Meat product sterilization and cooking process quality stability control production system

By monitoring and dynamically controlling the thermophysical state of meat products in real time, the problems of uneven heat distribution and dielectric property fluctuations during the sterilization and cooking process of meat products have been solved, achieving a balance between internal sterilization and surface quality, and ensuring product stability and consistency.

CN123113773APending Publication Date: 2026-11-17NORTHEAST AGRICULTURAL UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611472900.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-21
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In the current meat product sterilization and cooking process, it is impossible to obtain the internal thermophysical state in real time, resulting in uneven heat distribution, which affects the stability of product quality. Furthermore, the dielectric characteristics fluctuation during radio frequency heating leads to impedance mismatch, making it impossible to simultaneously guarantee internal sterilization and surface quality.

Method used

The parameter configuration module receives material specification parameters, and the data acquisition module monitors temperature, pressure, radio frequency reflection coefficient and condensate flow rate in real time. The state estimation module calculates the actual water loss and thermal diffusivity, the prediction and judgment module determines the core lethal value and surface damage, the intervention and control module forms a steam insulation layer and adjusts the radio frequency power, and the model calibration module optimizes the calculation deviation, thus realizing non-destructive measurement and dynamic control.

Benefits of technology

It enables non-destructive real-time monitoring of the internal thermophysical state of meat products, ensuring central sterilization and avoiding surface overheating, thus guaranteeing the stability and consistency of product quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113773A_ABST
    Figure CN123113773A_ABST
Patent Text Reader

Abstract

This invention relates to the field of food processing technology and discloses a production system for quality stability control during the sterilization and cooking process of meat products. The system includes a data acquisition module that dynamically acquires absolute temperature, absolute pressure, reflection coefficient, and instantaneous conductivity of condensate; a state estimation module that calculates the actual water loss based on the differential conductivity signal and updates the actual thermal diffusivity by combining the first-order time-domain derivative of the complex reflection coefficient; a prediction and judgment module that forward-predicts the internal lethal trajectory and the surface quality deterioration trend; and an intervention interruption signal that is sent when conditions are met. An intervention control module receives the interruption signal and executes a nonlinear pulsed stepwise pressure reduction to form a steam insulation layer on the surface. This drives an impedance matching unit to reconstruct the phase angle and apply radio frequency heating power to the core area. A model calibration module extracts the convergence integral deviation and optimizes the deviation weights. This invention resolves the contradiction between surface quality deterioration and failure to achieve sterilization targets in the core of meat products, improving the accuracy of internal thermal state estimation and quality stability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of food processing technology, specifically to a production system for quality stability control during the sterilization and cooking process of meat products. Background Technology

[0002] In the sterilization and cooking process of meat products, the basic requirement is to ensure that the center of the product reaches the safety standard for microbial inactivation, while maintaining the sensory quality of the surface as much as possible and reducing juice loss. Existing processing equipment sets fixed temperature and time process curves based on historical experience, or measures the temperature by inserting a metal probe into the product. However, the probe measurement method damages the physical integrity of the product's soft packaging and cannot be widely used in large-scale continuous production. Furthermore, the fixed process curve cannot obtain the real moisture loss and heat transfer state changes inside the material during processing in real time, and the thermodynamic state estimated by the system will deviate from the actual physical state.

[0003] Because it is impossible to obtain the true thermophysical state inside meat products in real time, traditional control systems face the problem of uneven heating when executing the heating program. In order to ensure that the internal cold point reaches the sterilization safety target, the system will continuously output heat or extend the heating time. This causes the surface of the meat products to be overheated, resulting in quality deterioration. Some equipment uses radio frequency heating technology to improve penetration, but when the meat products are heated and moisture migrates and juices are released, the dielectric properties of the material will fluctuate. The change in dielectric properties will cause the input impedance mismatch of the radio frequency system, generating a large amount of reactive reflected power, which will prevent the radio frequency system from maintaining energy injection into the core area. In addition, the internal composition of different batches of meat products is different. Conventional control models lack a post-feedback correction mechanism for multiple batches of actual operating data. The calculation error generated by the state equation in long-term operation is difficult to eliminate, resulting in a lack of stability in the quality of the final batches of meat products. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a production system for quality stability control during the sterilization and cooking process of meat products. This system solves the problems in existing technologies, such as the inability to non-destructively monitor the internal thermophysical state of meat products and the difficulty in simultaneously achieving core sterilization and surface quality due to radio frequency impedance mismatch caused by material phase transitions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a production system for quality stability control during the sterilization and cooking process of meat products. The system includes: The parameter configuration module is used to receive the material specification parameters of the meat products to be processed, and to establish the microbial inactivation safety target value and the surface quality damage equivalent threshold of the meat products according to the set calculation criteria. The data acquisition module is used to dynamically acquire the absolute temperature, absolute pressure, reflection coefficient and phase angle of the RF input matching terminal in the heating chamber, as well as the instantaneous mass flow rate and conductivity of the discharged condensate. The state estimation module is used to calculate the actual water loss of meat products during the processing stage based on the differential removal of interference from the basic condensate fluid using the real-time conductive signal of condensate. It updates the actual thermal diffusivity by combining the first derivative of the complex reflection coefficient in the time domain and iteratively outputs the current temperature, cumulative lethal value, and cumulative damage value of the core cold point inside the meat product. The prediction and judgment module is used to calculate the internal core lethal trajectory and surface quality deterioration trend within the forward prediction window from the current real-time time step starting point, and send an over-limit intervention interruption signal to the system main control bus when it is determined that the cumulative lethal value is insufficient to the safety target value and the cumulative surface damage will reach the equivalent threshold. The intervention and control module is used to receive the over-limit intervention interruption signal, perform nonlinear pulse step-down to flash evaporation to form a vapor insulation layer on the packaging surface, drive the impedance matching unit to reconstruct the phase angle, and continuously apply frequency-modulated radio frequency heating power to the core area. The model calibration module is used to extract the convergence integral deviation of the current batch processing and adaptively optimize the deviation weights for calculating the state equation.

[0006] By differentially extracting fluid electrical signals and high-frequency reflection characteristics, electromagnetic fluctuations and water loss fluctuations caused by phase change in meat products are transformed into real-time thermophysical parameters, enabling thermophysical state estimation under probe-free conditions. The prediction and judgment module performs forward prediction with thermal diffusion parameters as boundary conditions, actively interrupting the conventional temperature control loop before surface quality deteriorates. The intervention and control module uses a pressure reduction flash vaporization mechanism to form a physical insulation layer on the outside of the packaging and simultaneously adjusts the spatial distribution of the radio frequency field to inject heat into the center. Finally, the model calibration module updates the deviation weights to achieve closed-loop correction of control parameters.

[0007] Furthermore, the parameter configuration module contains an initial physical property mapping model. The initial physical property mapping model takes the internal fat mass ratio and inorganic salt mass fraction of the offline collected physical samples as input feature variables, and takes the true value of the initial thermal diffusivity coefficient of the corresponding sample and the true value of the characteristic conductivity of the pure precipitated juice as output labels. It is obtained by iterative training using a multilayer perceptron or support vector regression algorithm until the verification error converges.

[0008] A further approach involves pre-establishing a correlation model between characteristic variables and basic physical property parameters to provide a calculation benchmark for online parameter estimation and reduce the deviation of initial values ​​set on-site.

[0009] Furthermore, the data acquisition module is specifically used to: acquire the absolute temperature inside the heating cavity using platinum resistance temperature sensors distributed inside the cavity, and acquire the absolute pressure inside the heating cavity using a pressure transmitter; extract electromagnetic data on the same nanosecond time segment using high-frequency clock phase-locked loop technology based on field-programmable gate array, and calculate the reflection coefficient and phase angle of the RF input matching terminal; drive the measuring instrument to perform continuous synchronous measurement, acquire the instantaneous mass flow rate and conductivity of the discharged condensate, and encapsulate the parameters in a real-time data frame containing a timestamp.

[0010] A further approach is adopted, utilizing the synchronous clock alignment mechanism of multi-channel sensors to ensure the synchronization of high-frequency electromagnetic parameters with thermodynamic and fluid dynamic parameters at the same time cross section, thereby improving the data foundation accuracy of calculus calculations.

[0011] Furthermore, when the state estimation module performs differential calculation of the actual water loss separated from the condensate, it sets the pure condensate generated by the phase change of pure water vapor as a zero conductivity base, and then removes the interference of the basic condensate fluid based on the instantaneous conductivity signal of the condensate. At the same time, it introduces the characteristic conductivity coefficient to establish a mass conservation differential relationship based on the fluid mixing law, thereby calculating the actual water loss of the meat product during the processing stage as water loss data.

[0012] A further approach is adopted, utilizing the physical properties of pure steam condensate insulation and the presence of ions in the juices precipitated from meat products, combined with the law of fluid mixing, to calculate and separate the actual water loss of meat products precipitated by heat in real time.

[0013] Furthermore, when updating the true thermal diffusivity, the state estimation module substitutes the true water loss reflecting the shrinkage of the grid structure and the first derivative of the complex reflection coefficient reflecting electrophysical fluctuations in the time domain into the data-driven dynamic thermodynamic correction model; based on the initial calibrated thermal diffusivity constant of the material, it performs compensation calculations by combining the pre-fitted water loss coupling correction weight constant and the dielectric reflection coupling correction weight constant, calibrates in real time, and outputs the updated true thermal diffusivity.

[0014] A further approach is adopted to combine the shrinkage characteristics caused by the actual water loss with the disturbance characteristics caused by the change in the complex reflection coefficient, and to calculate and calibrate the thermal diffusivity through weight compensation, thereby reducing the impact of internal phase change on the calculation of thermal conductivity.

[0015] Furthermore, the prediction and determination module is specifically used to: import the current temperature, cumulative lethal value, and cumulative damage value of the product surface of the output internal core cold point into a multivariate model prediction controller based on the TwinCAT control platform; set a forward rolling prediction time window, continuously solve the parameter changes based on the built-in finite difference calculation matrix to predict the internal core lethal trajectory; combine the reaction kinetic equation to forward extrapolate and calculate the surface quality deterioration trend, and establish a bivariate threshold determination rule to trigger the mode switching intervention logic.

[0016] A further approach is adopted, which calculates the state trend within future time steps through a forward prediction window, identifies states with insufficient central lethal value and surface deterioration risk, triggers control intervention, and solves the problem of system temperature control lag.

[0017] Furthermore, when the intervention and control module performs nonlinear pulsed stepwise depressurization, it causes the attached moisture and surface exudate on the surface of the meat product soft packaging to undergo asynchronous flash vaporization, cutting off the heat conduction channel from the external steam environment to the interior of the meat product to generate the steam insulation layer; it calculates the cumulative value of the reflection feature sliding integral by performing a sliding window integral on the first derivative of the complex reflection coefficient in the time domain; when the cumulative value of the reflection feature sliding integral crosses the set coverage tolerance threshold, it verifies the thermal isolation state of the entire system based on the change of the integral of the time domain derivative of the radio frequency reflection.

[0018] A further approach is adopted, which utilizes the mechanism that the liquid boiling point decreases due to the decrease in gas pressure to cause water to flash vaporize and generate a vapor insulation layer with low dielectric constant and high thermal resistance characteristics, thereby cutting off the external heat conduction channel. At the same time, the integral value of the reflection coefficient is used to verify whether the thermal insulation state is valid.

[0019] Furthermore, when the intervention control module drives the impedance matching unit to perform phase angle reconstruction and apply power: after verifying that the thermal isolation state is established, it outputs a non-heating stage sweep frequency signal to calculate the reactance deviation value between the current impedance mismatch and the standard matching point; according to the reactance deviation value, it drives the impedance matching unit to eliminate the reactive reflected power introduced by the air cavity thermal insulation barrier to complete the phase angle reconstruction; after completing the phase angle reconstruction, it increases the radio frequency output power and introduces a sinusoidal frequency modulation mechanism to continuously apply frequency-modulated radio frequency heating power to the core area.

[0020] Further measures were taken to eliminate the reactive reflected power introduced by the steam insulation layer and to use a frequency modulation mechanism to change the distribution of standing wave nodes during radio frequency injection, so as to prevent arc breakdown at the dielectric interface and ensure stable injection of radio frequency power into the core area of ​​meat products.

[0021] Furthermore, when extracting the convergence integral deviation of batch processing, the model calibration module: controls the fluid loop to switch to the pressure-maintaining spray cooling stage; calculates the algebraic difference between the actual cumulative measured value of conductivity and the theoretical predicted value of conductivity for the current batch processing to obtain the global conductivity integral error; extracts the actual time span from the moment the pressure reduction command is issued to the cavity pressure reaching the target low pressure as the pressure reduction response time constant; and integrates the global conductivity integral error and the pressure reduction response time constant into a state observation vector by matrix splicing, and directly assigns it as the convergence integral deviation.

[0022] A further approach is adopted, which combines the fluid conductivity error with the pressure reduction response time constant of the pressure regulating fluid loop to form a state observation vector, providing observation data containing the time constant for model calibration.

[0023] Furthermore, when the model calibration module adaptively optimizes the deviation weights of the calculated state equation, it extracts the state observation vector as a supervision error signal and substitutes it into the Kalman observation feedback equation as the true measurement value of the Kalman filter; it dynamically solves and updates the optimal Kalman gain matrix by minimizing the mean square error criterion of the posterior state estimation; and it injects the updated optimal Kalman gain matrix as a dynamic learning rate into the weight update formula to drive the posterior closed-loop iterative update of the deviation weights of the calculated state equation.

[0024] A further approach is adopted, which uses the Kalman filter algorithm to handle convergence integral deviations and optimizes the deviation weights by iteratively updating the optimal Kalman gain matrix, thereby improving the calculation accuracy of the state equation in multi-batch production.

[0025] This invention provides a production system for stabilizing and controlling the quality of meat products during the sterilization and cooking process. It offers the following advantages: 1. This invention calculates the actual water loss by differentially removing interference from the basic condensate fluid based on the real-time conductivity signal of condensate, and updates the actual thermal diffusivity by combining the first derivative of the complex reflection coefficient in the time domain. By utilizing the correlation between the conductivity characteristics of condensate and the physical changes of radio frequency reflection parameters, it is possible to obtain real-time data on water loss and heat transfer parameters inside meat products without damaging the packaging or inserting probes. This solves the problem that the internal thermodynamic state cannot be directly measured in traditional processing, and improves the accuracy of processing state estimation.

[0026] 2. When the cumulative lethal value inside the meat is predicted to be insufficient and the cumulative damage on the surface will reach the equivalent threshold, the present invention performs a nonlinear pulse stepwise depressurization to induce surface moisture to flash evaporate and form a steam insulation layer. After phase angle reconstruction, frequency-modulated radio frequency heating power is applied to the core area. The steam layer generated by the depressurization cuts off the heat conduction from the external environment to the interior of the meat product. At the same time, impedance matching and frequency modulation ensure that the radio frequency power is stably injected into the cold spots inside the meat product. Under the premise of ensuring that the microbial inactivation standard is met at the center, the invention avoids juice loss and quality deterioration of the meat product surface due to excessive heating.

[0027] 3. This invention extracts the global conductivity integral error and voltage reduction response time constant of the current batch processing and integrates them into a state observation vector. This vector is then used as the actual measured value of the Kalman filter and substituted into the observation feedback equation. The optimal Kalman gain matrix is ​​dynamically solved to update the deviation weights of the calculated state equation. Multidimensional posterior data is introduced into the control loop for closed-loop iteration, enabling the system to automatically correct the calculation deviation of the state equation based on the actual measurement status of multiple batches of production. This overcomes the cumulative error caused by the fixed process model when facing the differences in the thermal properties of materials from different batches, ensuring the stability of product quality during long-term production. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the system timing interaction of the present invention; Figure 2 This is a schematic diagram of the method flow of the present invention; Figure 3 This is a schematic diagram illustrating the structural steps of the present invention for establishing the safety target value for microbial inactivation of meat products and the equivalent threshold value for surface quality damage. Figure 4 This is a schematic diagram of the steps for starting the power supply unit and dynamically acquiring process parameters in the heating chamber according to the present invention. Figure 5 This is a schematic diagram of the steps in this invention to calculate the actual water loss, update the actual thermal diffusivity, and iteratively output the state results. Figure 6 This is a schematic diagram of the structural steps of forming a steam insulation layer, reconstructing the phase angle, and continuously applying thermal power according to the present invention. Figure 7 This is a quality stability control diagram for the sterilization and cooking process of meat products according to the present invention. Detailed Implementation

[0029] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] See attached document Figure 1 This invention provides a production system for quality stability control during the sterilization and cooking process of meat products. The system may include an automated control cabinet deployed with a Siemens TIA Portal system, and the automated control cabinet integrates: The parameter configuration module is used to receive the material specification characteristic data of the meat products to be processed, and to establish the cumulative lethality safety target value and the upper limit of quality damage tolerance during the heating process based on the material specification characteristic data. The range of the cumulative lethality safety target value is limited to 3.0 to 6.0 min, and the range of the upper limit of quality damage tolerance is limited to 3% to 5% of the water loss equivalent. The data acquisition module is used to synchronously acquire environmental state parameters, electrical reflection parameters of the RF load matching network, and the mass and conductivity characteristics of the condensate drain at the bottom of the containment cavity according to a set time period while maintaining the heating and sterilization states. The set time period is limited to a range of 100 to 500 ms. The data acquisition module uses the NI industrial data acquisition platform. The state estimation module is used to separate the actual water loss of meat products due to thermal denaturation from the total condensate data through discrete difference calculation. It updates the dynamic thermal diffusivity of meat products by combining the time-domain derivative characteristics of electrical reflection parameters, and outputs the cumulative sterilization value of the cold point at the geometric center of the meat product and the degree of damage to the product surface quality based on finite difference calculation. The prediction and judgment module is used to predict the cumulative sterilization value trajectory and quality damage trajectory within a set time window based on the dynamic thermal diffusivity and the current internal state. When the cumulative sterilization value at the cold point does not meet the safety target value and the predicted surface damage reaches the upper limit of quality damage tolerance, a mode switching intervention signal is generated. The intervention and control module is used to receive mode switching intervention signals, control the pressure relief channel to perform stepped pressure reduction to build a water vapor insulation layer on the packaging surface, and after verifying that the thermal resistance isolation state of the full load is established, control the automatic matching circuit to reconstruct the impedance phase angle, drive the radio frequency generation unit to output high-frequency alternating power with frequency modulation characteristics, and directly inject heat into the center of the meat product. The model calibration module is used to extract the system convergence error and heat transfer characteristic parameters during the entire working period of heating, sterilization and cooling after the cumulative sterilization value at the cold point reaches the standard and the cooling operation begins. It automatically corrects the deviation weight coefficient of the state equation in the phase change mapping calculation and saves the updated configuration model.

[0031] As a collaborative operation mode of the overall architecture, the parameter configuration module establishes the basic constraint range and outputs the basic constraint range to the data acquisition module and the state estimation module.

[0032] After sterilization is initiated, the data acquisition module continuously sends the measured pressure, temperature, reflection coefficient, and electrical signals of the fluid to the state estimation module. The state estimation module then translates the electrophysical fluctuations and water loss fluctuations caused by the phase change of the material into real-time thermophysical parameters, solving the problem of thermal state estimation under probe-less conditions.

[0033] The prediction and judgment module performs time-domain iteration using the thermal diffusion parameters output by the state estimation module as boundary conditions, monitors the relative margins of safety constraints and quality constraints, and once it is determined that surface quality deterioration is about to occur, the system control logic directly interrupts the basic temperature regulation loop and wakes up the intervention and control module.

[0034] The intervention and control module controls the pressure regulating mechanism and the generation network according to the phase transition law. The flash phase change on the material surface caused by the pressure reduction forms a physical thermal barrier for the system. At the same time, the resulting full electrical mutation is used as the trigger criterion for impedance phase angle reconstruction and radio frequency energy leap. The intervention and control module adjusts the spatial distribution of the high-frequency field and compensates for mismatch energy, and forcibly replenishes the central target heat.

[0035] After a single work batch is terminated, the model calibration module receives the processing records of the currently completed batch and optimizes the parameters of the estimation equation. It then feeds back the updated patterns to the parameter configuration module for use in the subsequent ripening control of goods of the same specifications.

[0036] See attached document Figure 2 This invention provides a production method for stabilizing and controlling the quality of meat products during the sterilization and cooking process. The method includes the following steps: S1 receives the material specifications of the meat products to be processed, and establishes the safety target value for microbial inactivation and the equivalent threshold value for surface quality damage of the meat products according to the set calculation criteria. S2, start the steam environment heating and high frequency electromagnetic power supply unit, and dynamically collect the absolute temperature, absolute pressure, reflection coefficient and phase angle of the radio frequency input matching terminal in the heating chamber according to a fixed clock cycle during the processing, as well as the instantaneous mass flow rate and conductivity of the discharged condensate. S3, based on the differential removal of the basic condensate fluid interference by the instantaneous conductive signal of condensate, calculates the actual water loss of meat products during the processing stage, updates the actual thermal diffusivity by using the time-domain calculus of the actual water loss and the radio frequency reflection characteristics, and iteratively outputs the current temperature, cumulative lethal value and cumulative damage value of the core cold point inside the meat product. S4: Calculate the internal core lethal trajectory and surface quality deterioration trend within the forward prediction window from the current real-time time step starting point. When it is determined that the cumulative lethal value is insufficient to the safety target value and the cumulative surface damage will reach the equivalent threshold, trigger the mode switching intervention logic. S5 performs nonlinear pulse step-down to flash vaporize and form a vapor insulation layer on the packaging surface. After verifying the thermal isolation state of the entire system based on the integral change of the time-domain derivative of the radio frequency reflection, it drives the impedance matching unit to reconstruct the phase angle and continuously applies frequency-modulated radio frequency heating power to the core area. S6, after the cumulative lethal value at the center reaches the safety target value, the heating is stopped, the fluid circuit is switched to the pressure-maintaining spray cooling stage, and finally the convergence integral deviation of the current batch is extracted to adaptively optimize the deviation weight of the calculated state equation.

[0037] See attached document Figure 3 In specific implementations, step S1 provided by the present invention may include the following steps: S101 receives the material specification parameters of the meat products to be processed from the Siemens touch operation terminal through the parameter configuration module, and analyzes and extracts the geometric dimensions, internal fat mass ratio, inorganic salt mass fraction and material type information of the batch material, combined with the empirical comparison of physical property parameters established in advance based on conventional orthogonal experimental statistics. The empirical reference table for physical property parameters is specifically embodied in the pre-trained offline initial physical property mapping model. The method for creating and training the initial physical property mapping model is as follows: During the R&D phase before system deployment, an offline training dataset is constructed. Researchers collect and prepare a large number of meat product samples containing different internal fat mass percentages and different inorganic salt mass fraction gradients. These samples are heated on a standard heating test bench, and the true values ​​of the initial thermal diffusivity coefficient and the true values ​​of the characteristic conductivity of the pure precipitated juice are measured using a precision thermal property analyzer and an online conductivity meter, respectively. Subsequently, using the internal fat mass percentage and inorganic salt mass fraction as input feature variables, and the true values ​​of the initial thermal diffusivity coefficient and the true values ​​of the characteristic conductivity as output labels, an initial physical property mapping model is established using a multilayer perceptron or support vector regression algorithm. The offline training dataset is divided into a training set and a validation set in an 8:2 ratio. Mean squared error is used as the loss function, and the model weights are iteratively trained using the Adam optimization algorithm and gradient descent method until the validation error on the validation set converges and falls below a set tolerance threshold, with the set tolerance threshold limited to a range of 10. -4 Up to 10 -3 The interval is then used to solidify and deploy the trained static inference model in the parameter configuration module. The extracted geometric dimensions are converted into mesh boundary parameters calculated by finite difference, and the internal fat mass percentage is converted into the thermal diffusivity coefficient value in the initial stage. At the same time, the inorganic salt mass fraction is mapped to the characteristic conductivity coefficient of the corresponding meat product characteristic juice, and the characteristic conductivity coefficient is defined as an empirical constant between 0.5 and 2.5 S / m.

[0038] S102, combining the characteristic strain type specified in the process requirements, and obtaining the standard heat resistance time constant of the target microorganism at the reference temperature based on existing thermal lethality theories, and according to the set initial logarithm of the total viable count and the logarithm of the allowable residual viable count, the absolute sterilization intensity boundary that the corresponding batch processing must meet is calculated through the first set calculation criteria. Combining the equivalent thermal sterilization time relationship, the absolute sterilization intensity boundary is defined as the heating time for exposing the product to standard constant temperature conditions. Based on the existing thermal standard calculation criteria as the first set calculation criteria, the calculation formula corresponding to the microbial inactivation safety target value is established as follows: ; In the formula, The target value for microbial inactivation of meat products; The baseline microbial lethal time at the reference temperature; The initial microbial load of the meat product to be processed; The target microbial residue level is defined by safety standards.

[0039] Each variable is pre-calibrated according to the existing microbial sampling and testing procedures and used as a steady-state input. The microbial inactivation safety target value is input into the prediction and judgment module as the basic comparison benchmark for the underlying hardware state switching, and the range of the microbial inactivation safety target value is set to 3.0 to 6.0 min.

[0040] S103, establish the reaction kinetic relationship between the thermal shrinkage and denaturation process of surface tissue proteins and the mass equivalent of juice outflow. Use the product weight loss ratio as the benchmark for measuring the deterioration of surface sensory quality. Based on the existing centrifugal drip loss test method, set an upper limit limit for the water loss rate, limiting the value of the upper limit limit to 3% to 5%. Then, through the existing reaction kinetic equation as the second set calculation criterion, convert the upper limit limit of the water loss rate into the integral upper limit value of the irreversible chemical reaction under the corresponding thermal process, establish the equivalent threshold of surface quality damage, and establish the limiting thermal integral formula corresponding to the equivalent threshold of surface quality damage as follows: ; In the formula, This is the threshold value for surface quality damage. This is the set maximum permissible processing time boundary; The reaction rate constant is at the absolute reference temperature; The activation energy for protein denaturation; It is the ideal gas constant; This is an absolute reference temperature. The absolute surface temperature of the product corresponding to the integration time variable; For integration time; It is a differential operator.

[0041] The calculated surface quality damage equivalent threshold is transmitted to the prediction and judgment module as a rigid boundary condition for decoupling intervention, and the value range of the surface quality damage equivalent threshold is limited to 1500 to 3500 dimensionless integral units.

[0042] See attached document Figure 4 In specific implementations, step S2 provided by the present invention may include the following steps: S201, through the system main control bus, synchronously sends start commands to the steam proportional regulating valve and the high-frequency electromagnetic power supply unit, then starts the steam environment heating and the high-frequency electromagnetic power supply unit. Combined with the pre-stored specific heat capacity calibration table for different meat products, the initial opening of the steam proportional regulating valve is set to perform a linear increase action according to the set heating slope. The set heating slope value range is 3 to 8℃ / min. At the same time, the solid-state radio frequency generator inside the high-frequency electromagnetic power supply unit is controlled to output the set initial base power. The initial base power is determined according to the total mass of the processing batch according to the ratio of 400 to 800W / kg. The high-frequency electromagnetic power supply unit uses the MKS solid-state radio frequency generator system. Then, the steam sensible heat convection and radio frequency volume penetration heat field inside the heating chamber are synchronously applied to the meat products to be processed on the time axis, establishing the basic thermophysical boundary conditions of the heating stage.

[0043] S202 dynamically acquires the absolute temperature, absolute pressure, and reflection coefficient and phase angle of the RF input matching terminal within the heating cavity during processing, according to a fixed clock cycle set to 100 to 500 ms. Specifically, it uses platinum resistance temperature sensors distributed inside the cavity to acquire the absolute temperature, and a pressure transmitter to acquire the absolute pressure. A bidirectional directional coupler is used to extract the forward and reflected wave voltage amplitudes and phase difference of the RF input matching terminal. Then, based on a field-programmable gate array (FPGA) high-frequency clock phase-locked loop (PLL) technology, electromagnetic data is extracted and aligned on the same nanosecond time segment. The reflection coefficient and phase angle of the RF input matching terminal are calculated based on existing microwave transmission line theory, and the complex number calculation formula corresponding to the reflection coefficient is established as follows: ; In the formula, The complex reflection coefficient is the current physical timestamp. The current physical timestamp; The reflected wave voltage amplitude is the value collected at the current physical timestamp; The incident wave voltage amplitude is the value collected at the current physical timestamp; It is a natural constant; The imaginary unit; The phase angle of the reflected wave acquired at the current physical timestamp; The incident wave phase angle is acquired at the current physical timestamp; finally, the separated and extracted reflection coefficient and phase angle, along with the absolute temperature and absolute pressure, are encapsulated in a real-time data frame containing the timestamp.

[0044] S203, under the same clock trigger pulse that performs multi-channel hardware interrupt sampling, drives the EndressHauser mass flow meter and the MettlerToledo online conductivity meter installed in the condensate pipe at the bottom of the heating chamber to perform continuous synchronous measurement. Combined with the electrical characteristic deviation data caused by electrolyte dissolution during heating, it dynamically collects the instantaneous mass flow rate and conductivity of the discharged condensate. Finally, it appends the instantaneous mass flow rate and conductivity to the end of the real-time data frame, completing the synchronous acquisition of environmental parameters and fluid electrical signals and transmitting them to the state estimation module as the basic calculation data source.

[0045] See attached document Figure 5 In specific implementations, step S3 provided by the present invention may include the following steps: S301, the instantaneous mass flow rate and conductivity of the condensate received by the state estimation module are extracted. Considering the high concentration of inorganic salt ions in the juices released from heated meat products and the near-insulating nature of pure steam condensate, the pure condensate generated by the phase change of pure water vapor is set as a zero-conductivity base. Then, interference from the base condensate fluid is removed based on the differential instantaneous conductivity signal of the condensate. Simultaneously, a characteristic conductivity coefficient is introduced to establish a mass conservation differential relationship based on the fluid mixing law. Thus, the actual water loss of the meat products during processing is calculated as the water loss data, and the integral difference formula corresponding to the actual water loss is established as follows: ; In the formula, This is the cumulative value of the actual water loss extracted from the current physical timestamp; The current physical timestamp; To determine the instantaneous conductivity of the discharged condensate corresponding to the integration time variable, the integration time step is set to be consistent with the fixed clock period of the hardware sampling. The instantaneous mass flow rate of discharged condensate corresponding to the integral time variable; The characteristic conductivity calibration constant for the pure precipitated juice; For integration time; It is a differential operator.

[0046] S302 updates the true thermal diffusivity coefficient using the time-domain calculus of the actual water loss and radio frequency reflection characteristics. It also introduces the physical mechanism that water loss in meat products leads to changes in macroscopic dielectric properties, which in turn causes high-frequency distortion of the radio frequency reflection signal. It extracts the complex reflection coefficient from the real-time data frame, calculates the first derivative of the complex reflection coefficient in the time domain, and uses it as the physical quantity of the dielectric constant mutation rate under the high-frequency electromagnetic field. Then, it substitutes the water loss data reflecting the shrinkage of the grid structure and the dielectric constant mutation rate reflecting electrophysical fluctuations into the preset thermophysical property empirical mapping equation. The thermophysical property empirical mapping equation is essentially a data-driven dynamic thermodynamic correction model. The preset thermophysical property empirical mapping equation is essentially a data-driven dynamic thermodynamic correction model. The method for constructing and training the dynamic thermodynamic correction model is as follows: a synchronous monitoring experimental platform integrating a microwave vector network analyzer and a multi-point fiber optic temperature probe is built in a laboratory environment; meat products with known initial parameters are placed on the synchronous monitoring experimental platform for multiple complete radio frequency heating cycle experiments; during the experiment, the actual dynamic water loss, the first derivative of the reflection coefficient in the time domain, and the real-time true thermal diffusivity calculated by the fiber optic probe and Fourier inverse operation are collected synchronously to form a parameter fitting training set; then, multiple linear regression or least squares method is used for model fitting training, with water loss data and the first derivative of the reflection coefficient as input independent variables, and the correction amount of the true thermal diffusivity coefficient as the output dependent variable. By minimizing the sum of squared residuals between the calculated model value and the true value calculated by the fiber optic probe, the globally optimal correction weight is solved, thereby obtaining the high-precision equation parameters of the fixed loaded system. Therefore, the internal nonlinear thermal diffusivity is calibrated in real time, and the formula for updating the true thermal diffusivity is established as follows: ; In the formula, The true thermal diffusivity updated to the current physical timestamp; The current physical timestamp; The initial thermal diffusivity of the material is determined; To correct the weighting constant for water loss coupling, the value range of the water loss coupling compensation coefficient is limited to 0.01 to 0.05g. -1 , is one of the optimal parameters output by the least squares model during training; To correct the dielectric reflection coupling weighting constant, the range of the electromagnetic disturbance thermal resistance correction coefficient is limited to 0.1 to 0.5 m. 2 / s is one of the optimal parameters output by the least squares model during training; This is the cumulative value of the actual water loss extracted from the current physical timestamp; The complex reflection coefficient modulus of the current physical timestamp; It is a differential operator.

[0047] S303 imports the updated true thermal diffusivity into the built-in three-dimensional thermal conduction finite difference calculation matrix and calls the previously updated adaptive calculation state equation deviation weights stored in the system controller's non-volatile storage module. These deviation weights are then multiplied into the finite difference evolution equation as a state compensation matrix to compensate for historical experience errors in the fluid heat transfer boundary and internal node conduction. Furthermore, the absolute temperature extracted from the real-time data frame within the heating cavity is used as the outer convection heat transfer boundary condition. The external convection temperature boundary and the continuously changing internal thermal diffusivity jointly drive the thermal evolution of discrete nodes. The temperature field distribution of the grid nodes is iteratively solved using the existing Fourier thermal conduction finite difference numerical solution method. Then, the temperature values ​​at the geometric center nodes of the spatial grid are extracted and time integration calculations are performed with a fixed step size. This establishes the integral relationship of the cumulative lethal value as follows: ; In the formula, The cumulative lethal value of the core cold spot inside the product, calculated based on the current physical timestamp; The current physical timestamp; The absolute temperature of the internal core cold spot corresponding to the integration time variable; This is an absolute reference temperature. The temperature sensitivity characteristic constant is defined as the thermostability constant of the target microorganism, and its value is taken from 8 to 12℃ based on the existing thermodynamic spectrum of the characteristic strains. This is the time variable for integration.

[0048] S304 synchronously extracts the temperature sequence of the outermost grid node of the meat product in the three-dimensional heat conduction finite difference calculation matrix as the surface dynamic absolute temperature variable, and uses the node spatial coordinate mapping method to track the physical surface. Then, based on the reaction kinetic principle that the cumulative effect of surface temperature positively promotes irreversible biochemical deterioration, the existing reaction kinetic equation is applied to transform the surface dynamic absolute temperature variable into the result of irreversible protein deterioration and calculate the cumulative damage value of the product surface. Finally, combined with the current temperature of the internal core cold point and the cumulative lethal value, the current temperature of the internal core cold point, the cumulative lethal value, and the cumulative damage value of the product surface are iteratively output.

[0049] In specific implementations, step S4 provided by this invention may include the following steps: S401: The current temperature, cumulative lethal value, and cumulative surface damage value of the internal core cold point output by the state estimation module are imported into a multivariable model predictive controller based on the TwinCAT control platform. Combining the state-space extrapolation evolution mechanism in control theory, a forward rolling prediction time window is set, limiting its span to 3 to 10 minutes after the current calculation physical time slice. Using the current real-time time step start point as the initial state boundary, the parameter changes within future time slices are continuously solved based on the built-in finite difference calculation matrix while maintaining the current thermophysical state. Then, the internal core lethal trajectory and surface quality degradation trend within the forward prediction window are calculated from the current real-time time step start point. Thus, the forward extrapolation formula corresponding to the surface quality degradation trend is established as follows: ; In the formula, This represents the projected value of future surface quality degradation of the terminal at the current physical timestamp. The current physical timestamp; This represents the total number of forward prediction steps. This is the cumulative damage value of the product surface calculated at the current physical timestamp; Index of prediction steps for discrete time steps; The reaction rate constant is at the absolute reference temperature; The activation energy for protein denaturation; It is the ideal gas constant; This is an absolute reference temperature. The predicted value of the surface dynamic absolute temperature at the future time step obtained from the forward extrapolation of the prediction model; To predict the discrete time step, the final value of the internal core lethal trajectory and the final value of the surface quality degradation inference of the forward prediction window terminal are extracted and output to the decision execution module.

[0050] S402 establishes a bivariate threshold judgment rule in the judgment execution module, and establishes an over-limit trigger switching mechanism based on the physical phenomenon of surface overheating and core undercooking under thermodynamic non-equilibrium conduction state. It compares the final value of the internal core lethal trajectory with the safety target value of microbial inactivation, and simultaneously compares the final value of surface quality deterioration with the equivalent threshold of surface quality damage. Then, when it is determined that the cumulative lethal value is insufficient for the safety target value and the cumulative surface damage will reach the equivalent threshold, an over-limit intervention interruption signal is sent to the system main control bus, thereby triggering the mode switching intervention logic. Furthermore, by executing a hardware-level interrupt instruction, the conventional temperature and pressure closed-loop control link of the ambient absolute temperature and absolute pressure during the steam heating stage is forcibly cut off, thereby releasing the thermodynamic fluid boundary state into the nonlinear intervention control range.

[0051] In specific implementations, step S5 provided by the present invention may include the following steps: S501, after the system main control bus trigger mode switching intervention logic, controls the Burkert proportional pressure relief valve installed at the exhaust end of the heating chamber to output a pulse duty cycle signal, and performs nonlinear pulse step-down pressure reduction to flash vaporize on the packaging surface to form a steam insulation layer. At the same time, combined with the physical property that the liquid boiling point decreases due to pressure change, it causes the attached moisture and surface seepage juice on the surface of the soft packaging of meat products to undergo asynchronous flash vaporization, thereby generating a water vapor insulation barrier with low dielectric constant and high thermal resistance, thereby cutting off the heat conduction channel from the external steam environment to the interior of the meat products.

[0052] S502, during the construction of the steam insulation layer, the first derivative of the complex reflection coefficient output by the state estimation module in the time domain is extracted. The physical mechanism of microwave boundary condition changes caused by gas-liquid phase transition, leading to high-frequency oscillations of the reflection signal, is introduced. Then, a sliding window integral is performed on the first derivative of the complex reflection coefficient. Based on this, the time-domain integral verification formula corresponding to the insulation layer coverage rate is established as follows: ; In the formula, The accumulated value of the sliding integral of the reflection feature calculated using the current physical timestamp; The sliding time window length is set to be between 2 and 5 seconds. For integration time; The modulus of the complex reflection coefficient corresponding to the integral time variable; This is the current physical timestamp.

[0053] Next, it is determined whether the accumulated value of the sliding integral of the reflection feature crosses the set coverage tolerance threshold, and the coverage tolerance threshold is limited to the range of 0.85 to 0.95. Finally, when the accumulated value of the sliding integral of the reflection feature crosses the coverage tolerance threshold, the thermal isolation state of the entire system is verified based on the change of the integral of the time domain derivative of the radio frequency reflection.

[0054] After verifying the establishment of the thermal isolation state of the entire system based on the integral change of the time-domain derivative of the radio frequency reflection, the generator inside the high-frequency electromagnetic power supply unit outputs a set non-heating stage sweep frequency signal to the radio frequency input matching terminal, and limits the power of the non-heating stage sweep frequency signal to a test range of 10 to 50W. Then, it obtains the current impedance mismatch of the radio frequency input matching terminal under the interference of the steam insulation gas layer, and then calculates the reactance deviation value between the current impedance mismatch and the standard matching point according to the existing transmission line impedance matching calculation theory. Based on the reactance deviation value, it outputs an adjustment command to the automatic matching network to drive the impedance matching unit to perform phase angle reconstruction, thereby eliminating the reactive reflected power introduced by the gas cavity insulation barrier.

[0055] After completing phase angle reconstruction, S504 issues a power jump command to the high-frequency electromagnetic power supply unit. It also combines sinusoidal frequency modulation to change the spatial distribution of electromagnetic wave standing wave nodes, thereby disrupting the physical mechanism of arc breakdown. Then, while increasing the RF output power, a sinusoidal frequency modulation mechanism is introduced to control the generator to output a frequency-modulated RF signal, which penetrates the air cavity heat insulation barrier in a low dielectric constant state. Combined with the high-frequency ionization disturbance characteristics of the spatial nodes, frequency-modulated RF heating power is continuously applied to the core area, thereby achieving concentrated heat replenishment and temperature rise of the internal cold points under the state of external thermal environment isolation.

[0056] See attached document Figure 6 In specific implementations, step S6 provided by the present invention may include the following steps: S601 stops heating after the cumulative lethal value at the center reaches the safety target value and sends a heat source cut-off command to the system main control bus. Then, it controls the high-frequency electromagnetic power supply unit and the steam circuit to close synchronously, thereby controlling the pressure relief valve at the exhaust end of the heating chamber to maintain the set back pressure state, limiting the set back pressure state pressure value to the range of 0.12 to 0.15 MPa. At the same time, it drives the circulating water pump and opens the spray solenoid valve, switching the fluid circuit to the pressure-maintaining spray cooling stage. Combined with the forced convection heat transfer physical phenomenon between the ambient cold fluid and the packaging surface, it causes the absolute temperature inside the meat product to drop to the safe storage temperature range according to the exponential decay trajectory, and limits the safe storage temperature range to 0 to 4°C, thereby eliminating the continuous change in internal biological activity.

[0057] S602, after the pressure-maintaining spray cooling stage ends, the actual cumulative measured value of conductivity of the current batch is extracted from the theoretical predicted value of conductivity output by the state estimation module, and the algebraic difference between the actual cumulative measured value of conductivity and the theoretical predicted value of conductivity is calculated to obtain the global conductivity integral error. At the same time, the hysteresis response mechanism between pressure change and actual physical state evolution in the thermodynamic system is introduced. Then, the actual time span from the time the pressure reduction command is issued to the cavity pressure reaching the target low pressure is extracted as the pressure reduction response time constant. The global conductivity integral error and the pressure reduction response time constant are matrix-concatenated and integrated into the state observation vector. Finally, the convergence integral deviation of the current batch is extracted and the state observation vector is directly assigned as the convergence integral deviation.

[0058] S603, after obtaining the convergence integral deviation, combines the existing Kalman filter state update algorithm in the control system and uses the convergence integral deviation to adaptively optimize the deviation weights of the calculated state equation. Here, the adaptive optimization is essentially an online continuous training and calibration model for calculating the state equation. The training data of the online continuous training and calibration model comes from the batch operation result dataset generated by the continuous operation of the production system. The online training mechanism and evolution method are as follows: when the system is initially running, a state covariance matrix containing multi-dimensional features of physical parameters is established, and the deviation weights are initialized to the identity matrix. After each batch of work is completed, the system automatically extracts the state observation vector as a supervision error signal, which serves as the true measurement value of the Kalman filter. Then, during the state update training phase, the supervision error signal is substituted into the Kalman observation feedback equation. By minimizing the mean square error of the posterior state estimate as the objective criterion, the Kalman gain matrix is ​​dynamically solved and updated. Finally, the Kalman gain matrix is ​​injected as the dynamic learning rate into the weight update formula, so that the bias weights continuously evolve towards approximating the real physical conditions as production batches accumulate, achieving closed-loop lifelong training of the computational model. Thus, the iterative update formula for the bias weights is established as follows: ; In the formula, The updated calculation of the deviation weights of the state equation. The bias weights are used to calculate the state equations for the current batch, and the initial bias weights for the first iteration are limited to the identity matrix. The optimal Kalman gain matrix is ​​calculated based on existing Kalman filtering algorithms; This is the convergence integral deviation vector extracted from the current batch processing; This represents the number of batch iterations.

[0059] Then, the updated deviation weights of the calculated state equation are written into the non-volatile storage module of the system controller. At the same time, the cumulative effect of the system's historical operating data is combined to complete the posterior closed-loop iteration of the calculated state equation.

[0060] Reference Figure 7 To aid in understanding the technical solution of this invention, the following is an application example of quality stability control during the sterilization and cooking process of 250g soft-packaged braised beef chunks.

[0061] Before batch processing begins, the operating terminal receives the material specifications of this batch of soft-packaged braised beef chunks. The material specifications include a total mass of 250 grams, an internal fat mass percentage of 12%, an inorganic salt mass fraction of 1.8%, and a high-temperature resistant transparent composite soft film packaging material. The system's internal static inference model maps the inorganic salt mass fraction to the characteristic conductivity coefficient of the corresponding meat product's characteristic juices based on the input information, and maps the internal fat mass percentage to the initial thermal diffusivity coefficient. At the same time, based on the target microbial standard, it establishes a microbial inactivation safety target value of 4.0 minutes for the internal core cold point, and based on reaction kinetics, it converts the 4% upper limit of surface water loss rate into a surface quality damage equivalent threshold for the irreversible chemical reaction under the corresponding thermal process.

[0062] After the sterilization process is started, the system main control bus synchronously sends a start command to activate the steam proportional regulating valve and the solid-state radio frequency generator. The absolute temperature of the steam inside the heating chamber increases linearly at a slope of 5 degrees Celsius per minute. At the same time, the solid-state radio frequency generator outputs an initial base power of 200 watts according to the set ratio. During the heating period, the system's underlying hardware synchronously collects the absolute temperature, absolute pressure, instantaneous mass flow rate and conductivity of the discharged condensate in the heating chamber at a fixed clock cycle of 200 milliseconds. Based on the differential removal of the interference of the basic condensate fluid from the instantaneous conductivity signal of the condensate, the actual water loss of the meat product during the processing stage is calculated. Combining the water loss data reflecting the shrinkage of the grid structure and the dielectric constant change rate reflecting electrophysical fluctuations, the true thermal diffusivity of the meat product is updated. The three-dimensional thermal conduction finite difference calculation matrix is ​​then driven to iteratively output the current absolute surface temperature, internal central cold point absolute temperature and cumulative lethal value of the meat product.

[0063] When heating reaches the 15th minute, the prediction and judgment module uses the current real-time time step as the initial state boundary and extrapolates a forward rolling prediction time window spanning 5 minutes into the future physical time slice. Based on the finite difference calculation matrix, the system extrapolates that the cumulative lethal value of the internal central cold point at the 5-minute prediction window terminal is 3.2 minutes, which does not meet the 4.0-minute microbial inactivation safety target value. Simultaneously, the system extrapolates that the final value of surface quality deterioration exceeds the set surface quality damage equivalent threshold. Based on the over-limit extrapolation result, the system sends an over-limit intervention interruption signal to the system main control bus, forcibly cutting off the conventional temperature and pressure closed-loop control link of the absolute temperature and absolute pressure of the environment during the steam heating stage and triggering the mode switching intervention logic.

[0064] After the mode switching intervention logic takes over, the system controls the pressure relief valve to perform a step-by-step pressure reduction action from 0.2 MPa to 0.12 MPa. The sudden pressure drop causes the attached moisture and exudate on the packaging surface to undergo asynchronous flash vaporization, generating a water vapor heat insulation barrier between the inside of the external flexible packaging film and the surface of the meat product. The system performs sliding window integration calculation on the first derivative of the complex reflection coefficient to verify that the heat insulation layer coverage exceeds the tolerance threshold. Subsequently, the automatic matching network drives the impedance matching unit to perform phase angle reconstruction based on the reactance deviation value to eliminate reactive reflected power. After the phase angle reconstruction is completed, the system controls the generator to continuously apply radio frequency heating power with a jump to 600 watts and sinusoidal frequency modulation characteristics to the core area, penetrating the water vapor heat insulation barrier and directly concentrating heat to the cold point in the center of the meat product.

[0065] After the cumulative lethal value at the internal central cold point reaches the safety target value of 4.0 minutes, the system cuts off the heat source and maintains the set cavity back pressure state. At the same time, it switches to the pressure-maintaining spray cooling stage to reduce the internal absolute temperature of the meat products to a safe storage temperature of 4°C. After the current batch operation is terminated, the model calibration module extracts the algebraic difference between the actual cumulative conductivity measurement value and the theoretical conductivity prediction value, as well as the actual pressure drop response time constant, and combines them to form a state observation vector. The system uses the extracted convergence integral deviation to substitute into the Kalman filter state update algorithm to solve for the optimal Kalman gain matrix to adaptively optimize and iterate the deviation weights of the calculated state equation. The updated calculated state equation deviation weights are written into the non-volatile storage module of the system controller for the next production batch to call for estimation.

[0066] The control method provided in this application embodiment establishes a gaseous heat insulation barrier by triggering physical depressurization flash evaporation through forward prediction of over-limit signals, and performs internal heat replenishment in conjunction with radio frequency phase angle reconstruction and frequency conversion power leap. At the same time, it combines the cumulative historical operating deviations of the production process to perform adaptive correction of the state equation weights, thereby achieving decoupled optimization control of the sterilization safety and surface quality stability of meat products.

[0067] While specific embodiments of the present invention have been described above, those skilled in the art should understand that these specific embodiments are merely illustrative. Those skilled in the art can omit, substitute, and modify the details of the above methods and systems in various ways without departing from the principles and essence of the present invention. For example, combining the above method steps to perform substantially the same function and achieve substantially the same result using substantially the same method falls within the scope of the present invention.

Claims

1. A production system for the stable control of the quality of products of the sterilization and cooking process of meat products, characterized by, The system includes: The parameter configuration module is used to receive the material specification parameters of the meat products to be processed, and to establish the microbial inactivation safety target value and the surface quality damage equivalent threshold of the meat products according to the set calculation criteria. The data acquisition module is used to dynamically acquire the absolute temperature, absolute pressure, reflection coefficient and phase angle of the RF input matching terminal in the heating chamber, as well as the instantaneous mass flow rate and conductivity of the discharged condensate. The state estimation module is used to calculate the actual water loss of meat products during the processing stage based on the differential removal of interference from the basic condensate fluid using the real-time conductive signal of condensate. It updates the actual thermal diffusivity by combining the first derivative of the complex reflection coefficient in the time domain and iteratively outputs the current temperature, cumulative lethal value, and cumulative damage value of the core cold point inside the meat product. The prediction and judgment module is used to calculate the internal core lethal trajectory and surface quality deterioration trend within the forward prediction window from the current real-time time step starting point, and send an over-limit intervention interruption signal to the system main control bus when it is determined that the cumulative lethal value is less than the microbial inactivation safety target value and the cumulative surface damage will reach the surface quality damage equivalent threshold. The intervention and control module is used to receive the over-limit intervention interruption signal, perform nonlinear pulse step-down to flash evaporation to form a vapor insulation layer on the packaging surface, drive the impedance matching unit to perform phase angle reconstruction, and continuously apply frequency-modulated radio frequency heating power to the core area.

2. The meat product sterilization and cooking process quality stability control production system according to claim 1, characterized in that, The system also includes a model calibration module, which is used to extract the convergence integral deviation of the current batch processing and adaptively optimize the deviation weights for calculating the state equation; The parameter configuration module contains a fixed initial physical property mapping model. The initial property mapping model uses the internal fat mass ratio and inorganic salt mass fraction of the offline collected physical samples as input feature variables, and the true value of the initial thermal diffusivity coefficient and the true value of the characteristic conductivity of the pure precipitated juice as the output labels. It is obtained by iterative training using a multilayer perceptron or support vector regression algorithm until the verification error converges.

3. The meat product sterilization and cooking process quality stability control production system according to claim 1, characterized in that, The data acquisition module is specifically used for: The absolute temperature inside the heating chamber is collected using platinum resistance temperature sensors distributed inside the chamber, and the absolute pressure inside the heating chamber is obtained using a pressure transmitter. High-frequency clock phase-locked loop technology based on field-programmable gate arrays extracts electromagnetic data on the same nanosecond time segment and calculates the reflection coefficient and phase angle at the RF input matching end; The driving measuring instrument performs continuous synchronous measurement to obtain the instantaneous mass flow rate and conductivity of the discharged condensate, and encapsulates the parameters in a real-time data frame containing a timestamp.

4. The meat product sterilization and cooking process quality stability control production system according to claim 1, characterized in that, When the state estimation module performs differential calculation to separate and separate the actual water loss: The pure condensate generated by the phase change of pure water vapor is set as the zero conductivity substrate. Then, the interference of the basic condensate fluid is removed by differentially removing the instantaneous conductivity signal of the condensate. At the same time, the characteristic conductivity coefficient is introduced to establish a mass conservation differential relationship based on the fluid mixing law. Thus, the actual water loss of meat products during the processing stage is calculated as water loss data.

5. The meat product sterilization and cooking process quality stability control production system according to claim 4, characterized in that, When updating the true thermal diffusivity, the state estimation module: The actual water loss reflecting the shrinkage of the grid structure and the first derivative of the complex reflection coefficient reflecting electrophysical fluctuations in the time domain are substituted into the data-driven dynamic thermodynamic correction model. Based on the initial calibrated thermal diffusivity of the material, compensation calculations are performed using pre-fitted weighted constants for water loss coupling and dielectric reflection coupling, and the updated true thermal diffusivity is calibrated and output in real time.

6. The meat product sterilization and cooking process quality stability control production system according to claim 1, characterized in that, The prediction and determination module is specifically used for: The current temperature, cumulative lethal value, and cumulative surface damage value of the output internal core cold point are imported into a multivariable model predictive controller based on the TwinCAT control platform. A forward rolling prediction time window is set, and parameter changes are continuously solved based on the built-in finite difference calculation matrix to predict the internal core lethal trajectory. The surface quality degradation trend is calculated by forward extrapolation using the reaction kinetic equation, and a bivariate threshold determination rule is established to trigger the mode switching intervention logic.

7. The meat product sterilization and cooking process quality stability control production system according to claim 1, characterized in that, When the intervention and control module performs nonlinear pulse stepwise blood pressure reduction: This causes the adhering moisture and surface exudate on the soft packaging of meat products to undergo asynchronous flash vaporization, cutting off the heat conduction channel from the external steam environment to the interior of the meat products to generate the steam insulation layer. The cumulative value of the sliding integral of the reflection feature is obtained by performing a sliding window integral on the first derivative of the complex reflection coefficient in the time domain; When the cumulative value of the sliding integral of the reflection feature crosses the set coverage tolerance threshold, the thermal isolation state of the entire system is verified based on the change of the integral of the time-domain derivative of the radio frequency reflection.

8. The meat product sterilization and cooking process quality stability control production system according to claim 7, characterized in that, When the intervention control module performs phase angle reconstruction and applies power to the driving impedance matching unit: After verifying that the thermal isolation state is established, a non-heating stage sweep frequency signal is output to calculate the reactance deviation value between the current impedance mismatch and the standard matching point. The impedance matching unit is driven by the reactance deviation value to eliminate the reactive reflected power introduced by the air cavity thermal insulation barrier in order to complete the phase angle reconstruction. After completing the phase angle reconstruction, the RF output power is increased and a sinusoidal frequency modulation mechanism is introduced to continuously apply frequency-modulated RF heating power to the core area.

9. The meat product sterilization and cooking process quality stability control production system according to claim 2, characterized in that, When the model calibration module extracts the convergence integral deviation of batch processing: Switch the control fluid circuit to the pressure-maintaining spray cooling stage; Calculate the algebraic difference between the cumulative measured value of actual conductivity and the theoretical predicted value of conductivity for the current batch to obtain the global conductivity integral error; The actual time span from the moment the pressure reduction command is issued to the moment the cavity pressure reaches the target low pressure is extracted as the pressure reduction response time constant. The global conductivity integral error and the voltage drop response time constant are matrix-concatenated and integrated into a state observation vector, which is then directly assigned the convergence integral deviation.

10. The meat product sterilization and cooking process quality stability control production system according to claim 9, characterized in that, When the model calibration module adaptively optimizes the deviation weights for calculating the state equation: The state observation vector is extracted as a supervision error signal and substituted into the Kalman observation feedback equation as the actual measurement value of the Kalman filter. The optimal Kalman gain matrix is ​​dynamically solved and updated by minimizing the mean square error criterion of the posterior state estimate; The updated optimal Kalman gain matrix is ​​injected into the weight update formula as a dynamic learning rate, driving the bias weights of the calculated state equation to undergo posterior closed-loop iterative update.