Dough humidity and temperature dynamic regulation and control system based on multi-modal sensor fusion

Abstract

The invention discloses a dough humidity and temperature dynamic regulation and control system based on multi-modal sensor fusion, and relates to the technical field of operation parameter regulation and control, and the method comprises the steps: carrying out the non-blind area monitoring of a three-dimensional temperature and humidity field in a fermentation box, and eliminating the local failure risk of a point-type sensor; capturing a space-time coupling relationship of temperature and humidity-carbon dioxide concentration-dough volume through a self-attention mechanism; constructing a differential equation set comprising at least one state variable; establishing a fermentation equipment thermodynamic model, optimizing refrigeration and heating time sequences through reinforcement learning, and designing an energy consumption optimization strategy based on digital twinning; performing cross-device cooperative control; constructing a hierarchical diagnosis tree containing fault features; performing microbial activity enhanced humidity regulation and control, and forming a protective layer on the surface of the dough based on a humidity gradient control algorithm of a water molecule diffusion coefficient. And by arranging the environment regulation and control module, the time compression module and the time sequence optimization module, the product quality, the production efficiency and the energy utilization rate are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of operating parameter regulation, and specifically relates to a dynamic regulation system for dough humidity and temperature based on multi-modal sensor fusion. Background Art

[0002] The scale of the baking market is expanding day by day. Data-driven decision optimization requires building a closed-loop control system from raw materials to finished products through multi-modal sensor fusion and dynamic regulation algorithms.

[0003] Most systems only implement basic parameter monitoring such as temperature and humidity, and do not establish a multi-variable coupling relationship model, making it impossible to meet the dynamic process requirements. Consumers' requirements for the safety and consistency of baking products are increasing. It is necessary to reduce the contamination of miscellaneous bacteria through enhanced microbial activity and skin protection technology, with a target inhibition rate of ≥95%, and reduce the risk of manual operation. Summary of the Invention

[0004] To solve the above technical problems, a dynamic regulation system for dough humidity and temperature based on multi-modal sensor fusion is provided, and this technical solution solves the problems raised in the above background art.

[0005] To achieve the above object, the technical solution adopted by the present invention is as follows: A dynamic regulation system for dough humidity and temperature based on multi-modal sensor fusion, comprising: A parameter monitoring module, which is used to adopt a distributed optical fiber sensing network to realize non-blind area monitoring of the three-dimensional temperature and humidity field in the fermentation chamber, eliminating the risk of local failure of point sensors; An environmental regulation module, which is used to introduce a deep learning architecture into the fermentation environment regulation, and capture the spatio-temporal coupling relationship of temperature and humidity - carbon dioxide concentration - dough volume through a self-attention mechanism; A time compression module, which is used to construct a differential equation system containing at least one state variable, quantify the influence of temperature, humidity, and pH value on the yeast gas production rate, and dynamically compress the fermentation time based on non-linear fermentation process control; A timing optimization module, which is used to establish a thermodynamic model of the fermentation equipment, optimize the refrigeration and heating timing through reinforcement learning, and design an energy consumption optimization strategy based on digital twin; A collaborative control module, which is used to develop a communication protocol interface to realize full-process parameter linkage of mixer - fermentation chamber - proofing box - oven, and perform cross-device collaborative control; A fault identification module, which is used to construct a hierarchical diagnostic tree containing fault features, and perform pattern recognition in combination with a deep belief network, and the fault features include sensor-level, equipment-level, and process-level features; Gradient control module, which is used to perform microbial activity-enhanced humidity regulation, and form a protective layer on the dough surface based on the humidity gradient control algorithm of the water molecule diffusion coefficient.

[0006] Preferably, the environmental regulation module specifically includes: Record the temperature, humidity, carbon dioxide concentration and dough volume data at fixed time intervals to form a time series data set; Normalize the temperature, humidity, carbon dioxide concentration and dough volume data to the [0, 1] interval to eliminate the influence of dimensions and accelerate the convergence of the model; Select the long short-term memory network as the deep learning architecture for processing time series data; For the input of each time step of temperature, humidity, carbon dioxide concentration and dough volume, generate query, key and value vectors through linear transformation; Calculate the dot product of the query vector and all key vectors, obtain the attention weights through normalization, and then perform weighted summation on the value vectors to obtain the output of the current time step; Adopt the multi-head self-attention mechanism to map the input vector to at least one subspace, calculate the self-attention respectively, and then splice the results to enhance the expression ability of the model; Adopt the encoder-decoder structure. The encoder is responsible for processing the input sequence, and the decoder generates the predicted value of the dough volume at the current time step based on the output of the encoder and the predicted value of the dough volume before the current time step; Select the mean square error function as the loss function to measure the difference between the predicted dough volume and the real dough volume; The self-attention mechanism automatically captures the spatio-temporal coupling relationship between temperature, humidity, carbon dioxide concentration and dough volume by calculating the attention weights between different time steps; Use the trained model to predict the temperature, humidity and carbon dioxide concentration during the fermentation process in real time, and adjust the fermentation environment parameters according to the prediction results; Compare the model prediction results with the actual measured values to form a closed-loop control system to achieve precise regulation of the fermentation environment.

[0007] Preferably, the time compression module specifically includes: Based on the non-linear expression of the gas production rate, construct a dynamic model of yeast gas production rate; Set the coupling mechanism between temperature-humidity-pH state variables; Based on the mass conservation and energy conservation, establish a three-variable coupling equation set and calculate the real-time optimal parameter values. The three-variable coupling equation set includes the cell growth equation, substrate consumption equation, temperature change equation, pH dynamic equation and humidity dynamic equation; Adjust the real-time temperature through the steam and cooling water flow rates in the jacket layer; Based on the Arrhenius acceleration effect, the metabolic rate is increased exponentially by increasing the temperature, and a dynamic balance is established between enzyme activity and stability; Automatically add alkali or acid solution, and adjust the real-time pH value by controlling the flow rate through a peristaltic pump; Based on the pH buffer synergistic effect, near the optimal pH, the enzyme activity and substrate absorption efficiency are synergistically improved, accelerating the metabolic process; The humidification amount is controlled by the atomizing nozzle on the top of the tank, and the real-time humidity is adjusted in conjunction with the ventilation amount.

[0008] Preferably, the timing optimization module specifically includes: Establish a three-dimensional non-steady-state heat balance equation for the temperature field in the fermentation tank and set the energy interaction term; Set the metabolic heat production term to quantify the exponential correlation between bacterial metabolic rate and temperature, and introduce the heat production coefficient to characterize the Joule heat generated per unit bacterial biomass; Set the heat conduction term, use Fourier's law to describe the radial heat conduction of the tank wall, and establish the functional relationship between the thermal conductivity of the composite material and the temperature; The convective heat transfer term is set, and based on Newton's cooling law, a dynamic correlation model between the turbulent convective heat transfer coefficient of the fermentation liquid and the flow velocity and temperature gradient is constructed; Set the radiation heat transfer term, introduce the surface emissivity correction factor based on the radiation heat exchange between the tank surface and the environment; Set the latent heat of evaporation term, establish the relationship model between humidity gradient and evaporation rate, and quantify the latent heat absorbed during the water phase change process; Based on circadian rhythms and seasonal fluctuations, define time-varying functions of ambient temperature, humidity, and wind speed; Establish a dynamic input model of cooling medium temperature, flow rate and heating steam pressure; Set the initial distribution of fermentation liquid temperature, bacterial concentration, and substrate concentration during inoculation.

[0009] Compared with the prior art, the present invention has the following beneficial effects: This system solves the core pain points of uncontrollable fermentation process, high energy consumption, high defective rate and delayed fault diagnosis in the baking industry through multimodal sensor fusion, spatiotemporal coupling modeling, nonlinear fermentation process control and cross-device collaborative optimization technology. It is suitable for large-scale industrial production scenarios, which not only improves product quality and production efficiency, but also promotes the transformation of the baking industry towards intelligence, greening and standardization, providing key technical support for the upgrading of the baking industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a flow chart of the dough humidity and temperature dynamic control system based on multi-modal sensor fusion of the present invention; Figure 2Flow chart for realizing non-blind area monitoring of three-dimensional temperature and humidity fields in the fermentation chamber of the present invention; Figure 3 Flow chart for capturing the spatio-temporal coupling relationship of temperature and humidity - carbon dioxide concentration - dough volume of the present invention; Figure 4 Flow chart for dynamically compressing the fermentation time based on non-linear fermentation process control of the present invention; Figure 5 Flow chart for designing an energy consumption optimization strategy based on digital twin of the present invention; Figure 6 Flow chart for constructing a hierarchical diagnostic tree containing fault characteristics of the present invention; Figure 7 Flow chart for performing microbial activity-enhanced humidity regulation of the present invention. Detailed implementation mode

[0011] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.

[0012] Referring to Figure 1 As shown, the dynamic regulation system for dough humidity and temperature based on multi-modal sensor fusion includes: A parameter monitoring module, which is used to adopt a distributed optical fiber sensing network to realize non-blind area monitoring of the three-dimensional temperature and humidity fields in the fermentation chamber and eliminate the risk of local failure of point sensors; An environment regulation module, which is used to introduce a deep learning architecture into the fermentation environment regulation and capture the spatio-temporal coupling relationship of temperature and humidity - carbon dioxide concentration - dough volume through a self-attention mechanism; A time compression module, which is used to construct a differential equation system containing at least one state variable, quantify the influence of temperature, humidity, and pH value on the yeast gas production rate, and dynamically compress the fermentation time based on non-linear fermentation process control; A timing optimization module, which is used to establish a thermodynamic model of the fermentation equipment, optimize the refrigeration and heating timings through reinforcement learning, and design an energy consumption optimization strategy based on digital twin; A collaborative control module, which is used to develop a communication protocol interface to realize the full-process parameter linkage of the mixer - fermentation chamber - proofing box - oven and perform cross-device collaborative control; A fault identification module, which is used to construct a hierarchical diagnostic tree containing fault characteristics and perform pattern recognition in combination with a deep belief network, and the fault characteristics include sensor-level, device-level, and process-level characteristics; Gradient control module, which is used for microbial activity-enhanced humidity regulation, and forms a protective layer on the dough surface based on the humidity gradient control algorithm of the water molecule diffusion coefficient.

[0013] As shown in Figure 2 The parameter monitoring module specifically includes: Based on the actual size and shape of the fermentation chamber, draw a three-dimensional space model, mark the monitoring area, and determine the accuracy requirements for temperature and humidity monitoring; Select an optical fiber sensor based on Raman scattering as the sensor for temperature and humidity monitoring; Based on the three-dimensional space model of the fermentation chamber, design the layout of the optical fiber sensing network, and ensure that the optical fiber can evenly cover all monitoring areas by controlling the direction, bending radius and fixing method of the optical fiber; For large fermentation chambers, at least one layer of optical fibers is arranged in parallel; Integrate the distributed optical fiber sensor with hardware devices such as a data acquisition card and a signal conditioning circuit.

[0014] Add spiral optical fibers around the stirring shaft to monitor the local temperature and humidity fluctuations generated by stirring, densely arrange them at the air inlet / outlet to capture the influence of air flow on temperature and humidity, fix the optical fibers with high-temperature resistant tapes or cable ties to avoid direct contact with the heating pipe and the stirring shaft, set optical fiber guiding grooves at the four corners of the box body to prevent the optical fibers from shifting due to vibration, wrap the outer layer of the optical fibers with stainless steel hoses to resist dough extrusion or mechanical collision, and install a dust filter at the air inlet to avoid flour adhesion affecting the scattering signal As shown in Figure 3 The environment regulation module specifically includes: Record the temperature and humidity, carbon dioxide concentration and dough volume data at fixed time intervals to form a time series data set; Normalize the temperature and humidity, carbon dioxide concentration and dough volume data to the [0,1] interval to eliminate the influence of dimension and accelerate the convergence of the model; Select the long short-term memory network as the deep learning architecture for processing time series data; For the input of each time step of temperature and humidity, carbon dioxide concentration and dough volume, generate query, key and value vectors through linear transformation; Calculate the dot product of the query vector and all key vectors, obtain the attention weights through normalization, and then perform weighted summation on the value vectors to obtain the output of the current time step; Adopt the multi-head self-attention mechanism to map the input vector to at least one subspace, calculate the self-attention respectively, and then splice the results to enhance the expression ability of the model; The encoder-decoder structure is adopted. The encoder is responsible for processing the input sequence, and the decoder generates the dough volume prediction value at the current time step based on the encoder output and the dough volume prediction value before the current time step. The mean square error function is selected as the loss function to measure the difference between the predicted dough volume and the true dough volume; The self-attention mechanism automatically captures the spatiotemporal coupling relationship between temperature and humidity, carbon dioxide concentration, and dough volume by calculating the attention weights between different time steps; Use the trained model to make real-time predictions of temperature, humidity, and carbon dioxide concentration during the fermentation process, and adjust the fermentation environment parameters based on the prediction results; The model prediction results are compared with the actual measured values ​​to form a closed-loop control system to achieve precise regulation of the fermentation environment.

[0015] The mean square error function is: , In the formula, is the mean square error function, is the total number of samples, is the predicted value of the i-th sample, is the true value of the i-th sample.

[0016] Reference Figure 4 As shown, the time compression module specifically includes: Based on the nonlinear expression of gas production rate, a dynamic model of yeast gas production rate was constructed; Set up the coupling mechanism between temperature-humidity-pH state variables; A three-variable coupling equation group is established based on mass conservation and energy conservation, and a real-time optimal parameter value is calculated, wherein the three-variable coupling equation group includes a bacterial growth equation, a substrate consumption equation, a temperature change equation, a pH dynamic equation, and a humidity dynamic equation; Real-time temperature regulation through jacket steam and cooling water flow; Based on the Arrhenius acceleration effect, the metabolic rate is increased exponentially by increasing the temperature, and a dynamic balance is established between enzyme activity and stability; Automatically add alkali or acid solution, and adjust the real-time pH value by controlling the flow rate through a peristaltic pump; Based on the pH buffer synergistic effect, near the optimal pH, the enzyme activity and substrate absorption efficiency are synergistically improved, accelerating the metabolic process; The humidification amount is controlled by the atomizing nozzle on the top of the tank, and the real-time humidity is adjusted in conjunction with the ventilation amount.

[0017] Coupling relationship modeling: Temperature - metabolic acid production: An increase in temperature accelerates glycolysis, leading to the accumulation of lactic acid or acetic acid and a linear relationship between the pH decline rate and temperature; Humidity - carbon dioxide partial pressure: An increase in humidity reduces the evaporation rate on the dough surface, resulting in an increase in the CO2 partial pressure in the tank, which in turn affects the yeast respiration rate; pH - substrate absorption: When the pH deviates from the optimal value, such as pH < 4.5, the plasma membrane permeability decreases and the glucose absorption efficiency reduces. The three - variable coupling equations are as follows: , In the formula, is the yeast cell concentration, is the time, is the maximum specific growth rate, is the function of temperature on the growth rate, is the function of pH on the growth rate, is the function of humidity on the growth rate, is the substrate concentration, is the half - saturation constant, taking 0.5, is the cell death rate constant, taking 0.01, is the cell yield coefficient, is the product yield coefficient, is the product formation rate constant, taking 0.02, is the temperature in the fermenter, is the steam flow rate, is the cooling water flow rate, is the metabolic heat production, is the density of the fermentation broth, is the volume of the fermenter, is the specific heat capacity of the fermentation broth, are the temperatures of steam and cooling water respectively, is the real - time pH value, is the amount of alkali solution added, is the amount of acid solution added, is the metabolic acid production, is the buffering capacity, are the pH values of the alkali solution and acid solution respectively, is the humidity in the fermenter, is the humidification amount, is the ventilation rate, is the air density, is the air volume in the fermenter, is the specific heat capacity of air, are the humidification and exhaust humidities respectively.

[0018] Referring to Figure 5 as shown, the timing optimization module specifically includes: Establish a three-dimensional unsteady-state heat balance equation for the temperature field inside the fermenter and set the energy interaction term; Set the metabolic heat generation term, quantify the exponential correlation between the bacterial metabolism rate and temperature, and introduce a heat generation coefficient to characterize the joule heat generated per unit bacterial biomass; Set the heat conduction term, use Fourier's law to describe the radial heat conduction of the tank wall, and establish a functional relationship between the thermal conductivity of the composite material and temperature; Set the convective heat transfer term, and based on Newton's cooling law, construct a dynamic correlation model for the turbulent convective heat transfer coefficient of the fermentation broth with flow velocity and temperature gradient; Set the radiative heat transfer term, and based on the radiative heat exchange between the tank surface and the environment, introduce a surface emissivity correction factor; Set the latent heat of evaporation term, establish a relationship model between the humidity gradient and the evaporation rate, and quantify the latent heat absorbed during the water phase change process; Based on circadian rhythm and seasonal fluctuations, define the time-varying functions of environmental temperature, humidity, and wind speed; Establish a dynamic input model for the temperature, flow rate of the cooling medium, and the pressure of the heating steam; Set the initial distributions of the fermentation broth temperature, bacterial concentration, and substrate concentration at the time of inoculation.

[0019] The three-dimensional unsteady-state heat balance equation is: , In the formula, is the density of the fermentation broth, is the specific heat capacity, is the temperature, is the time, is the thermal conductivity of the composite material, is the metabolic heat generation term, is the heat conduction term, is the convective heat transfer term, is the radiative heat transfer term, is the latent heat of evaporation term.

[0020] Referring to Figure 6 as shown, the fault identification module specifically includes: Input the sensor data into the first layer of the hierarchical diagnostic tree and output the abnormal patterns of the vibration signal, where the abnormal patterns include impact pulses and harmonic distortion; Input the equipment status indicators into the second layer of the hierarchical diagnostic tree, and combined with the intermediate event logic in the fault tree, judge the equipment fault type, where the fault types include bearing faults and gear wear; Input the process parameter characteristics into the third layer of the hierarchical diagnostic tree, and based on the process mechanism model, evaluate the impact of the fault on the process target; The top-level deep belief network fuses multi-modal features and outputs the final diagnosis result.

[0021] The fault identification module adopts a hybrid architecture of hierarchical diagnostic tree + deep belief network. Through multi-level feature extraction and logical reasoning, it realizes the full-link diagnosis of faults from sensor signals to process impacts. The module is divided into four layers: The first layer (sensor level): recognition of abnormal patterns of original signals; The second layer (equipment level): judgment of equipment fault types; The third layer (process level): assessment of the impact of faults on process objectives; The top layer (fusion layer): multi-modal feature fusion and final diagnosis.

[0022] Refer to Figure 7 As shown, the gradient control module specifically includes: By controlling the diffusion rate of water molecules on the dough surface and the humidity gradient distribution, the water exchange rate between the dough epidermis and the interior is dynamically adjusted during the fermentation process to form a low-permeability epidermal layer with a microbial protection function. Compare the self-diffusion coefficients of water molecules in the dough before and after fermentation to verify the inhibitory effect of the protective layer on water molecule penetration. Observe the thickness and density of the epidermal layer, with the goal of forming a continuous and pore-free hydrated layer. Determine the total number of microorganisms in the epidermal layer by the plate counting method to ensure that the concentration of the bacteriostatic agent effectively inhibits miscellaneous bacteria.

[0023] In the high-humidity stage, the starch granules absorb water and swell to form a viscoelastic matrix in the epidermal layer. In the stage of decreasing humidity gradient, the water in the epidermal layer slowly loses, and the starch molecular chains rearrange to form a continuous and pore-free dense structure with a porosity < 5%.

[0024] Furthermore, this solution also proposes a computer-readable storage medium, on which a computer-readable program is stored. When the computer-readable program is called, it executes the above-mentioned dynamic regulation system for dough humidity and temperature based on multi-modal sensor fusion.

[0025] It can be understood that the storage medium can be a magnetic medium, such as a floppy disk, a hard disk, a magnetic tape; an optical medium, such as a DVD; or a semiconductor medium, such as a solid-state disk (SSD), etc.

[0026] In summary, the advantages of the present invention are as follows: Through multi-modal sensor fusion, spatio-temporal coupling modeling, non-linear fermentation process control, and cross-device collaborative optimization technologies, this system solves the core pain points of uncontrollable fermentation process, high energy consumption, high defective product rate, and lagging fault diagnosis in the baking industry. It is applicable to large-scale industrial production scenarios, not only improving product quality and production efficiency, but also promoting the transformation of the baking industry towards intelligence, greenness, and standardization, providing key technical support for the upgrading of the baking industry.

[0027] The basic principles, main features and advantages of the present invention have been shown and described above. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, various changes and improvements will occur to the present invention, and all these changes and improvements fall within the scope of the present invention claimed. The scope of protection required by the present invention is defined by the appended claims and their equivalents.

Claims

1. A dough humidity and temperature dynamic control system based on multimodal sensor fusion, characterized in that: include: A parameter monitoring module, which is used to use a distributed optical fiber sensor network to achieve blind-spot monitoring of the three-dimensional temperature and humidity field in the fermentation box, eliminating the risk of local failure of point sensors; An environmental control module, which is used to introduce a deep learning architecture into fermentation environment control and capture the spatiotemporal coupling relationship of temperature and humidity, carbon dioxide concentration, and dough volume through a self-attention mechanism; A time compression module, wherein the time compression module is used to construct a differential equation group including at least one state variable, quantify the influence of temperature, humidity and pH value on the yeast gas production rate, and dynamically compress the fermentation time based on nonlinear fermentation process control; A timing optimization module, which is used to establish a thermodynamic model of the fermentation equipment, optimize the cooling and heating timings through reinforcement learning, and design an energy consumption optimization strategy based on digital twins; A collaborative control module, which is used to develop a communication protocol interface to achieve full-process parameter linkage of a mixer-fermentation box-proofing box-oven and perform cross-device collaborative control; A fault identification module, which is used to construct a hierarchical diagnosis tree containing fault features and perform pattern recognition in combination with a deep belief network. The fault features include sensor-level, device-level, and process-level features; A gradient control module is used to perform microbial activity-enhanced humidity control, and a humidity gradient control algorithm based on the diffusion coefficient of water molecules is used to form a protective layer on the dough surface.

2. The dough humidity and temperature dynamic control system based on multimodal sensor fusion according to claim 1, characterized in that: The parameter monitoring module specifically includes: Based on the actual size and shape of the fermentation box, draw a three-dimensional space model, mark the monitoring area, and determine the accuracy requirements for temperature and humidity monitoring; Select the optical fiber sensor based on Raman scattering as the sensor for temperature and humidity monitoring; Based on the three-dimensional spatial model of the fermentation box, the layout of the optical fiber sensor network is designed. By controlling the direction, bending radius and fixing method of the optical fiber, it is ensured that the optical fiber can evenly cover all monitoring areas. For large fermentation tanks, at least one layer of optical fibers is arranged in parallel; Integrate distributed fiber optic sensors with data acquisition cards and signal conditioning circuit hardware equipment.

3. The dough humidity and temperature dynamic control system based on multimodal sensor fusion according to claim 2, characterized in that: The environmental control module specifically includes: The temperature and humidity, carbon dioxide concentration, and dough volume data were recorded at fixed time intervals to form a time series data set; Normalize the temperature and humidity, carbon dioxide concentration, and dough volume data to the [0,1] interval to eliminate the dimension effect and accelerate model convergence; Choose Long Short-Term Memory Network as the deep learning architecture for processing time series data; For the input of temperature and humidity, carbon dioxide concentration, and dough volume at each time step, query, key, and value vectors are generated through linear transformation; Calculate the dot product of the query vector and all key vectors, obtain the attention weight by normalization, and then perform weighted summation on the value vector to obtain the output of the current time step; A multi-head self-attention mechanism is used to map the input vector to at least one subspace, calculate the self-attention separately, and then concatenate the results to enhance the expressiveness of the model; The encoder-decoder structure is adopted. The encoder is responsible for processing the input sequence, and the decoder generates the dough volume prediction value at the current time step based on the encoder output and the dough volume prediction value before the current time step. The mean square error function is selected as the loss function to measure the difference between the predicted dough volume and the true dough volume; The self-attention mechanism automatically captures the spatiotemporal coupling relationship between temperature and humidity, carbon dioxide concentration, and dough volume by calculating the attention weights between different time steps; Use the trained model to make real-time predictions of temperature, humidity, and carbon dioxide concentration during the fermentation process, and adjust the fermentation environment parameters based on the prediction results; The model prediction results are compared with the actual measured values ​​to form a closed-loop control system to achieve precise regulation of the fermentation environment.

4. The dough humidity and temperature dynamic control system based on multimodal sensor fusion according to claim 3 is characterized in that: The time compression module specifically includes: Based on the nonlinear expression of gas production rate, a dynamic model of yeast gas production rate was constructed; Set up the coupling mechanism between temperature-humidity-pH state variables; A three-variable coupling equation group is established based on mass conservation and energy conservation, and a real-time optimal parameter value is calculated, wherein the three-variable coupling equation group includes a bacterial growth equation, a substrate consumption equation, a temperature change equation, a pH dynamic equation, and a humidity dynamic equation; Real-time temperature regulation through jacket steam and cooling water flow; Based on the Arrhenius acceleration effect, the metabolic rate is increased exponentially by increasing the temperature, and a dynamic balance is established between enzyme activity and stability; Automatically add alkali or acid solution, and adjust the real-time pH value by controlling the flow rate through a peristaltic pump; Based on the pH buffer synergistic effect, near the optimal pH, the enzyme activity and substrate absorption efficiency are synergistically improved, accelerating the metabolic process; The humidification amount is controlled by the atomizing nozzle on the top of the tank, and the real-time humidity is adjusted in conjunction with the ventilation amount.

5. The dough humidity and temperature dynamic control system based on multimodal sensor fusion according to claim 4 is characterized in that: The timing optimization module specifically includes: Establish a three-dimensional non-steady-state heat balance equation for the temperature field in the fermentation tank and set the energy interaction term; Set the metabolic heat production term to quantify the exponential correlation between bacterial metabolic rate and temperature, and introduce the heat production coefficient to characterize the Joule heat generated per unit bacterial biomass; Set the heat conduction term, use Fourier's law to describe the radial heat conduction of the tank wall, and establish the functional relationship between the thermal conductivity of the composite material and the temperature; Set the convective heat transfer term and build a dynamic correlation model between the turbulent convective heat transfer coefficient of the fermentation liquid and the flow rate and temperature gradient based on Newton's cooling law; Set the radiation heat transfer term, introduce the surface emissivity correction factor based on the radiation heat exchange between the tank surface and the environment; Set the latent heat of evaporation term, establish the relationship model between humidity gradient and evaporation rate, and quantify the latent heat absorbed during the water phase change process; Based on circadian rhythms and seasonal fluctuations, define time-varying functions of ambient temperature, humidity, and wind speed; Establish a dynamic input model of cooling medium temperature, flow rate and heating steam pressure; Set the initial distribution of fermentation liquid temperature, bacterial concentration, and substrate concentration during inoculation.

6. The dough humidity and temperature dynamic control system based on multimodal sensor fusion according to claim 5, characterized in that: The fault identification module specifically includes: Inputting sensor data into a first layer of a hierarchical diagnostic tree, and outputting abnormal patterns of vibration signals, wherein the abnormal patterns include shock pulses and harmonic distortion; Input the equipment status index into the second-level hierarchical diagnosis tree, and combine the intermediate event logic in the fault tree to determine the equipment fault type, which includes bearing fault and gear wear; Input the process parameter characteristics into the third-level hierarchical diagnosis tree to evaluate the impact of the fault on the process target based on the process mechanism model; The top-level deep belief network fuses multimodal features and outputs the final diagnosis result.

7. The dough humidity and temperature dynamic control system based on multimodal sensor fusion according to claim 6, characterized in that: The gradient control module specifically includes: By controlling the diffusion rate of water molecules on the dough surface and the distribution of humidity gradient, the water exchange rate between the dough surface and the interior is dynamically adjusted during the fermentation process to form a low-permeability surface layer with microbial protection function. Compare the self-diffusion coefficient of water molecules in the dough before and after fermentation to verify the inhibitory effect of the protective layer on the penetration of water molecules; Observe the thickness and density of the epidermis, with the goal of forming a continuous, pore-free hydrated layer; The total number of microorganisms in the epidermis was determined by the plate counting method to ensure that the concentration of the antibacterial agent was effective in inhibiting miscellaneous bacteria.

Images