitoring system for co-production of microbial protein using hydroxide bacteria according to claim 2, characterized in that, The corrected tank wall thermal radiation matrix is ​​determined based on the results of the differential correction calculation, specifically including: Extract the corresponding real-time ambient temperature value from the background environment data, and multiply the real-time ambient temperature data with the preset wall reflectivity parameter to construct an environmental reflection noise matrix with the same dimension as the tank wall thermal radiation matrix; Subtract the corresponding coordinate position value in the environmental reflection noise matrix from the value of each pixel in the tank wall thermal radiation matrix to obtain the net radiation difference matrix. The preset distance attenuation compensation coefficient is retrieved to perform linear gain calculation on the net radiation difference matrix, thereby determining the corrected tank wall thermal radiation matrix.

4. The fermentation monitoring system for co-production of microbial protein using hydroxide bacteria according to claim 1, characterized in that, When generating a metabolic intensity distribution map, the metabolic field reconstruction module performs the following steps: Recall pre-stored tank wall thermal resistance model data, which includes the tank wall thickness and the thermal conductivity of the tank wall material; By combining the thermal radiation matrix of the tank wall and the thermal resistance model data of the tank wall, the temperature value corresponding to the liquid film adhering to the tank wall is calculated. Then, a two-dimensional array is constructed based on the temperature value corresponding to the liquid film adhering to the tank wall, and the two-dimensional array is set as a metabolic intensity distribution map.

5. A fermentation monitoring system for co-producing microbial protein using hydroxide bacteria according to claim 4, characterized in that, Calculate the temperature value corresponding to the liquid film adhering to the wall inside the tank, then construct a two-dimensional array based on the temperature value of the liquid film adhering to the wall inside the tank, and set the two-dimensional array as a metabolic intensity distribution map, specifically including: The tank wall thermal resistance model data is analyzed to extract the pre-stored tank wall thickness value and the thermal conductivity of the tank wall material. Divide the tank wall thickness by the thermal conductivity of the tank wall material to calculate the thermal resistance parameter that characterizes the resistance to heat penetration through the tank wall. Obtain a preset heat flux density estimation factor, and multiply the preset heat flux density estimation factor with the thermal resistance parameter to generate a temperature compensation increment; The temperature compensation increment is accumulated point by point into the value of each pixel in the thermal radiation matrix of the tank wall, thereby determining the temperature value corresponding to the liquid film adhering to the tank wall.

6. A fermentation monitoring system for co-producing microbial protein using hydroxide bacteria according to claim 1, characterized in that, The specific formula for calculating the temperature field uniformity index is as follows: Calculate the ratio of the temperature standard deviation to the temperature average value of all valid pixels in the metabolic intensity distribution map, and use the ratio as the temperature field uniformity index.

7. A fermentation monitoring system for co-production of microbial protein using hydroxide bacteria according to claim 1, characterized in that, When the abnormal state determination module executes the region feature recognition logic to generate a metabolic abnormality partition mask, it specifically performs the following steps: Traverse all pixels in the metabolic intensity distribution map, filter out the set of pixels with temperature values ​​higher than a preset hotspot threshold, and mark the connected region where the set of pixels is located as a metabolic hotspot region. Filter out the set of pixels in the metabolic intensity distribution map whose temperature values ​​are lower than a preset cold point threshold, and mark the connected region where the set of pixels is located as a metabolic dead zone region; The location information of the metabolic hotspots and metabolic dead zones in the metabolic intensity distribution map is extracted, and then the location information is mapped to the physical hierarchical coordinate system of the fermenter to generate a metabolic anomaly partition mask containing physical hierarchical location information.

8. A fermentation monitoring system for co-production of microbial protein using hydroxide bacteria according to claim 7, characterized in that, The specific logic for generating the flow field correction vector is as follows: The physical level at which the metabolic hotspot region is located is determined, and commands to increase the blade angle of attack and decrease the micro-injection opening are generated. The physical level of the metabolic dead zone is located, and commands to increase the micro-injection air opening and fine-tune the blade angle of attack are generated.

9. A fermentation monitoring system for co-production of microbial protein using hydroxide bacteria according to claim 1, characterized in that, It also includes a closed-loop feedback verification module, used for: After the execution control interface module sends the flow field correction vector and a preset response time has elapsed, the thermal data acquisition module is triggered to acquire a new tank wall thermal radiation matrix again. The metabolic field reconstruction module is triggered to calculate a new temperature field uniformity index; Determine whether the new temperature field uniformity index falls within the preset uniformity threshold range. If it does not, generate a secondary correction coefficient and feed it back to the flow field control generation module.

10. A fermentation monitoring system for co-production of microbial protein using hydroxide bacteria according to claim 1, characterized in that, The flow field control generation module also includes a global perturbation submodule, used for: When the temperature field uniformity index exceeds a preset emergency threshold, a global blade angle of attack command is generated.

Images