A dynamic feedback control method and system in ganoderma lucidum pill fermentation process

Abstract

The application provides a dynamic feedback control method and system in the fermentation process of ganoderma pill, wherein the method comprises: obtaining the concentration dynamic change information of ganoderma polysaccharide, ganoderma acid and dissolved oxygen in the fermentation broth in the fermentation process of ganoderma pill; based on the concentration dynamic change information, a time-space spectrum reflecting the distribution state of metabolites is generated through a multi-modal data interaction fusion method; the time-space spectrum is input into a pre-trained concentration correlation change prediction model, and concentration correlation change data between ganoderma polysaccharide and ganoderma acid in a future time window is output; based on the concentration correlation change data, the adjustment amount of the fermentation control parameters in the time sequence dimension and the spatial dimension is iteratively corrected, a feedback control signal is generated, and the fermentation process of the ganoderma pill is dynamically adjusted, wherein the fermentation control parameters include the feeding rate and the aeration amount. The application improves the output stability and proportion controllability of effective components in the fermentation process of ganoderma pill.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biopharmaceuticals, and in particular to a dynamic feedback control method and system in the fermentation process of ganoderma pill. BACKGROUND

[0002] As a traditional Chinese medicine preparation, the content and proportion of ganoderma polysaccharide and ganoderma acid, the core active ingredients of ganoderma pill, directly affect the product quality. In industrialized fermentation production, the concentration changes of the two components need to be monitored in real time, and the feed rate and aeration rate need to be accurately adjusted according to their dynamic correlation to ensure efficient synthesis and stable accumulation of target metabolites.

[0003] Currently, there is a control method that uses offline detection combined with static model prediction, which is specifically as follows: periodically sampling and detecting the concentrations of polysaccharide and acid in the fermentation broth, generating a fixed-length control instruction through a pre-set concentration-parameter mapping table, and adjusting the feed and aeration operations.

[0004] The above method has a control lag due to the dependence on offline detection; meanwhile, the static model is prone to mismatch between feed and aeration adjustment when the dissolved oxygen concentration suddenly changes, resulting in uncontrollable proportion of target components. SUMMARY

[0005] The present application provides a dynamic feedback control method and system in the fermentation process of ganoderma pill to solve the problems of poor stability and weak controllability of effective components in the fermentation process of ganoderma pill in the prior art.

[0006] In a first aspect, the present application provides a dynamic feedback control method in the fermentation process of ganoderma pill, comprising:

[0007] In the fermentation process of ganoderma pill, the dynamic change information of the concentrations of ganoderma polysaccharide, ganoderma acid and dissolved oxygen in the fermentation broth is obtained;

[0008] Based on the concentration dynamic change information, a spatiotemporal atlas reflecting the distribution state of metabolites is generated through a multi-modal data interaction fusion method;

[0009] The spatiotemporal atlas is input into a pre-trained concentration correlation change prediction model, and concentration correlation change data between the ganoderma polysaccharide and the ganoderma acid in a future time window is output;

[0010] Based on the concentration correlation change data, the adjustment amount of fermentation control parameters in the time sequence dimension and the spatial dimension is iteratively corrected, a feedback control signal is generated to dynamically adjust the fermentation process of ganoderma pill, and the fermentation control parameters include the feed rate and the aeration rate.

[0011] Optionally, based on the concentration correlation change data, the adjustment amount of fermentation control parameters in the time sequence dimension and the spatial dimension is iteratively corrected, and a feedback control signal is generated, comprising:

[0012] Based on the concentration correlation change data and the dynamic change information of dissolved oxygen concentration, and combined with the preset fuzzy compensation rules, a control command sequence is generated.

[0013] Based on the control command sequence and the concentration-related change data, an adaptive matching relationship is constructed between the prediction deviation and the compensation threshold parameter;

[0014] Based on the adaptive matching relationship, a feedback control signal is generated.

[0015] Optionally, generating a feedback control signal based on the adaptive matching relationship includes:

[0016] Based on the degree of correspondence between the predicted deviation and the compensation threshold parameter in the adaptive matching relationship, the adjustment direction of the fermentation control parameter is determined. The adjustment direction includes the incremental change trend in the time dimension and the parameter distribution offset pattern in the spatial dimension.

[0017] Feedback control signals are generated based on the incremental change trend in the time dimension and the parameter distribution offset pattern in the spatial dimension.

[0018] Optionally, generating a feedback control signal based on the incremental change trend in the time dimension and the parameter distribution offset pattern in the spatial dimension includes:

[0019] Based on the incremental change trend in the time dimension, calculate the single-step adjustment factor of the control parameter;

[0020] By imposing an upper limit constraint on the adjustment range of the single-step adjustment factor, an amplitude-limited adjustment factor is obtained;

[0021] A smoothness threshold check is performed on the amplitude-limited adjustment factor to generate a single-step adjustment factor sequence;

[0022] Generate a spatial coefficient matrix based on the parameter distribution offset pattern in the spatial dimension;

[0023] Amplitude upper limit pruning is performed on the matrix elements of the spatial coefficient matrix;

[0024] The single-step adjustment factor sequence is combined with the clipped spatial coefficient matrix according to a preset spatiotemporal coupling rule;

[0025] The physical feasibility of the combined fermentation environment is verified, and if the verification is successful, a feedback control signal corresponding to the combined result is output.

[0026] Optionally, the step of generating a control command sequence based on the concentration correlation change data and the dynamic change information of dissolved oxygen concentration, combined with a preset fuzzy compensation rule, includes:

[0027] Based on the concentration ratio of Ganoderma lucidum polysaccharide to Ganoderma lucidum acid in the concentration correlation change data, and the dynamic change information of dissolved oxygen concentration, a compensation intensity value is generated.

[0028] Based on the product relationship between the compensation intensity value and the base value of the control parameter in the preset fuzzy compensation rule, the feeding rate is generated;

[0029] The feeding rate is extended into discrete command points according to the time series;

[0030] Based on the rate of change of compensation intensity between adjacent discrete command points, transition command points are inserted in the time sequence to generate a feeding rate command sequence.

[0031] Based on the distribution differences of the dynamic changes in dissolved oxygen concentration within the fermentation vessel, the spatial priority of the fermentation zones is determined, and an independent aeration correction coefficient is assigned to each fermentation zone.

[0032] The ventilation correction coefficient is superimposed onto the ventilation base value of the corresponding fermentation region according to the spatial priority to generate a ventilation instruction sequence for different fermentation regions.

[0033] The feeding rate command sequence is combined with the ventilation rate command sequence to generate a control command sequence.

[0034] Optionally, the step of inputting the spatiotemporal map into a pre-trained concentration correlation change prediction model and outputting the concentration correlation change data between the Ganoderma lucidum polysaccharide and the Ganoderma lucidum acid within a future time window includes:

[0035] The spatiotemporal map is input into a pre-trained concentration correlation change prediction model. Through a multi-step recursive prediction algorithm, the predicted concentration ratio and spatial distribution offset of Ganoderma lucidum polysaccharide and Ganoderma lucidum acid at each future time point are output sequentially.

[0036] The predicted concentration ratio and spatial distribution offset of Ganoderma lucidum polysaccharides and ganoderic acids at all future time points are integrated according to the time window length to generate concentration correlation change data between Ganoderma lucidum polysaccharides and ganoderic acids within the future time window.

[0037] Optionally, the step of generating a spatiotemporal map reflecting the distribution status of metabolites based on the concentration dynamic change information through multimodal data interaction and fusion includes:

[0038] The dynamic changes in the concentrations of Ganoderma lucidum polysaccharides, ganoderic acids, and dissolved oxygen were divided into time series segments according to the time dimension;

[0039] Metabolic association features were labeled for each time series segment;

[0040] The fermentation container used for fermenting Ganoderma lucidum pills was divided into multiple independent fermentation zones. The spatial distribution differences of Ganoderma lucidum polysaccharides, Ganoderma lucidum acids and dissolved oxygen in each fermentation zone were extracted to generate spatial distribution feature maps of each fermentation zone.

[0041] The metabolic association feature labeling results are cross-linked with the spatial distribution feature maps at the corresponding time points to generate a spatiotemporal map.

[0042] Secondly, this application provides a dynamic feedback control system for the fermentation process of Ganoderma lucidum pills, comprising:

[0043] The acquisition module is used to acquire dynamic change information on the concentrations of Ganoderma lucidum polysaccharides, Ganoderma lucidum acids and dissolved oxygen in the fermentation broth during the fermentation process of Ganoderma lucidum pills.

[0044] The generation module is used to generate a spatiotemporal map reflecting the distribution status of metabolites based on the concentration dynamic change information through multimodal data interaction and fusion.

[0045] The input module is used to input the spatiotemporal map into a pre-trained concentration correlation change prediction model and output the concentration correlation change data between Ganoderma lucidum polysaccharide and Ganoderma lucidum acid within a future time window;

[0046] The correction module is used to iteratively correct the adjustment amount of the fermentation control parameters in the temporal and spatial dimensions based on the concentration-related change data, and generate feedback control signals to dynamically adjust the fermentation process of Ganoderma lucidum pills. The fermentation control parameters include feeding rate and aeration rate.

[0047] Thirdly, this application provides a computing device, including a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute a dynamic feedback control method for the fermentation process of Ganoderma lucidum pills as described in any of the first aspects.

[0048] Fourthly, this application provides a computer storage medium storing computer program instructions thereon, wherein the computer program instructions, when executed by a processor, implement a dynamic feedback control method for the fermentation process of Ganoderma lucidum pills as described in any one of the first aspects.

[0049] This application provides a dynamic feedback control method for the fermentation process of Ganoderma lucidum pills. The method includes: acquiring dynamic change information on the concentrations of Ganoderma lucidum polysaccharides, ganoderic acids, and dissolved oxygen in the fermentation broth during the fermentation process; generating a spatiotemporal map reflecting the distribution state of metabolites based on the dynamic change information of the concentrations through multimodal data interaction and fusion; inputting the spatiotemporal map into a pre-trained concentration correlation change prediction model and outputting concentration correlation change data between Ganoderma lucidum polysaccharides and ganoderic acids within a future time window; iteratively correcting the adjustment amount of fermentation control parameters in the temporal and spatial dimensions based on the concentration correlation change data, generating a feedback control signal to dynamically regulate the fermentation process of Ganoderma lucidum pills, wherein the fermentation control parameters include feeding rate and aeration rate.

[0050] The technical solution provided in this application has the following beneficial effects:

[0051] This application enables real-time monitoring of key parameters (polysaccharides, acids, and dissolved oxygen) during the fermentation process, providing a data foundation for dynamic regulation and improving the stability of active ingredient production during Ganoderma lucidum pill fermentation. Through multimodal data fusion, the distribution of metabolites in time and space is presented intuitively, overcoming the limitations of single-dimensional analysis. Based on model prediction of the synergistic change trends of polysaccharides and acids in future periods, feedforward control of the fermentation process is achieved. Through iterative correction of parameters in both spatiotemporal dimensions, precise synergistic regulation of feeding rate and aeration rate is realized.

[0052] Furthermore, this application also generates preliminary control commands by fusing concentration-related change data with dissolved oxygen dynamic information and combining fuzzy compensation rules; then, it establishes an adaptive matching mechanism between the predicted deviation and the compensation threshold, and finally outputs a dynamically optimized feedback control signal. This process realizes the upgrade of control parameters from static rules to dynamic adaptation.

[0053] Furthermore, this solution effectively solves the regulation mismatch problem caused by response lag in traditional control methods. Through an adaptive compensation mechanism, it improves the regulation accuracy and stability of the proportion of effective components during the fermentation of Ganoderma lucidum pills, making it particularly suitable for complex working conditions such as dissolved oxygen fluctuations.

[0054] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 A flowchart illustrating a dynamic feedback control method during the fermentation process of Ganoderma lucidum pills, provided as an embodiment of this application;

[0057] Figure 2 A schematic diagram of the structure of a dynamic feedback control system in the fermentation process of Ganoderma lucidum pills provided in this application embodiment;

[0058] Figure 3 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation

[0059] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0060] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0061] Researchers have found that existing methods for controlling the fermentation process of Ganoderma lucidum pills struggle to capture the dynamic correlation between polysaccharides and ganoderic acids in real time, and cannot effectively predict the spatiotemporal impact of dissolved oxygen changes on metabolic processes, leading to lag and spatial inhomogeneity in feeding and aeration regulation. Based on this, this application provides a dynamic feedback control method for the fermentation process of Ganoderma lucidum pills. This method constructs a spatiotemporal map through multimodal data fusion, combines a predictive model to proactively output concentration correlation trends, and iteratively optimizes control parameters based on spatiotemporal dual-dimensional optimization, achieving precise and coordinated regulation of feeding rate and aeration. The technical solution of this application is applicable to medicinal fungal fermentation scenarios requiring precise control of the proportions and spatial distribution of multiple metabolites.

[0062] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0063] Figure 1 A flowchart of a dynamic feedback control method for the fermentation process of Ganoderma lucidum pills provided in this application embodiment is shown below. Figure 1 As shown, the method includes:

[0064] Step 101: During the fermentation of Ganoderma lucidum pills, obtain information on the dynamic changes in the concentrations of Ganoderma lucidum polysaccharides, Ganoderma lucidum acids, and dissolved oxygen in the fermentation broth.

[0065] In this step, the dynamic change information of Ganoderma lucidum polysaccharide concentration represents the content of polysaccharides in the fermentation broth (unit: g / L), reflecting the metabolic activity of the microorganisms. The dynamic change information of Ganoderma lucidum acid concentration represents the content of triterpenoids in the fermentation broth (unit: mg / L), characterizing the degree of product accumulation. The dynamic change information of dissolved oxygen concentration represents the oxygen content in the fermentation broth (unit: % saturation), affecting the respiratory metabolism of the microorganisms.

[0066] In this embodiment, fermentation broth parameters are collected in real time by an online sensor array (including a near-infrared spectral probe, dissolved oxygen electrode, etc.) installed in the fermenter. The polysaccharide and acid concentrations are collected every 5 minutes using spectral analysis, while dissolved oxygen is continuously monitored by electrodes. The collected data is processed by a signal conversion module and then transmitted to the central control system to form a three-parameter time-series dataset containing timestamps.

[0067] For example, when culturing Ganoderma lucidum mycelium in a 500-liter stainless steel fermenter, three sets of online monitoring devices were installed: a near-infrared sensor to monitor the polysaccharide concentration in real time (the detection wavelength was selected at the characteristic absorption peak of Ganoderma lucidum polysaccharides at 2100 nm, and the current value was measured to be 15.3 g / L); a high-performance liquid chromatography probe to detect the concentration of ganoderic acid (by comparing the retention time of the standard, the current value was measured to be 1.2 mg / mL); and a dissolved oxygen electrode (using a polarized probe, the dissolved oxygen level in the middle of the tank was measured to be 45% saturation). The system automatically collected data every 10 minutes and stored the monitored values ​​of the three parameters along with the collection timestamps into the database. For example, at the 36th hour of cultivation, the system recorded a set of typical data: polysaccharide 15.3 g / L, acid 1.2 mg / mL, and dissolved oxygen 45%.

[0068] Step 102: Based on the concentration dynamic change information, a spatiotemporal map reflecting the distribution status of metabolites is generated through multimodal data interaction and fusion.

[0069] In this step, multimodal data includes monitoring data from different sources (spectral, electrode, etc.) and types (concentration values, rate of change, etc.). Metabolite distribution refers to the spatiotemporal distribution characteristics of three key substances—Ganoderma lucidum polysaccharides, ganoderic acids, and dissolved oxygen—in the fermentation system. The spatiotemporal map represents a visualization matrix of metabolite distribution along the time axis and spatial coordinates.

[0070] In this embodiment, time-series concentration data is bound to the three-dimensional coordinates of the fermenter, and a continuous distribution field is generated using a spatiotemporal interpolation algorithm. Tensor decomposition technology is employed to fuse multi-parameter data, constructing a structure containing a time layer (one slice per hour) and a spatial layer (1 cm). 3 The three-dimensional maps of voxels and parameter layers (polysaccharides / acids / dissolved oxygen) are obtained; finally, the dimensions are unified through normalization, and a standardized spatiotemporal distribution matrix is ​​output.

[0071] For example, based on data collected over 36 hours, the system first calculates the rate of change in polysaccharide and acid concentrations (by comparing data from the previous six time points, it is found that polysaccharide increases by 0.5 grams per hour and acid increases by 0.08 milligrams per hour). Then, the fermenter is divided into three layers (upper, middle, and lower), with five monitoring points deployed in each layer. A three-dimensional concentration distribution map is generated using a spatial interpolation algorithm. Finally, the time-series data and spatial distribution data are fused to generate a dynamic spatiotemporal map. This map shows that the lower layer of the tank has a higher polysaccharide concentration (up to 16.1 g / L) but lower dissolved oxygen (only 38%), while the upper layer has active acid synthesis (1.25 mg / mL) and sufficient dissolved oxygen (52%).

[0072] Step 103: Input the spatiotemporal map into the pre-trained concentration correlation change prediction model and output the concentration correlation change data between Ganoderma lucidum polysaccharide and Ganoderma lucidum acid within the future time window.

[0073] In this step, the predictive model represents a deep learning model based on a long short-term memory-transformer hybrid architecture. Concentration-related changes indicate the trend of the ratio of polysaccharide to acid concentrations over future periods.

[0074] In this embodiment, the spatiotemporal map is split into time series segments and spatial feature maps, which are input into the model's time encoder and spatial encoder, respectively; cross-regional metabolic associations are captured through an attention mechanism, and spatiotemporal features are fused in a fully connected layer; the predicted polysaccharide / acid concentration ratio and confidence intervals at 15-minute intervals every 3 hours are output to form a change curve.

[0075] For example, after inputting the generated spatiotemporal map into the prediction model, the model analysis found that, based on the current trend, the accumulation rate of polysaccharides in the lower layer will slow down in the next 3 hours (predicted value decreases from 0.5 g / h to 0.3 g / h), while the synthesis rate of acids in the upper layer will accelerate (increase from 0.08 mg / h to 0.12 mg / h). The model specifically marked the lower right quadrant of the tank (coordinate x3-y4-z1 region) as a risk point for insufficient dissolved oxygen (predicted to drop below 35%) leading to metabolic imbalance, and output a warning signal.

[0076] Step 104: Based on the concentration-related change data, iteratively correct the adjustment amount of the fermentation control parameters in the temporal and spatial dimensions, and generate feedback control signals to dynamically adjust the fermentation process of Ganoderma lucidum pills. The fermentation control parameters include feeding rate and aeration rate.

[0077] In this step, the adjustment amount in the time dimension represents the gradient of the feed rate over time (mL / min). 2 The spatial dimension adjustment represents the distribution coefficient (0-1 normalized value) of ventilation volume in different areas. Feedback control signals refer to the set of execution instructions automatically generated by the system, including timing control instructions, spatial control instructions, and safety constraints.

[0078] In this embodiment, the basic feed amount is calculated using a composite control algorithm of proportional-integral-derivative and fuzzy logic based on the predicted concentration ratio change; the ventilation weight of each region is determined by fluid dynamics simulation in combination with the spatial distribution of dissolved oxygen; the parameters are iterated every 30 minutes, and the adjustment instructions are sent to the distributed actuator (feed pump zone valve group), while the control parameters are updated synchronously through the OPC protocol.

[0079] For example, based on the predicted results, the control system formulated an adjustment plan: increasing the feeding rate from the current 50 ml / min to 65 ml / min (the increase was obtained by multiplying the prediction deviation value by a compensation coefficient of 0.8), and simultaneously adjusting the ventilation distribution, increasing the proportion of lower-layer ventilation from 30% to 40% (calculated based on the area volume ratio and dissolved oxygen deficit). After 2 hours of execution, monitoring data showed that the dissolved oxygen in the lower layer recovered to 42%, the polysaccharide accumulation rate recovered to 0.45 g / h, the acid synthesis rate stabilized at 0.09 mg / h, and the metabolism in each area returned to a balanced state.

[0080] This solution improves the consistency of target metabolites (polysaccharides and acids) in different areas of the fermentation tank during the fermentation process of Ganoderma lucidum pills by establishing a complete technology chain from real-time monitoring to intelligent control, making the overall distribution more uniform; it effectively maintains the stability of metabolite ratios and reduces quality fluctuations between production batches. At the same time, this solution reduces the ineffective consumption of energy and raw materials through precise control.

[0081] To address the issue of insufficient precision in the dynamic regulation of metabolite concentration during the fermentation of Ganoderma lucidum pills, in some embodiments, step 104: iteratively correcting the adjustment amounts of the fermentation control parameters in the temporal and spatial dimensions based on the concentration-related change data, and generating a feedback control signal, includes:

[0082] Step 201: Based on the concentration correlation change data and the dynamic change information of dissolved oxygen concentration, and combined with the preset fuzzy compensation rules, generate a control command sequence.

[0083] In step 201, the fuzzy compensation rule refers to an empirical comparison table of "deviation degree - compensation intensity" established based on historical fermentation data. The deviation degree includes the percentage deviation of the polysaccharide to acid concentration ratio from the target range, and the compensation intensity corresponds to the adjustment level of the feeding rate and aeration rate (divided into three levels: strong, medium, and weak). The control command sequence is a time-series command set composed of control parameter adjustment schemes for multiple time nodes. Each node includes the feeding rate adjustment value and the aeration rate allocation ratio for each region.

[0084] In this embodiment, the system first analyzes the concentration correlation data of Ganoderma lucidum polysaccharides and ganoderic acids, and combines this with the spatial distribution characteristics of dissolved oxygen to perform matching analysis using a preset fuzzy compensation rule library. By comparing the deviation of the real-time concentration ratio from the target interval, the basic compensation level is determined, and a regional correction coefficient is calculated based on the distribution differences of dissolved oxygen in each region. The basic compensation value and the regional correction coefficient are then superimposed to generate a preliminary control command sequence containing multiple control time nodes. This sequence clearly defines the changing trend of the feeding rate at different time periods and the allocation scheme of ventilation volume in each region.

[0085] Step 202: Based on the control command sequence and the concentration-related change data, construct an adaptive matching relationship between the prediction deviation and the compensation threshold parameter.

[0086] In step 202, the prediction deviation refers to the degree of difference between the actual concentration change value and the model prediction value. The generation process involves comparing the prediction parameters in the control command sequence with the actual detection values ​​of the online sensor array, which are updated in real time, to calculate the prediction deviation. The compensation threshold parameter is an upper limit of allowable deviation set according to different fermentation stages. The generation process involves dynamically generating a compensation threshold parameter adapted to the current fermentation stage based on the fluctuation characteristics of the concentration-related change data. The adaptive matching relationship is a feedback mechanism that dynamically adjusts the compensation parameters to keep the prediction deviation within the threshold range.

[0087] In this embodiment, during the execution of control commands, the system continuously monitors the deviation between the actual metabolite concentration change and the predicted value. When a continuous periodic deviation exceeds a preset threshold, an adaptive adjustment mechanism is activated. By dynamically evaluating the correspondence between the deviation and the compensation effect, the matching coefficient of the compensation parameters is recalculated, and the correction range of the feeding rate and aeration rate is adjusted. This process establishes a real-time deviation-compensation mapping relationship, ensuring that the control parameters always maintain optimal matching with the current fermentation state.

[0088] Step 203: Generate a feedback control signal based on the adaptive matching relationship.

[0089] In this embodiment, based on the optimized compensation parameter matching relationship, the system integrates the timing control requirements and spatial distribution characteristics to generate the final feedback control signal. This signal transforms the feeding rate adjustment scheme into a specific flow control curve, specifying the flow rate variation over time; simultaneously, it transforms the ventilation volume allocation scheme into precise opening commands for the air intake valves in each area, forming a spatiotemporally coordinated closed-loop control strategy. The control signal is transmitted to the actuator in real time via the industrial bus, completing the entire process from analysis and decision-making to physical control.

[0090] Here is a specific example:

[0091] During the cultivation of Ganoderma lucidum mycelium in a 500-liter stainless steel fermenter, when the system detected that the polysaccharide concentration reached 18.5 g / L at the 48th hour of fermentation (data was collected every 10 minutes by a near-infrared sensor, and the average of the last 6 measurements was taken) and the ganoderic acid concentration was 1.4 mg / mL (high performance liquid chromatography detection value), the current polysaccharide / acid concentration ratio was calculated to be 13.2 (18.5 ÷ 1.4), which deviates from the target ratio of 12.0 by +10%. The system combines dissolved oxygen data from different areas of the tank (40% at the bottom, 45% in the middle, and 50% at the top) and queries a pre-set fuzzy compensation rule base (this rule base is built based on historical data, specifying that a deviation of 10-15% corresponds to intermediate compensation). It determines that the basic feeding rate needs to be increased by 15% (the current feeding rate is 60 ml / min, so it is increased by 9 ml to 69 ml / min). At the same time, based on the dissolved oxygen gradient distribution, it calculates that the bottom area needs to have an additional 5% of ventilation (the compensation coefficient is calculated as: basic compensation value 15% × regional dissolved oxygen deviation coefficient 0.33, where 0.33 is determined by the ratio of the difference between the bottom dissolved oxygen of 40% and the average value of 45%). During execution, the system monitored the actual concentration changes in real time and found that the acid synthesis rate was 0.05 mg / h lower than the predicted value (predicted value 0.08, actual value 0.03). The compensation parameters were recalculated through the deviation feedback algorithm (new feeding rate = original compensation value × (1 + deviation correction coefficient 0.2) = 69 × 1.2 ≈ 83 mL / min). Finally, a phased control command was generated: the feeding rate was linearly increased from 60 to 70 mL / min in the first 30 minutes, and increased to 83 mL / min in the next 30 minutes; the aeration distribution was adjusted to 45% at the bottom, 40% in the middle, and 15% at the top, so that the ratio returned to 12.1 after 50 hours of cultivation, and the dissolved oxygen in each area was stabilized at 42±3%.

[0092] In this embodiment, by establishing a complete control chain from deviation detection to parameter adaptation, the optimal ratio of polysaccharide to acid concentration is ensured, the dissolved oxygen balance in each area of ​​the fermenter is maintained, and the impact of environmental fluctuations on the fermentation process is eliminated through a dynamic compensation mechanism.

[0093] To address the issue of unclear adjustment direction of control parameters during the fermentation of Ganoderma lucidum pills, in some embodiments, step 203: generating a feedback control signal based on the adaptive matching relationship includes:

[0094] Step 301: Based on the degree of correspondence between the predicted deviation and the compensation threshold parameter in the adaptive matching relationship, determine the adjustment direction of the fermentation control parameter. The adjustment direction includes the incremental change trend in the time dimension and the parameter distribution offset pattern in the spatial dimension.

[0095] In step 301, the incremental change trend in the time dimension refers to the direction and magnitude of the increase or decrease in the feeding rate over time, which is divided into five levels: rapid increase, slow increase, stable, slow decrease, and rapid decrease. The parameter distribution offset pattern in the spatial dimension refers to the distribution and adjustment scheme of the aeration volume in different areas of the fermenter, including three basic modes: center-gathering type (strengthening aeration in the center), edge-compensation type (strengthening aeration at the edge), and uniform diffusion type (proportional adjustment).

[0096] In this embodiment, the system determines the adjustment direction of fermentation control parameters by analyzing the relative relationship between the predicted deviation and the compensation threshold parameter. In the temporal dimension, it classifies incremental change trends into different levels based on the degree of deviation, including characteristics such as the rate of change and duration. In the spatial dimension, it identifies parameter distribution shift patterns based on abnormal metabolite distribution areas, determining key locations in each region where regulation needs to be strengthened or weakened. This step transforms the abstract adaptive matching relationship into a specific regulatory strategy guide by establishing a mapping rule between the degree of deviation and the adjustment direction.

[0097] Step 302: Generate a feedback control signal based on the incremental change trend in the time dimension and the parameter distribution offset pattern in the spatial dimension.

[0098] In this embodiment, based on a determined adjustment direction, the system transforms the incremental change trend in the time dimension into a continuous control curve for the feeding rate, setting the slope and duration of the curve according to the trend level. Simultaneously, it converts the spatial distribution offset pattern into a regional allocation scheme for aeration volume, adjusting the aeration ratio according to identified key areas. Finally, the time-series control curve and the spatial allocation scheme are fused and calibrated to generate a feedback control signal that simultaneously contains time-series control commands and spatial distribution parameters, achieving precise control of the fermentation process.

[0099] Here is a specific example:

[0100] During the cultivation of Ganoderma lucidum mycelium in a 500-liter stainless steel fermenter, when the system detected a deviation of +12% in the polysaccharide / acid concentration ratio at the 50th hour of fermentation (current ratio 13.4, target ratio 12.0), it first determined, based on the ratio of the deviation degree to the compensation threshold parameter of 1.2 (12% ÷ 10% threshold), that a "rapidly rising" temporal adjustment trend and a "centrally clustered" spatial offset mode were required. The system determined the feeding rate adjustment scheme through a preset adjustment intensity calculation formula (basic adjustment amount × deviation coefficient 1.2): the first stage (0-20 minutes) was increased at a slope of 0.6 ml per minute (basic slope 0.5 ml × 1.2), and the second stage (20-40 minutes) was increased at a slope of 0.4 ml per minute. Meanwhile, based on the dissolved oxygen distribution (42% in the central area, 45% in the transition area, and 48% in the edge area), and following the calculation rules of the central aggregation mode (central area compensation = basic compensation × (1 + central area dissolved oxygen deviation coefficient 0.8)), the ventilation allocation was adjusted as follows: the ventilation ratio in the central area was increased to 46% (originally 40% + 6%), the transition area was maintained at 40%, and the edge area was reduced to 14%. After implementing this control scheme for 2 hours, the dissolved oxygen in the central area increased to 45%, and the polysaccharide / acid ratio deviation was reduced to +6%.

[0101] In this embodiment of the application, by establishing clear rules for determining the adjustment direction and a control signal generation mechanism, the accuracy and consistency of the control parameter adjustment are ensured, the response speed to sudden fermentation anomalies is improved, and the uniformity of material distribution in the fermenter is effectively improved through spatiotemporal coordinated regulation.

[0102] To further improve the accuracy and reliability of the fermentation control signal for Ganoderma lucidum pills, in some embodiments, step 302: generating a feedback control signal based on the incremental change trend in the time dimension and the parameter distribution offset pattern in the spatial dimension includes:

[0103] Step 401: Calculate the single-step adjustment factor of the control parameter based on the incremental change trend in the time series dimension.

[0104] In step 401, the single-step adjustment factor refers to the change in the control parameter within a single time unit, used to quantify the adjustment magnitude of each control step in the time series dimension. Its value is determined by the type and intensity of the incremental change trend. For example, the calculation process is as follows: extract the concentration change data from the current moment to the previous N time steps, calculate the difference in the rate of change between Ganoderma lucidum polysaccharide and Ganoderma lucidum acid concentrations (e.g., Ganoderma lucidum acid increases 0.2% / min faster than polysaccharide); according to a preset trend-adjustment mapping table, convert the rate of change difference into an initial adjustment factor base value (e.g., a difference of 0.2% / min corresponds to a base value + 0.5). The initial base value is weighted based on the direction of dissolved oxygen concentration change (increasing / decreasing): if dissolved oxygen decreases, the weight of the Ganoderma lucidum acid increase is increased, and the adjustment factor base value is multiplied by 1.2; otherwise, it is multiplied by 0.8; output the unconstrained single-step adjustment factor (e.g., final value = 0.5 × 1.2 = 0.6).

[0105] In this embodiment, the system calculates the basic single-step adjustment factor based on the determined "rapid rise" trend type using a trend-factor conversion model. This model considers factors such as the current fermentation stage and historical control effects, and outputs the theoretical adjustment amount of the feeding rate within each control cycle (e.g., 15 minutes).

[0106] Step 402: Apply an upper limit constraint on the adjustment range of the single-step adjustment factor to obtain an amplitude-limited adjustment factor.

[0107] In step 402, the adjustment range upper limit constraint refers to the maximum allowable adjustment amount per step set to ensure fermentation safety, including two types of restrictions: absolute value upper limit and relative change rate upper limit. For example, the implementation process is as follows: read the adjustment range upper limit for the current fermentation stage (e.g., maximum allowable adjustment of feed rate per step ±10 L / h) from a predefined process parameter library; convert the upper limit value into an equivalent constraint range for the adjustment factor (e.g., ±0.8). If the single-step adjustment factor > the upper limit value (e.g., 0.6 > 0.8), retain the original value; if the single-step adjustment factor < the negative upper limit value (e.g., -1.0 < -0.8), force it to be set to -0.8; output the amplitude-limited adjustment factor (e.g., 0.6 becomes 0.6, -1.0 becomes -0.8). The amplitude-limited adjustment factor refers to the single-step adjustment factor after the adjustment range upper limit constraint processing, used to ensure that the change in control parameters (e.g., feed rate adjustment) within each time unit does not exceed the maximum safe range allowed by the equipment or process.

[0108] In this embodiment, the system retrieves the safety parameter range for the current stage from the fermentation process database and performs double verification on the calculated single-step adjustment factor: first, it compares whether the absolute value exceeds the equipment's allowable limit, and then it calculates whether the increase ratio relative to the current value exceeds the limit. Factors that pass the verification are converted into amplitude-limited adjustment factors.

[0109] Step 403: Perform a smoothness threshold check on the amplitude-limited adjustment factor to generate a single-step adjustment factor sequence.

[0110] In step 403, the smoothness threshold check is a filtering condition used to ensure that the adjustment amounts of adjacent time units change smoothly, avoiding drastic fluctuations in the control parameters. The single-step adjustment factor sequence is a continuous set of control instructions composed of amplitude-limited adjustment factors from multiple time units arranged in chronological order.

[0111] In this embodiment, the system establishes a sliding time window (e.g., 3 adjustment cycles) and calculates the gradient of the adjustment factor within the window. When a new factor causes the gradient to exceed a threshold, a smoothing correction is performed based on the average rate of change of the previous two cycles, ultimately outputting a single-step adjustment factor sequence that meets the continuity requirement.

[0112] Step 404: Generate a spatial coefficient matrix based on the parameter distribution offset pattern in the spatial dimension.

[0113] In step 404, the spatial coefficient matrix is ​​a two-dimensional array that quantifies the control intensity of each region of the fermenter. The matrix rows and columns correspond to spatial positions, and the element values ​​represent the parameter adjustment weights of that region.

[0114] In this embodiment, based on the "central clustering" distribution offset pattern, the system generates an initial coefficient matrix: the central region is assigned the highest coefficient, decreasing layer by layer towards the periphery. The coefficient value is determined by the regional importance weight and the current dissolved oxygen deviation. Specifically, the fermentation vessel is divided into K independent regions (e.g., three upper and lower layers × three left and right zones), and the concentration gradient difference between ganoderic acid and polysaccharides in each region is calculated (e.g., gradient difference in region A = 0.3% / cm). The offset direction is determined based on the sign (positive / negative) of the gradient difference (e.g., a positive value indicates clustering towards the region center). The main diagonal elements represent the self-adjustment intensity of each region, with a value equal to the absolute value of the gradient difference in that region (e.g., element corresponding to region A = 0.3). The off-diagonal elements represent the synergistic effect between regions; if two regions have the same offset direction, the value is assigned to the average gradient difference between the two regions (e.g., regions A and B are in the same direction, element = 0.3 + 0.2 / 2 = 0.25). The unpruned spatial coefficient matrix (K×K dimensions) is output. Example:

[0115] Example Matrix (Simplified 3-Region Version)

[0116]

[0117] Step 405: Perform amplitude upper limit pruning on the matrix elements of the spatial coefficient matrix.

[0118] In step 405, amplitude upper limit pruning is a normalization process for the elements of the spatial coefficient matrix to prevent over-adjustment in local areas.

[0119] In this embodiment, the system sets a maximum adjustment coefficient limit for a single region, and scales out-of-limit elements in the matrix proportionally to maintain the relative proportion between regions. For example, amplitude upper limit pruning means applying an absolute value upper limit constraint to all elements (such as a maximum value of 0.5 for a single element), and retaining the upper limit value for out-of-limit elements by sign (such as 0.6 corresponding to 0.5, and -0.7 corresponding to -0.5).

[0120] Step 406: Combine the single-step adjustment factor sequence with the pruned spatial coefficient matrix according to a preset spatiotemporal coupling rule.

[0121] In step 406, the spatiotemporal coupling rule is an operational method that integrates the temporal adjustment sequence with the spatial coefficient matrix to ensure that temporal and spatial regulation are coordinated and consistent.

[0122] In this embodiment, the system performs tensor multiplication on the single-step adjustment factor sequence and the spatial coefficient matrix to generate a spatiotemporal joint control tensor. The time dimension of this tensor records the adjustment amount for each time period, and the spatial dimension records the allocation ratio for each region.

[0123] Step 407: Verify the physical feasibility of the fermentation environment for the combined results. If the verification is successful, output the feedback control signal corresponding to the combined results.

[0124] In step 407, physical feasibility verification includes three levels: equipment performance verification, process compliance check, and safety assessment.

[0125] In this embodiment, the system will jointly control the tensor input verification module to sequentially check: whether the actuator can achieve the change rate required by the instruction, whether the adjustment scheme conforms to the process specifications, and whether the key parameters exceed the safety threshold. The tensor that passes the verification is converted into the final control signal.

[0126] Here is a specific example:

[0127] During the cultivation of Ganoderma lucidum mycelium in a 500-liter stainless steel fermenter, when the system detected a polysaccharide / acid concentration ratio deviation of +15% at the 52nd hour of fermentation (current ratio 13.8, target ratio 12.0), based on the ratio of the deviation degree to the compensation threshold parameter of 1.5 (15% ÷ 10% threshold), the system determined that a "rapidly rising" temporal adjustment trend and a "centrally clustered" spatial offset mode should be adopted. First, the single-step adjustment factor was calculated: the base adjustment amount was 0.5 ml / min (a baseline value determined from historical data), multiplied by the deviation coefficient of 1.5 to obtain an initial single-step adjustment factor of 0.75 ml / min. After being constrained by the upper limit of amplitude (the equipment allows a maximum single-step adjustment of 0.7 ml / min), it was corrected to 0.7 ml / min. After smoothness verification (adjustment changes between adjacent time periods not exceeding 0.2 ml), a single-step adjustment factor sequence was generated: 0-15 minutes +0.6 ml / min, 15-30 minutes +0.7 ml / min, 30-45 minutes +0.7 ml / min. Simultaneously, based on the dissolved oxygen distribution (central zone 41%, transition zone 44%, edge zone 47%), a spatial coefficient matrix was generated according to the central aggregation pattern: central zone coefficient 0.5 (calculated as: base coefficient 0.4 × (1 + dissolved oxygen deviation coefficient 0.25)), transition zone 0.3, edge zone 0.2. After amplitude clipping (maximum single-region coefficient 0.45), the matrix was corrected to: central zone 0.45, transition zone 0.3, edge zone 0.2. After spatiotemporal coupling, the actual increase in feed volume in the central zone during the first 15 minutes was 0.6 × 0.45 = 0.27 ml / min. The scheme was verified to meet all safety constraints. After the final control signal was executed for 2 hours, the dissolved oxygen in the central area increased from 41% to 44%, and the polysaccharide / acid ratio deviation decreased to +8%.

[0128] In this embodiment of the application, by establishing a complete spatiotemporal parameter processing and verification mechanism, the stability and security of the control signal are ensured, the accuracy of multi-regional coordinated regulation is improved, and the risk of misoperation is effectively prevented through physical feasibility verification.

[0129] To further improve the accuracy and adaptability of control command generation during the fermentation of Ganoderma lucidum pills, in some embodiments, step 201: generating a control command sequence based on the concentration correlation change data and the dynamic change information of dissolved oxygen concentration, combined with a preset fuzzy compensation rule, includes:

[0130] Step 501: Based on the concentration ratio of Ganoderma lucidum polysaccharide to Ganoderma lucidum acid in the concentration correlation change data, and the dynamic change information of dissolved oxygen concentration, generate a compensation intensity value.

[0131] In step 501, the compensation intensity value refers to a quantitative index of the control intensity calculated based on the current fermentation state. Its value ranges from 0 to 1, where 0 indicates no compensation is needed and 1 indicates the maximum compensation intensity. This value is jointly determined by the degree of deviation in the polysaccharide / acid concentration ratio and the trend of dissolved oxygen changes.

[0132] In this embodiment, the system first calculates the percentage deviation between the current polysaccharide / acid ratio and the target value. Then, combining the instantaneous value of dissolved oxygen and its rate of change, it queries a three-dimensional fuzzy rule table (deviation degree × dissolved oxygen value × dissolved oxygen change trend) to obtain the compensation intensity value. For example, when the deviation is +15% and the dissolved oxygen is 45% and showing a decreasing trend, the output compensation intensity value is 0.8.

[0133] Step 502: Generate the feeding rate based on the product relationship between the compensation intensity value and the base value of the control parameter in the preset fuzzy compensation rule.

[0134] In step 502, the baseline control parameter refers to the preset standard control values ​​for each fermentation stage, including the baseline feeding rate and baseline aeration rate, which are preset according to process requirements. The feeding rate refers to the flow rate of fresh culture medium or nutrients added to the fermenter per unit time (e.g., per minute), and its value directly affects the cell growth rate and the accumulation of metabolites. For example, based on the preset standard feeding amount for the current fermentation stage (e.g., 50 ml / min); a coefficient reflecting the current required control intensity (e.g., 0.8); the actual feeding rate = baseline value × (1 ± compensation intensity value). For positive compensation, the feeding amount is increased (e.g., 50 × 1.8 = 90 ml / min), and for negative compensation, the feeding amount is decreased (e.g., 50 × 0.2 = 10 ml / min).

[0135] In this embodiment, the system retrieves the baseline feeding rate (e.g., 50 ml / min) for the current stage from the process database, multiplies it by the compensation intensity value to obtain the actual feeding rate (50 × 0.8 = 40 ml / min). This calculation ensures that the controlled amount is precisely matched with the fermentation state.

[0136] Step 503: Expand the feeding rate into discrete command points according to the time series.

[0137] In step 503, discrete command points refer to key control nodes of control parameters on the time axis, and each node contains a specific timestamp and parameter value.

[0138] In this embodiment, the system divides the total control duration (e.g., 2 hours) into several time periods (e.g., every 30 minutes), sets a feeding rate value at each time period node, forming a preliminary discrete control point sequence. The feeding rate value at each point is proportionally allocated according to the time period length and the total adjustment amount.

[0139] Step 504: Based on the rate of change of compensation intensity between adjacent discrete command points, insert transition command points in the time sequence to generate a feeding rate command sequence.

[0140] In step 504, the transition command point is an intermediate control point inserted between discrete command points to ensure control continuity and to smooth the parameter change curve.

[0141] In this embodiment, the system calculates the gradient of the feeding rate change between adjacent discrete points. When the gradient exceeds a set threshold (e.g., 0.5 ml per minute), 1-2 transition points are inserted between the two points to make the change curve a linear transition. The final generated feeding rate command sequence includes all discrete points and necessary transition points.

[0142] Step 505: Based on the distribution differences of the dynamic change information of dissolved oxygen concentration in the fermentation vessel, divide the spatial priority of the fermentation zone and assign an independent aeration correction coefficient to each fermentation zone.

[0143] In step 505, spatial priority refers to the ranking of the fermentation regions based on the degree of abnormality in dissolved oxygen distribution. The higher the priority, the more adjustment is needed for that region.

[0144] In this embodiment, the system vertically divides the fermenter into three layers: upper, middle, and lower. It calculates the deviation of dissolved oxygen from the target value for each layer and ranks them in descending order of deviation to determine priority (lower layer > middle layer > upper layer). The aeration correction coefficient is allocated according to priority, with higher priority layers having larger coefficients.

[0145] Step 506: The aeration correction coefficient is superimposed onto the aeration base value of the corresponding fermentation region according to the spatial priority to generate an aeration command sequence for different fermentation regions.

[0146] In step 506, the ventilation command sequence refers to the ventilation adjustment plan for each region at different time points, which includes time and spatial dimension information.

[0147] In this embodiment, the system multiplies the baseline ventilation rate (e.g., 30 cubic meters per hour) by the correction coefficient for each region to obtain the actual ventilation rate for the region (e.g., 36 cubic meters per hour for a lower-level coefficient of 1.2). Ventilation rate variation curves for each region at different time periods are generated according to the time sequence requirements, forming a complete regional instruction sequence.

[0148] Step 507: Combine the feeding rate command sequence with the ventilation rate command sequence to generate a control command sequence.

[0149] In this embodiment, the system aligns and merges the feeding rate command sequence and the ventilation volume command sequence of each region along the time axis. Each time point contains the feeding rate value and the ventilation volume value of each region, forming a control command table that can be directly issued to the actuator.

[0150] Here is a specific example:

[0151] During the cultivation of Ganoderma lucidum mycelium in a 500-liter stainless steel fermenter, when the system detected a polysaccharide / acid concentration ratio deviation of +15% at the 52nd hour of fermentation (current ratio 13.8, target ratio 12.0), the compensation intensity value of 0.9 was first obtained by querying the three-dimensional fuzzy compensation rule table based on the degree of deviation and the dynamic changes in dissolved oxygen (42% in the central area, 44% in the transition area, and 47% in the edge area). (Calculation process: a deviation of 15% corresponds to a base intensity value of 0.8, plus a correction coefficient of +0.1 for the decreasing trend of dissolved oxygen). Using a basic feeding rate of 60 ml / min as a benchmark, the actual feeding rate was calculated to be 60 × (1 + 0.9) = 114 ml / min. This feeding rate was then extended into discrete command points over time: 60 ml at 0 minutes and 114 ml at 60 minutes. Because the rate of change of compensation intensity between adjacent points (0.9 / 60 = 0.015) exceeds the threshold of 0.01, the system inserts a transition command point of 87 ml at 30 minutes (calculated as: 60 + (114 - 60) × 0.5), forming a feeding rate command sequence [0:60, 30:87, 60:114]. Simultaneously, based on the difference in dissolved oxygen distribution (lowest at 42% in the central area), spatial priority is divided as central area > transition area > edge area, with ventilation correction coefficients of 1.3, 1.1, and 0.8 assigned according to priority (calculated as: base coefficient 1.0 + priority weight × 0.3). Based on a base ventilation rate of 30 cubic meters per hour, the generated regional ventilation command sequence is: central area 39 (30 × 1.3), transition area 33 (30 × 1.1), and edge area 24 (30 × 0.8).

[0152] In this embodiment, by establishing a full-process method from parameter analysis to instruction generation, the dynamic matching of control instructions and fermentation state is ensured, the accuracy of multi-parameter collaborative regulation is improved, and the stability of the control process is guaranteed through the transition instruction point mechanism.

[0153] To further improve the prediction accuracy of metabolite concentration changes during Ganoderma lucidum pill fermentation, in some embodiments, step 103: inputting the spatiotemporal map into a pre-trained concentration correlation change prediction model and outputting the concentration correlation change data between Ganoderma lucidum polysaccharides and Ganoderma lucidum acids within a future time window includes:

[0154] Step 601: Input the spatiotemporal map into the pre-trained concentration correlation change prediction model, and output the predicted concentration ratio and spatial distribution offset between Ganoderma lucidum polysaccharide and Ganoderma lucidum acid at each future time point through a multi-step recursive prediction algorithm.

[0155] In step 601, the multi-step recursive prediction algorithm refers to a method that continuously predicts data at multiple time points through iteration, with the output of each prediction step serving as the input for the next step. The concentration ratio prediction value represents the concentration ratio of Ganoderma lucidum polysaccharides to ganoderic acids at a future time. The spatial distribution offset refers to the degree of variation in the distribution of metabolites in different fermentation regions.

[0156] In this embodiment, the prediction model first decomposes the spatiotemporal map into time-series data and spatial feature maps. A time encoder extracts the concentration change trends of polysaccharides and acids, while a spatial encoder analyzes the dissolved oxygen distribution patterns in each region. During the prediction phase, the model starts with the spatiotemporal characteristics of the current moment and progressively predicts the concentration ratios and regional offsets at subsequent time points. Each prediction step is 15 minutes long, and the model makes 12 consecutive prediction steps (3 hours). Each prediction step is dynamically corrected based on the results of the previous step.

[0157] Step 602: Integrate the predicted concentration ratio and spatial distribution offset of Ganoderma lucidum polysaccharide and ganoderic acid at all future time points according to the time window length to generate concentration correlation change data between Ganoderma lucidum polysaccharide and ganoderic acid within the future time window.

[0158] In step 602, the time window length refers to the total duration of the prediction coverage, and the concentration-related change data is a structured dataset that integrates the concentration ratio and spatial offset information of all time points within the prediction period.

[0159] In this embodiment, the system arranges the multi-step prediction results in chronological order, generates a continuous curve using cubic spline interpolation for the concentration ratio data, and performs cluster analysis on the spatial offset by region to mark the hotspot areas of the offset. The final output is a data combination containing the time-concentration ratio curve and the spatial offset heatmap.

[0160] Here is a specific example:

[0161] During the cultivation of Ganoderma lucidum mycelium in a 500-liter stainless steel fermenter, the current spatiotemporal map was input into the prediction model at the 72nd hour of system operation. The map showed: polysaccharide concentration 19.2 g / L (average of the last 6 measurements), ganoderic acid 1.5 mg / mL, and dissolved oxygen spatial distribution: bottom 42%, middle 45%, and top 48%. The model first analyzed the time encoder and found that polysaccharide increased by 0.35 g per hour (linear regression slope), acid increased by 0.07 mg per hour (cubic spline fitting), and dissolved oxygen decreased by 0.8% per hour (bottom region). In the multi-step recursive prediction, the model uses a 15-minute step size, dividing one hour into four steps. The first step predicts an increase of 0.0875 g of polysaccharide (0.35 ÷ 4), an increase of 0.0175 mg of acid (0.07 ÷ 4), and a decrease of 0.2% in dissolved oxygen at the bottom (0.8% ÷ 4). The calculated concentration ratio changes from 12.8 (19.2 ÷ 1.5) to 12.9 ((19.2 + 0.0875) ÷ (1.5 + 0.0175)), with a bottom region offset of +0.5% (current 42% - predicted 41.8%). After 12 consecutive prediction steps (3 hours), the integrated data shows that after 3 hours, the concentration ratio will reach 13.4 (calculated as: 12.8 + (0.1 × 6)), the dissolved oxygen at the bottom region will decrease to 40.4%, and the cumulative offset in the middle region will increase by 1.2%.

[0162] In the embodiments of this application, by using multi-step recursive prediction and spatiotemporal data integration, the effective early warning time window is extended, the accuracy of regional offset prediction is improved, and sufficient data support is provided for proactive regulation.

[0163] To further improve the spatiotemporal resolution of metabolite monitoring during the fermentation of Ganoderma lucidum pills, in some embodiments, step 102: generating a spatiotemporal map reflecting the distribution state of metabolites based on the concentration dynamic change information through multimodal data interaction and fusion includes:

[0164] Step 701: Divide the dynamic change information of the concentration of Ganoderma lucidum polysaccharides, Ganoderma lucidum acid and dissolved oxygen into time series segments according to the time dimension.

[0165] In step 701, a time series segment refers to a time period unit in which continuous monitoring data is divided according to a fixed duration. Each segment contains the concentration change curves and interaction data of Ganoderma lucidum polysaccharides, Ganoderma lucidum acid and dissolved oxygen during that time period.

[0166] In this embodiment of the application, the system segments the original monitoring data in 30-minute intervals and calculates three key indicators for each segment: the slope of polysaccharide concentration change (linear regression coefficient), the acceleration of acid concentration change (second derivative), and the response delay time of dissolved oxygen to the changes of both (cross-correlation analysis).

[0167] Step 702: Label each time series segment with metabolic association features.

[0168] In step 702, metabolic association feature markers refer to a set of features that quantify the interaction relationships between different metabolites, including association patterns such as promotion, inhibition, and synergy.

[0169] In this embodiment, the system analyzes the lead-lag relationship between polysaccharide and acid changes using Granger causality tests, and quantifies the influence of dissolved oxygen on both using correlation coefficients. The results are encoded using a three-color label: red indicates strong inhibition (correlation coefficient <-0.7), green indicates strong promotion (>0.7), and yellow indicates weak correlation (-0.3 to 0.3).

[0170] Step 703: Divide the fermentation container used for Ganoderma lucidum pill fermentation into multiple independent fermentation zones, and extract the spatial distribution difference ratio of Ganoderma lucidum polysaccharides, Ganoderma lucidum acids and dissolved oxygen in each fermentation zone to generate a spatial distribution feature map of each fermentation zone.

[0171] In step 703, the spatial distribution feature map is a quantitative map reflecting the non-uniformity of metabolite distribution in the fermentation vessel, including regional concentration gradients and boundary diffusion trends.

[0172] In this embodiment, the fermenter is vertically divided into three layers: upper, middle, and lower. Each layer is further divided into three annular zones: center, middle, and edge, resulting in a total of nine regions. For each region, the following are calculated: polysaccharide concentration region ratio (concentration in that region / average concentration), acid concentration coefficient of variation (standard deviation / mean), and dissolved oxygen gradient (concentration difference with adjacent regions). A continuous distribution surface is generated using Kriging interpolation.

[0173] Step 704: Cross-bind the metabolic association feature labeling results with the spatial distribution feature map at the corresponding time to generate a spatiotemporal map.

[0174] In step 704, cross-binding refers to the fusion process of establishing a mapping relationship between temporal features and spatial features to form a data entity with both temporal and spatial attributes.

[0175] In this embodiment, the system matches the spatial feature map corresponding to each time segment, and fuses the two through a spatiotemporal encoder: the time features are converted into positional codes, and the spatial features are converted into channel attention weights. The final output is a three-dimensional spatiotemporal map with dimensions of [time step × spatial partition × feature channel].

[0176] Here is a specific example:

[0177] During the cultivation of Ganoderma lucidum mycelium in a 500-liter stainless steel fermenter, when the system reached the 36th hour, the continuous monitoring data was first divided into time series segments at 30-minute intervals. Taking the 36-36.5 hour segment as an example: the polysaccharide concentration measured by the near-infrared sensor increased from 16.2 g / L to 16.5 g / L (slope calculation: (16.5-16.2)×2=+0.6 g / L / hour), the acid concentration measured by high performance liquid chromatography increased from 1.1 mg / mL to 1.12 mg / mL (acceleration calculation: [(1.12-1.1)×2-previous period growth rate 0.015]×2=+0.01 mg / mL / hour2), and the dissolved oxygen decreased from 46% to 44% (the delayed response time was determined to be 15 minutes through cross-correlation analysis). Based on the correlation coefficient of -0.65 between the decrease in dissolved oxygen and the increase in acid growth in this segment, it was marked as "dissolved oxygen moderately inhibits acid synthesis". The fermenter was divided into three independent regions: upper (z1), middle (z2), and lower (z3). Calculations showed that the lower region had a polysaccharide content of 118% (16.5 ÷ average concentration 14.0), an acid coefficient of variation of 0.18 (standard deviation 0.2 ÷ mean 1.12), and a dissolved oxygen gradient of -4% / cm (lower layer 44% - middle layer 48%). Finally, by binding the temporal segment characteristics with spatial distribution data, a spatiotemporal map was generated, showing that the lower region exhibited a metabolic state of "high polysaccharide (118%) - low oxygen (44%) - acid inhibition" after 36.5 hours, providing precise spatiotemporal localization for subsequent regulation.

[0178] In this embodiment of the application, by using spatiotemporal dual-dimensional data fusion and feature labeling, local metabolic abnormality regions that are difficult to detect by traditional methods are revealed, and a visual expression of metabolite interaction relationships is established, providing spatial positioning basis for precise regulation.

[0179] Figure 2 This application provides a schematic diagram of the structure of a dynamic feedback control system for the fermentation process of Ganoderma lucidum pills, as shown in the embodiment of the present application. Figure 2 As shown, the system includes:

[0180] The acquisition module 21 is used to acquire dynamic change information on the concentrations of Ganoderma lucidum polysaccharides, Ganoderma lucidum acid and dissolved oxygen in the fermentation broth during the fermentation process of Ganoderma lucidum pills.

[0181] The generation module 22 is used to generate a spatiotemporal map reflecting the distribution status of metabolites based on the concentration dynamic change information through multimodal data interaction and fusion.

[0182] Input module 23 is used to input the spatiotemporal map into a pre-trained concentration correlation change prediction model and output the concentration correlation change data between Ganoderma lucidum polysaccharide and Ganoderma lucidum acid within a future time window.

[0183] The correction module 24 is used to iteratively correct the adjustment amount of the fermentation control parameters in the temporal and spatial dimensions based on the concentration-related change data, and generate feedback control signals to dynamically adjust the fermentation process of Ganoderma lucidum pills. The fermentation control parameters include feeding rate and aeration rate.

[0184] Figure 2 The aforementioned dynamic feedback control system for the fermentation process of Ganoderma lucidum pills can execute... Figure 1 The implementation principle and technical effects of the dynamic feedback control method for the fermentation process of Ganoderma lucidum pills described in the above embodiment will not be repeated here. The specific methods by which each module and unit of the dynamic feedback control system for the fermentation process of Ganoderma lucidum pills in the above embodiment perform operations have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0185] In one possible design, Figure 2 The dynamic feedback control system in the fermentation process of Ganoderma lucidum pills shown in the embodiment can be implemented as a computing device, such as... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0186] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0187] The processing component 32 is described above Figure 1 The embodiment describes a dynamic feedback control method for the fermentation process of Ganoderma lucidum pills.

[0188] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0189] Storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0190] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0191] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0192] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0193] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0194] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The embodiment shown illustrates a dynamic feedback control method for the fermentation process of Ganoderma lucidum pills.

[0195] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0196] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0197] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0198] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A dynamic feedback control method for the fermentation process of Ganoderma lucidum pills, characterized in that, include: During the fermentation of Ganoderma lucidum pills, information on the dynamic changes in the concentrations of Ganoderma lucidum polysaccharides, Ganoderma lucidum acids, and dissolved oxygen in the fermentation broth was obtained. Based on the aforementioned concentration dynamic change information, a spatiotemporal map reflecting the distribution status of metabolites is generated through multimodal data interaction and fusion. The spatiotemporal map is input into a pre-trained concentration correlation change prediction model, which outputs the concentration correlation change data between Ganoderma lucidum polysaccharide and Ganoderma lucidum acid within a future time window. Based on the concentration-related change data, the fermentation control parameters are iteratively adjusted in both the temporal and spatial dimensions to generate feedback control signals, thereby dynamically regulating the fermentation process of Ganoderma lucidum pills. The fermentation control parameters include feeding rate and aeration rate.

2. The method according to claim 1, characterized in that, The step of iteratively correcting the adjustment amounts of fermentation control parameters in the temporal and spatial dimensions based on the concentration-related change data, and generating feedback control signals, includes: Based on the concentration correlation change data and the dynamic change information of dissolved oxygen concentration, and combined with the preset fuzzy compensation rules, a control command sequence is generated. Based on the control command sequence and the concentration-related change data, an adaptive matching relationship is constructed between the prediction deviation and the compensation threshold parameter; Based on the adaptive matching relationship, a feedback control signal is generated.

3. The method according to claim 2, characterized in that, The step of generating a feedback control signal based on the adaptive matching relationship includes: Based on the degree of correspondence between the predicted deviation and the compensation threshold parameter in the adaptive matching relationship, the adjustment direction of the fermentation control parameter is determined. The adjustment direction includes the incremental change trend in the time dimension and the parameter distribution offset pattern in the spatial dimension. Feedback control signals are generated based on the incremental change trend in the time dimension and the parameter distribution offset pattern in the spatial dimension.

4. The method according to claim 3, characterized in that, The step of generating a feedback control signal based on the incremental change trend in the temporal dimension and the parameter distribution offset pattern in the spatial dimension includes: Based on the incremental change trend in the time dimension, calculate the single-step adjustment factor of the control parameter; By imposing an upper limit constraint on the adjustment range of the single-step adjustment factor, an amplitude-limited adjustment factor is obtained; A smoothness threshold check is performed on the amplitude-limited adjustment factor to generate a single-step adjustment factor sequence; Generate a spatial coefficient matrix based on the parameter distribution offset pattern in the spatial dimension; Amplitude upper limit pruning is performed on the matrix elements of the spatial coefficient matrix; The single-step adjustment factor sequence is combined with the clipped spatial coefficient matrix according to a preset spatiotemporal coupling rule; The physical feasibility of the combined fermentation environment is verified, and if the verification is successful, a feedback control signal corresponding to the combined result is output.

5. The method according to claim 2, characterized in that, The control command sequence is generated based on the concentration correlation change data and the dynamic change information of dissolved oxygen concentration, combined with a preset fuzzy compensation rule, including: Based on the concentration ratio of Ganoderma lucidum polysaccharide to Ganoderma lucidum acid in the concentration correlation change data, and the dynamic change information of dissolved oxygen concentration, a compensation intensity value is generated. Based on the product relationship between the compensation intensity value and the base value of the control parameter in the preset fuzzy compensation rule, the feeding rate is generated; The feeding rate is extended into discrete command points according to the time series; Based on the rate of change of compensation intensity between adjacent discrete command points, transition command points are inserted in the time sequence to generate a feeding rate command sequence. Based on the distribution differences of the dynamic changes in dissolved oxygen concentration within the fermentation vessel, the spatial priority of the fermentation zones is determined, and an independent aeration correction coefficient is assigned to each fermentation zone. The ventilation correction coefficient is superimposed onto the ventilation base value of the corresponding fermentation region according to the spatial priority to generate a ventilation instruction sequence for different fermentation regions. The feeding rate command sequence is combined with the ventilation rate command sequence to generate a control command sequence.

6. The method according to claim 1, characterized in that, The step of inputting the spatiotemporal map into a pre-trained concentration correlation change prediction model and outputting the concentration correlation change data between Ganoderma lucidum polysaccharides and Ganoderma lucidum acids within a future time window includes: The spatiotemporal map is input into a pre-trained concentration correlation change prediction model. Through a multi-step recursive prediction algorithm, the predicted concentration ratio and spatial distribution offset of Ganoderma lucidum polysaccharide and Ganoderma lucidum acid at each future time point are output sequentially. The predicted concentration ratio and spatial distribution offset of Ganoderma lucidum polysaccharides and ganoderic acids at all future time points are integrated according to the time window length to generate concentration correlation change data between Ganoderma lucidum polysaccharides and ganoderic acids within the future time window.

7. The method according to claim 1, characterized in that, The generation of a spatiotemporal map reflecting the distribution of metabolites based on the dynamic concentration change information, through multimodal data interaction and fusion, includes: The dynamic changes in the concentrations of Ganoderma lucidum polysaccharides, ganoderic acids, and dissolved oxygen were divided into time series segments according to the time dimension; Metabolic association features are labeled for each time series segment; The fermentation container used for fermenting Ganoderma lucidum pills was divided into multiple independent fermentation zones. The spatial distribution differences of Ganoderma lucidum polysaccharides, Ganoderma lucidum acids and dissolved oxygen in each fermentation zone were extracted to generate spatial distribution feature maps of each fermentation zone. The metabolic association feature labeling results are cross-linked with the spatial distribution feature maps at the corresponding time points to generate a spatiotemporal map.

8. A dynamic feedback control system for the fermentation process of Ganoderma lucidum pills, characterized in that, include: The acquisition module is used to acquire dynamic change information on the concentrations of Ganoderma lucidum polysaccharides, Ganoderma lucidum acids and dissolved oxygen in the fermentation broth during the fermentation process of Ganoderma lucidum pills. The generation module is used to generate a spatiotemporal map reflecting the distribution status of metabolites based on the concentration dynamic change information through multimodal data interaction and fusion. The input module is used to input the spatiotemporal map into a pre-trained concentration correlation change prediction model and output the concentration correlation change data between Ganoderma lucidum polysaccharide and Ganoderma lucidum acid within a future time window; The correction module is used to iteratively correct the adjustment amount of the fermentation control parameters in the temporal and spatial dimensions based on the concentration-related change data, and generate feedback control signals to dynamically adjust the fermentation process of Ganoderma lucidum pills. The fermentation control parameters include feeding rate and aeration rate.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the dynamic feedback control method for the fermentation process of Ganoderma lucidum pills as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The device contains a computer program that, when executed by a computer, implements a dynamic feedback control method for the fermentation process of Ganoderma lucidum pills as described in any one of claims 1 to 7.

Images