A method and system for hierarchical modeling, dynamic prediction and uncertainty analysis of a microbial inactivation process

Abstract

The present application relates to a kind of microbial inactivation process hierarchical modeling, dynamic prediction and uncertainty analysis method and system.The method, first, obtains the experimental data including processing variable, microbial survival amount and process environmental factor;Call preset primary model library to carry out multi-model fitting, based on correction akaike information criterion AICc automatically determine optimal target primary model, and extract model parameters under different environmental conditions to construct data set;With the data set as the basis, call secondary model library to establish the response relationship between environmental factor and inactivation parameter.For time-varying process conditions, the present application uses secondary model to calculate instantaneous inactivation parameter, realizes dynamic survival amount prediction by numerical integration, and carries out uncertainty propagation analysis based on parameter variance-covariance matrix and process disturbance.The present application solves the problem of large manual transcription error and weak adaptability of traditional method, significantly improves the efficiency, prediction accuracy and reliability of food processing sterilization process design.

Description

Technical Field

[0001] This invention belongs to the fields of food processing and safety, predictive microbiology, food processing technology modeling, model uncertainty analysis and computer-aided decision-making technology, specifically involving a hierarchical modeling, dynamic prediction and uncertainty analysis method and system for microbial inactivation process. Background Technology

[0002] In food processing, sterilization, disinfection, and preservation, the inactivation effect of target microorganisms directly affects product safety, shelf life, and process stability. To determine reasonable conditions for heat treatment, high-pressure treatment, photo-inactivation, pulsed electric field treatment, chemical disinfection, or irradiation, it is usually necessary to quantitatively predict the survival changes of target microorganisms under different process conditions. Predictive microbiology typically employs a hierarchical modeling approach: a primary model describes the change in microbial survival as a function of treatment variables under fixed process conditions, while a secondary model describes the change in primary model parameters as a function of environmental factors. In actual processing, process conditions often change over time, thus requiring dynamic prediction of microbial survival changes during real-world processing. Existing methods typically use different software, tables, or scripts to complete primary model fitting, secondary model establishment, dynamic prediction, and uncertainty analysis, which suffers from problems such as manual parameter transcription, model selection relying on experience, difficulty in process reproducibility, and insufficient adaptability to various non-thermal processing technologies. Furthermore, traditional methods often only provide a single prediction result, failing to reflect the uncertainties caused by model parameter estimation errors and fluctuations in process control factors. Therefore, it is necessary to provide a method and system for predicting microbial inactivation that can integrate primary model fitting, automatic model selection, primary parameter automatic transfer, secondary response modeling, dynamic process prediction, and uncertainty analysis. Summary of the Invention

[0003] The purpose of this invention is to provide a hierarchical modeling, dynamic prediction and uncertainty analysis method and system for microbial inactivation process, so as to realize the integration, automation and quantification of microbial inactivation prediction from fixed condition fitting to time-varying process prediction and prediction reliability evaluation, thereby improving the efficiency of food processing sterilization or disinfection process design.

[0004] To achieve the above objectives, the technical solution of the present invention is: a method for hierarchical modeling, dynamic prediction, and uncertainty analysis of microbial inactivation processes, comprising:

[0005] Acquire microbial inactivation experimental data, which includes treatment variable data, microbial survival data, and corresponding process environmental factor data;

[0006] The preset primary extinction biomechanical model library is called to perform multi-model fitting on the microbial survival data, and the model parameters and fitting evaluation indexes corresponding to each candidate primary model are obtained.

[0007] The candidate linear models are ranked based on the Corrected Akaike Information Criterion (AICc), and the target linear model is determined based on the ranking results.

[0008] For experimental data under different process environmental factors, multi-model fitting and target model determination operations were performed, and corresponding parameters were extracted using the same target primary model structure to construct a primary model parameter dataset;

[0009] Using the obtained primary model parameters as response variables and the corresponding process environmental factors as independent variables, a quadratic response model between the primary model parameters and the process environmental factors is established by calling the preset quadratic model library.

[0010] Obtain the time-varying process condition curves for the processing to be predicted;

[0011] The instantaneous inactivation parameters corresponding to different time points or integral substeps in the dynamic process are calculated using the quadratic response model, and the instantaneous inactivation parameters are substituted into the target primary model to establish an equation for the change of microbial survival under time-varying process conditions.

[0012] The equation for the change in microbial survival is numerically integrated to output a dynamic process organism survival prediction curve, logarithmic reduction value, and cumulative lethality.

[0013] Parameter samples are generated based on the estimated values ​​of model parameters and their variance-covariance matrix. Dynamic inactivation prediction is repeatedly performed by combining the perturbation samples of process control factors to obtain the uncertainty distribution of the survival prediction results.

[0014] Furthermore, the processing variable data includes at least one of processing time, light flux, absorbed dose, high pressure holding time, electric field treatment time, and disinfectant contact time; the process environmental factor data includes at least one of temperature, pressure, water activity, pH value, light intensity, electric field intensity, disinfectant concentration, and dose rate.

[0015] Furthermore, the primary inactivation kinetic model library includes at least two of the following: log-linear model, log-linear model with tail residue, linear shoulder model, linear shoulder-tail model, Weibull model, Weibull model with tail, biphasic model, biphasic model with shoulder, hybrid Weibull model, improved Gompertz model, and Baranyi inactivation model; the multi-model fitting uses nonlinear least squares method for parameter estimation and Levenberg-Marquardt algorithm for optimization solution.

[0016] Furthermore, the primary model parameters include at least one of the following: D value, rate constant, scale parameter, shape parameter, shoulder length, residual bacterial count parameter, and subpopulation ratio parameter; and for the same set of microbial inactivation experimental data under different process environmental factors used to establish a secondary model, the corresponding parameters are extracted by using the same target primary model structure to ensure that the resulting primary model parameter dataset is comparable.

[0017] Furthermore, the quadratic model library includes at least one of the following: Bigelow model, extended Bigelow model, Arrhenius model, Arrhenius-Eyring model, Ratkowsky square root model, linear model, constant model, and user-defined factor model; and before establishing the quadratic response model, the parameters of the first-order model are subjected to logarithmic transformation, natural logarithmic transformation, square root transformation, or identity transformation, and the corresponding inverse transformation is performed during prediction.

[0018] Furthermore, when calculating the instantaneous inactivation parameters, piecewise linear interpolation is used to obtain the process environment factor values ​​at the integral substep; when numerically integrating the equation for the change in the number of microorganisms, the fourth-order Runge-Kutta method is used.

[0019] Furthermore, when generating parameter samples, the variance-covariance matrix is ​​decomposed using Cholesky decomposition to generate joint parameter samples that maintain the correlation of model parameters; for model parameters that are strictly positive, sampling is performed in logarithmic space, and physical boundary constraints are applied to the sampling results.

[0020] Furthermore, it also includes a sensitivity analysis step: performing one-way perturbation sensitivity analysis and / or standardized regression coefficient SRC variance contribution analysis on the input variables to obtain the ranking or variance contribution ratio of the influence of the input variables on the microbial survival prediction results; the input variables include at least two of the following: process control factors, primary model parameters, and secondary model parameters.

[0021] Furthermore, when the process technology type is photo-inactivation, the light flux is calculated based on the integral value of light intensity and treatment time, and the light flux is used as the processing variable data; when the process technology type is irradiation treatment, the absorbed dose is calculated based on the integral value of dose rate and treatment time, and the absorbed dose is used as the processing variable data.

[0022] Furthermore, the corrected Akaike Information Criterion (AICc) is calculated based on the sum of squared residuals of the candidate linear model, the number of samples, and the number of model parameters.

[0023] This invention also provides a hierarchical modeling, dynamic prediction, and uncertainty analysis system for microbial inactivation processes, including a processor, a memory, and a bus. The memory stores machine-readable instructions that the processor can execute. When executed, the machine-readable instructions implement the method described above to achieve the following functional modules:

[0024] The data acquisition module is used to acquire microbial inactivation experimental data, which includes processing variable data, microbial survival data, and corresponding process environmental factor data.

[0025] The primary model fitting module is used to call a preset primary extinction biomechanical model library to perform multi-model fitting on the microbial survival data, and obtain the model parameters and fitting evaluation indexes corresponding to each candidate primary model.

[0026] The model selection module is used to rank the candidate linear models based on the Corrected Akaike Information Criterion (AICc) and determine the target linear model.

[0027] The parameter extraction module is used to extract target primary model parameters from microbial inactivation experimental data under different process environmental factors, and to form a primary model parameter dataset using the same target primary model structure.

[0028] The quadratic model establishment module is used to establish a quadratic response model with the parameters of the primary model as response variables and the corresponding process environmental factors as independent variables.

[0029] The dynamic prediction module is used to calculate the instantaneous inactivation parameters based on the secondary response model, and substitute the instantaneous inactivation parameters into the target primary model to establish the microbial survival rate change equation, thereby performing dynamic inactivation prediction.

[0030] The uncertainty analysis module is used to generate parameter samples based on the estimated values ​​of model parameters and their variance-covariance matrix, and to obtain the uncertainty distribution of the prediction results by combining the disturbance samples of process control factors.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] (1) Integrate the primary model fitting, model selection, parameter extraction, secondary modeling, dynamic prediction and uncertainty analysis into a unified process to improve the degree of automation and reproducibility;

[0033] (2) The target primary model is determined based on multi-model fitting and AICc automatic selection mechanism to improve the objectivity of model selection;

[0034] (3) Improve parameter comparability and secondary model stability by automatically transferring primary model parameters and extracting a unified model structure;

[0035] (4) It can adapt to dynamic inactivation prediction under time-varying process conditions, and is closer to the actual processing process;

[0036] (5) It can quantify the prediction uncertainty caused by model parameters and process fluctuations, and provide a basis for process safety margin design. Attached Figure Description

[0037] Figure 1 This is a system architecture diagram of hierarchical modeling, dynamic prediction and uncertainty analysis of microbial inactivation process according to a preferred embodiment of the present invention.

[0038] Figure 2 The flowchart illustrates the method for hierarchical modeling, dynamic prediction, and uncertainty analysis of the microbial inactivation process, which is a preferred embodiment of the present invention.

[0039] Figure 3 This is a flowchart of an automatic model selection process according to a preferred embodiment of the present invention.

[0040] Figure 4 This is a schematic diagram illustrating the transfer of primary model parameters to a secondary model according to a preferred embodiment of the present invention.

[0041] Figure 5 This is a flowchart illustrating the time-varying process dynamic prediction of a preferred embodiment of the present invention.

[0042] Figure 6 This is a schematic diagram of the dynamic survival prediction results of a preferred embodiment of the present invention.

[0043] Figure 7 This is a flowchart illustrating the uncertainty propagation and sensitivity analysis of a preferred embodiment of the present invention.

[0044] Figure 8 This is a diagram showing the results of a single model fitting and automatic selection under heat treatment in Embodiment 1 of the present invention.

[0045] Figure 9 This is a diagram showing the quadratic response surface and fitting results of Embodiment 1 of the present invention.

[0046] Figure 10 This is a diagram showing the uncertainty prediction distribution and local and global sensitivity analysis of Embodiment 1 of the present invention.

[0047] Figure 11 This is a diagram showing the results of a single model fitting and automatic selection under non-thermal-optical processing in Embodiment 2 of the present invention. Detailed Implementation

[0048] The present invention will be further described below with reference to the accompanying drawings and embodiments. The following embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. Where there is no conflict, the technical features in the following embodiments can be combined with each other.

[0049] The method described in this invention is applicable to the prediction and analysis of the inactivation effect of target microorganisms during food processing, sterilization, disinfection and preservation, and can be used in processes such as heat treatment, high pressure treatment, photoinactivation, pulsed electric field treatment, chemical disinfection and irradiation treatment.

[0050] like Figure 1 As shown, the system provided by this invention includes a data input module, a primary model fitting module, a model selection module, an automatic parameter transfer module, a secondary model building module, a dynamic prediction module, an uncertainty analysis module, a sensitivity analysis module, and a result output module. These modules work together to complete the import of microbial inactivation experimental data, model fitting, parameter transfer, dynamic prediction, uncertainty analysis, and result output.

[0051] like Figure 2 As shown, the method of the present invention includes steps such as data acquisition, primary model fitting, determination of the target primary model, extraction of primary model parameters, establishment of secondary model, input of dynamic process conditions, dynamic inactivation prediction, output of prediction results and uncertainty analysis; sensitivity analysis can also be further performed as needed.

[0052] Step 1: Data Acquisition

[0053] Users input or import microbial inactivation experimental data, which includes treatment variable data, microbial survival data, and corresponding process environmental factor data. Treatment variable data may include treatment time, light flux, absorbed dose, high-pressure holding time, electric field treatment time, or disinfectant contact time; process environmental factors may include temperature, pressure, water activity, pH value, light intensity, electric field intensity, disinfectant concentration, or dose rate. The system preferably performs integrity checks and format validation on the input data.

[0054] Step 2: First-time model fitting

[0055] The system calls upon a pre-defined primary inactivation kinetic model library to perform multi-model fitting on microbial survival data under fixed process conditions, obtaining model parameters and fitting evaluation indices for each candidate primary model. Candidate primary models may include, for example, log-linear models, Weibull models, biphasic models, modified Gompertz models, and Baranyi inactivation models. The system preferentially employs nonlinear least squares method for parameter estimation and the Levenberg-Marquardt algorithm for parameter optimization. Boundary constraints can be set for parameters such as D-value, rate constant, scale parameter, and shape parameter.

[0056] In one embodiment, the log-linear model can be represented as:

[0057]

[0058] In another embodiment, the Weibull model can be represented as:

[0059]

[0060] Where N is the number of surviving microorganisms at time t, N0 is the initial number of microorganisms, and D is the decimal attenuation time. is the scale parameter, and p is the shape parameter.

[0061] Step 3: Determine the target model in one step

[0062] After fitting the candidate first-order models, the system sorts the multiple candidate first-order models based on the Corrected Akaike Information Criterion (AICc) and determines the target first-order model. The system can output the target first-order model name, model parameters, fitting curve, and evaluation index.

[0063] In one embodiment, the Akaike Information Correction Criterion (AICc) can be calculated using the following formula:

[0064]

[0065] Where n is the number of samples, SSE is the sum of squared residuals, and k is the number of model parameters. The candidate linear model with the smallest AICc value is preferably selected as the target linear model.

[0066] For multiple sets of process environment condition data used to establish a quadratic model, it is preferable to determine a unified primary model structure based on the comprehensive AICc sorting results and the consistency of cross-condition parameters.

[0067] Step 4: First-time model parameter extraction and automatic transfer

[0068] The system repeatedly executes steps 2 and 3 on experimental data under multiple process environmental factors to obtain the target primary model parameters under the corresponding conditions. These parameters are then automatically associated with the corresponding process environmental factors to form a primary model parameter dataset. To ensure consistency in parameter meaning and comparability across conditions, it is preferable to extract parameters using the same target primary model structure for the same set of data used to build a secondary model. The primary model parameters may include at least one of the following: D-value, rate constant, scale parameter, shape parameter, shoulder length, residual bacterial count parameter, or subpopulation ratio parameter. The system automatically transfers the primary model parameter dataset to the secondary model building module via a parameter transfer module.

[0069] Figure 3 This is a flowchart of an automatic model selection process according to a preferred embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the transfer of primary model parameters to a secondary model according to a preferred embodiment of the present invention.

[0070] Step 5: Establishing the secondary model

[0071] The system uses the primary model parameters obtained in step 4 as the response variables and the corresponding process environmental factors as independent variables, and calls a preset quadratic model library to establish a quadratic response model. The quadratic model library may include Bigelow models, extended Bigelow models, Arrhenius models, Arrhenius-Eyring models, Ratkowsky square root models, linear models, constant models, and user-defined factor models. For different types of primary model parameters, the system can perform logarithmic transformations, natural logarithmic transformations, square root transformations, or identity transformations before fitting, and perform corresponding inverse transformations during prediction.

[0072] In one embodiment, when the first-order model parameter is D, the Bigelow model can be represented as:

[0073]

[0074] Among them, D ref Reference temperature T ref The D value below, This represents the amount of temperature change corresponding to a logarithmic change in the D value caused by a temperature change.

[0075] Step 6: Input Dynamic Process Conditions

[0076] The user inputs a time-varying process condition curve for the processing to be predicted. This curve may include at least one of the following: temperature-time curve, pressure-time curve, light intensity-time curve, electric field intensity-time curve, disinfectant concentration-time curve, or dose rate-time curve. The system can establish piecewise curves based on the input time points and corresponding process environmental factor values ​​for use in calculating instantaneous parameters during subsequent integration.

[0077] In embodiments involving photoinactivation or irradiation treatment, the system can also convert the treatment variables according to the process technology type. When the process technology type is photoinactivation, the irradiance can be calculated based on the integral value of light intensity and treatment time, and used as the treatment variable; when the process technology type is irradiation treatment, the absorbed dose can be calculated based on the integral value of dose rate and treatment time, and used as the treatment variable; when the process conditions are constant or a discrete approximation is used, the calculation can also be performed based on the corresponding product value.

[0078] Step 7: Dynamic inactivation prediction

[0079] like Figure 5As shown, during the dynamic prediction process, the system reads the process environmental factor values ​​at the current time point or the integral substep, calculates the corresponding instantaneous parameters of the primary model based on the quadratic response model established in step 5, and then substitutes the instantaneous parameters into the target primary model determined in step 3 to establish the microbial survival rate change equation. For cases where the process environmental factors change continuously between adjacent sampling times, piecewise linear interpolation is preferably used to obtain the process environmental factor values ​​at the integral substep.

[0080] Under dynamic process conditions, the change in microbial survival can be expressed as:

[0081]

[0082] Where N represents the number of microorganisms that survive. The parameter vector of the time-varying linear model is obtained by mapping the process environmental factor u(t) through a quadratic response model, where u(t) is the time-varying process environmental factor, and f(·) is determined by the mathematical form of the target linear model. The system performs numerical integration on the equation to obtain the microbial survival prediction results under dynamic process conditions.

[0083] Step 8: Output the prediction results

[0084] like Figure 6 As shown, the system performs numerical integration on the microbial survival rate change equation established in step 7, preferably using the fourth-order Runge-Kutta method for integration calculation, to obtain the microbial survival prediction curve under dynamic process conditions over time. Based on this, the system further outputs the logarithmic reduction value and cumulative lethality; in one embodiment, it can also output the treatment time or process intensity required to achieve the predetermined logarithmic reduction value according to the user-defined target inactivation level.

[0085] Step 9: Uncertainty Analysis

[0086] like Figure 7 As shown, the system generates parameter samples based on the model parameter estimates and their variance-covariance matrix obtained in steps 2 and / or 5, and repeatedly performs dynamic inactivation prediction in conjunction with process control factor disturbance samples to obtain the uncertainty distribution of the prediction results. Preferably, the system uses Cholesky decomposition to generate joint parameter samples that maintain the correlation of model parameters; for strictly positive model parameters, sampling is preferably performed in logarithmic space with physical boundary constraints applied. The system can output the predicted mean, confidence interval, probability of achievement, and sample results required for sensitivity analysis.

[0087] In one embodiment, let the mean vector of the model parameters be... If the variance-covariance matrix is ​​Sigma, then the joint parameter sample can be expressed as:

[0088]

[0089] Where L is the lower triangular matrix of Sigma's Cholesky decomposition, and z is a standard normal random vector.

[0090] Step 10: Sensitivity Analysis

[0091] After completing step 9, the system can further perform sensitivity analysis. This sensitivity analysis can employ one-way perturbation analysis and / or standardized regression coefficient (SRC) variance contribution analysis to quantitatively assess the relationship between changes in input variables and changes in prediction results. The input variables can be categorized according to process control factors, primary model parameters, and secondary model parameters, and their impact on the uncertainty of the prediction results can be ranked or their variance contribution ratio can be output respectively.

[0092] Step 11: System Implementation

[0093] This invention also provides a system for hierarchical modeling, dynamic prediction, and uncertainty analysis of microbial inactivation processes to implement the above-described methods. The system includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the system runs, the machine-readable instructions are executed by the processor to implement the above-described methods. Further, the system can be divided into a data input module, a primary model fitting module, a model selection module, an automatic parameter transfer module, a secondary model building module, a dynamic prediction module, an uncertainty analysis module, a sensitivity analysis module, and a result output module. Each of these modules can be implemented by the same processor executing the machine-readable instructions.

[0094] Through the above implementation methods, the present invention can integrate primary model fitting, automatic target model selection, primary parameter automatic transfer, secondary response modeling, time-varying process dynamic prediction, uncertainty propagation and sensitivity analysis into a unified process, thereby achieving quantitative prediction and risk control of food processing or microbial control process conditions.

[0095] Example 1: Comprehensive Prediction and Uncertainty Analysis of Microbial Inactivation Based on Heat Treatment Process

[0096] This embodiment aims to illustrate the application of the hierarchical modeling, dynamic prediction, and uncertainty analysis method and system for microbial inactivation processes provided by the present invention in the entire process of typical heat treatment processes.

[0097] Step (1) Obtain microbial inactivation experimental data: Obtain experimental data on the survival of target microorganisms (such as Salmonella) in dried chili matrix under different constant temperature and water activity (aw) combinations, as well as corresponding process environmental factor data.

[0098] Step (2) Call the preset primary extinguishing dynamics model library to perform multi-model fitting: such as Figure 8 As shown, under specific experimental conditions of 60°C and water activity of 0.33, the system performs multi-model fitting on the survival data to obtain the model parameters and fitting evaluation indexes corresponding to each candidate linear model (such as log-linear model, Weibull model, biphasic model, etc.).

[0099] Step (3) Determine the target linear model based on AICc: The system calculates the correction Akaike Information Criterion (AICc) value of the log-linear model based on the fitting results. The value is -19.28, which is the smallest among all candidate models. The system automatically determines it as the target linear model based on this ranking result.

[0100] Step (4) Extracting primary model parameters to form a dataset: The system performs steps (2) and (3) on experimental data under different temperature and water activity conditions, and extracts corresponding parameters using the same objective log-linear model structure. For example, the inactivation D value under 60°C and aw=0.33 conditions is extracted to be 54.8 min, thus forming a comparable primary model parameter dataset.

[0101] Step (5) Establish a quadratic response model: Using the obtained primary model parameters (D value) as the response variable and the corresponding temperature and water activity as independent variables, call the preset quadratic model library. Combined with... Figure 9 From the three-dimensional response surface, it can be seen that the system establishes an extended Bigelow quadratic model, and the parameters are obtained by fitting: reference D value D ref = 12.88 min, temperature-sensitive parameter z T =18.4 °C, water activity sensitive parameter z aw = 0.392.

[0102] Step (6) Obtain the time-varying process condition curve: Input the dynamic time-varying process condition curve of the processing to be predicted. In this embodiment, it is set as follows: linearly increase the temperature from 55°C to 75°C for 10 min, hold at 75°C for 5 min, and then cool down to 55°C for 5 min. During the process, the water activity is kept constant at 0.45.

[0103] Step (7) Calculate the instantaneous inactivation parameter and establish the equation: Based on the extended Bigelow model established in step (5), the system calculates the instantaneous inactivation D value corresponding to different integral substeps in the above dynamic process, and substitutes the instantaneous inactivation parameter into the target log-linear model to establish the equation for the change of microbial survival under time-varying process conditions.

[0104] Step (8) Numerical Integration and Prediction Output: The system uses the fourth-order Runge-Kutta method to numerically integrate the equation for the change in the survival rate of the microorganism. For example... Figure 6As shown, the microbial survival prediction curve under dynamic process conditions is output. The results show that inactivation mainly occurs during the high-temperature holding stage, and the final output cumulative reduction in lethal log is approximately 4.5 log.

[0105] Step (9) Uncertainty Distribution Generation: Based on the estimated model parameters and their variance-covariance matrix, parameter samples are generated. Dynamic inactivation prediction is then repeatedly executed, incorporating set process control disturbances (e.g., a temperature standard deviation of 1°C). Reference Figure 10 It can be seen that the system obtains the uncertainty distribution of the prediction results and outputs a suggested safe processing time of 7.9 min at a 99% confidence level.

[0106] Sensitivity analysis steps (additional): Perform standardized regression coefficient (SRC) variance contribution analysis on the input variables. The results show that the uncertainty of the prediction results is mainly driven by the quadratic model parameters (variance contribution rate 87.8%) and process control factors (variance contribution rate 11.8%).

[0107] Example 2: Nonlinear microbial kinetics modeling based on nonthermal photo-inactivation process

[0108] This embodiment aims to illustrate the compatibility and accuracy of the system of the present invention in processing non-thermal physical techniques and nonlinear dynamic characteristic data.

[0109] Step (1) Data Acquisition and Processing Variable Transformation: Acquire the light pulse inactivation data of the target microorganism (e.g., Listeria monocytogenes) on the surface of fresh-cut broccoli. Since the process technology type is light inactivation, the system automatically converts the integral value of light intensity and time into the processing variable—light flux (J / cm²). 2 ).

[0110] Step (2) Automatic Model Selection and Parameter Extraction: The system calls a preset primary model library to fit the data. Combining the AICc criterion and overfitting prevention evaluation, the Weibull model has an AICc value of -33.016, exhibiting the best fitting performance while avoiding parameter overfitting. The system identifies it as the target model. Figure 11 The ranking results show that the model effectively captures the obvious nonlinear characteristics of the survival curve as light flux increases. The Weibull model parameters extracted by the system are: δ (J / cm²) = 0.0409±0.0155, p = 0.102±0.0114.

[0111] This embodiment demonstrates the system's high adaptability under different processing variable dimensions (time vs. flux) and different dynamic modes (linear vs. nonlinear), providing accurate quantitative evaluation support for non-thermal processing processes.

[0112] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for hierarchical modeling, dynamic prediction, and uncertainty analysis of microbial inactivation processes, characterized in that, include: Acquire microbial inactivation experimental data, which includes treatment variable data, microbial survival data, and corresponding process environmental factor data; The preset primary extinction biomechanical model library is called to perform multi-model fitting on the microbial survival data, and the model parameters and fitting evaluation indexes corresponding to each candidate primary model are obtained. The candidate linear models are ranked based on the Corrected Akaike Information Criterion (AICc), and the target linear model is determined based on the ranking results. For experimental data under different process environmental factors, multi-model fitting and target model determination operations were performed, and corresponding parameters were extracted using the same target primary model structure to construct a primary model parameter dataset; Using the obtained primary model parameters as response variables and the corresponding process environmental factors as independent variables, a quadratic response model between the primary model parameters and the process environmental factors is established by calling the preset quadratic model library. Obtain the time-varying process condition curves for the processing to be predicted; The instantaneous inactivation parameters corresponding to different time points or integral substeps in the dynamic process are calculated using the quadratic response model, and the instantaneous inactivation parameters are substituted into the target primary model to establish an equation for the change of microbial survival under time-varying process conditions. The equation for the change in microbial survival is numerically integrated to output a dynamic process organism survival prediction curve, logarithmic reduction value, and cumulative lethality. Parameter samples are generated based on the estimated values ​​of model parameters and their variance-covariance matrix. Dynamic inactivation prediction is repeatedly performed by combining the perturbation samples of process control factors to obtain the uncertainty distribution of the survival prediction results.

2. The method for hierarchical modeling, dynamic prediction, and uncertainty analysis of microbial inactivation processes according to claim 1, characterized in that, The processing variable data includes at least one of the following: processing time, light flux, absorbed dose, high pressure holding time, electric field treatment time, and disinfectant contact time; the process environmental factor data includes at least one of the following: temperature, pressure, water activity, pH value, light intensity, electric field intensity, disinfectant concentration, and dose rate.

3. The method for hierarchical modeling, dynamic prediction, and uncertainty analysis of microbial inactivation processes according to claim 1, characterized in that, The single-stage inactivation kinetic model library includes at least two of the following: log-linear model, log-linear model with tail residue, linear shoulder model, linear shoulder-tail model, Weibull model, Weibull model with tail, biphasic model, biphasic model with shoulder, hybrid Weibull model, improved Gompertz model, and Baranyi inactivation model; the multi-model fitting uses nonlinear least squares method for parameter estimation and Levenberg-Marquardt algorithm for optimization solution.

4. The method for hierarchical modeling, dynamic prediction, and uncertainty analysis of microbial inactivation processes according to claim 1, characterized in that, The primary model parameters include at least one of the following: D-value, rate constant, scale parameter, shape parameter, shoulder length, residual bacterial count parameter, and subpopulation ratio parameter; and for the same set of microbial inactivation experimental data under different process environmental factors used to establish a secondary model, the corresponding parameters are extracted by using the same target primary model structure to ensure that the resulting primary model parameter dataset is comparable.

5. The method for hierarchical modeling, dynamic prediction, and uncertainty analysis of microbial inactivation processes according to claim 1, characterized in that, The quadratic model library includes at least one of the following: Bigelow model, extended Bigelow model, Arrhenius model, Arrhenius-Eyring model, Ratkowsky square root model, linear model, constant model, and user-defined factor model; and before establishing the quadratic response model, the parameters of the first-order model are subjected to logarithmic transformation, natural logarithmic transformation, square root transformation, or identity transformation, and the corresponding inverse transformation is performed during prediction.

6. The method for hierarchical modeling, dynamic prediction, and uncertainty analysis of microbial inactivation processes according to claim 1, characterized in that, When calculating the instantaneous inactivation parameters, piecewise linear interpolation is used to obtain the process environment factor values ​​at the integral substep; when performing numerical integration on the microbial survival change equation, the fourth-order Runge-Kutta method is used.

7. The method for hierarchical modeling, dynamic prediction, and uncertainty analysis of microbial inactivation processes according to claim 1, characterized in that, When generating parameter samples, the variance-covariance matrix is ​​decomposed using Cholesky decomposition to generate joint parameter samples that preserve the correlation of model parameters; for model parameters that are strictly positive, sampling is performed in logarithmic space and physical boundary constraints are applied to the sampling results.

8. The method for hierarchical modeling, dynamic prediction, and uncertainty analysis of microbial inactivation processes according to claim 1, characterized in that, It also includes a sensitivity analysis step: performing one-way perturbation sensitivity analysis and / or standardized regression coefficient SRC variance contribution analysis on the input variables to obtain the ranking or variance contribution ratio of the input variables on the microbial survival prediction results; the input variables include at least two of the following: process control factors, primary model parameters, and secondary model parameters.

9. The method for hierarchical modeling, dynamic prediction, and uncertainty analysis of microbial inactivation processes according to claim 1, characterized in that, When the process technology type is photo-inactivation, the light flux is calculated based on the integral value of light intensity and treatment time, and the light flux is used as the processing variable data; when the process technology type is irradiation treatment, the absorbed dose is calculated based on the integral value of dose rate and treatment time, and the absorbed dose is used as the processing variable data.

10. A system for hierarchical modeling, dynamic prediction, and uncertainty analysis of microbial inactivation processes, characterized in that, The system includes a processor, a memory, and a bus. The memory stores machine-readable instructions that the processor can execute. When executed, the machine-readable instructions implement the method as described in any one of claims 1 to 9, to achieve the following functional modules: The data acquisition module is used to acquire microbial inactivation experimental data, which includes processing variable data, microbial survival data, and corresponding process environmental factor data. The primary model fitting module is used to call a preset primary extinction biomechanical model library to perform multi-model fitting on the microbial survival data, and obtain the model parameters and fitting evaluation indexes corresponding to each candidate primary model. The model selection module is used to rank the candidate linear models based on the Corrected Akaike Information Criterion (AICc) and determine the target linear model. The parameter extraction module is used to extract target primary model parameters from microbial inactivation experimental data under different process environmental factors, and to form a primary model parameter dataset using the same target primary model structure. The quadratic model establishment module is used to establish a quadratic response model with the parameters of the primary model as response variables and the corresponding process environmental factors as independent variables. The dynamic prediction module is used to calculate the instantaneous inactivation parameters based on the secondary response model, and substitute the instantaneous inactivation parameters into the target primary model to establish the microbial survival rate change equation, thereby performing dynamic inactivation prediction. The uncertainty analysis module is used to generate parameter samples based on the estimated values ​​of model parameters and their variance-covariance matrix, and to obtain the uncertainty distribution of the prediction results by combining the disturbance samples of process control factors.

Images