Data Analysis-Based Monitoring Method, System, and Medium for the Preparation of Fish Micronized Protein

Abstract

This application relates to the field of data processing technology, and discloses a data analysis-based monitoring method, system, and medium for the preparation of surimi protein. The method includes: real-time detection of plant protein melt within an extruder using multi-point temperature sensors and a near-infrared spectrometer to obtain an extrusion state data matrix; analysis of process parameter correlations using a weighted fusion algorithm to obtain a multi-parameter coupling coefficient table; establishment of a mapping relationship model between textural properties and process parameters through regression analysis, combined with texture analyzer data; prediction of textural deviation values ​​using a deviation compensation algorithm based on this model to generate quality deviation early warning indicators; and gradient correction of temperature and rotation speed using a graded control strategy based on the early warning indicators, outputting a real-time parameter adjustment instruction set. This application solves the problem of lacking real-time data monitoring and intelligent parameter adjustment in existing surimi protein preparation processes, improving product quality stability and the level of automation control in the production process.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, system and medium for monitoring the preparation of mimic fish surimi protein based on data analysis. Background Technology

[0002] With the rapid development of the plant-based protein food market, fish paste-like protein products have attracted widespread attention as alternatives to traditional animal proteins. Existing plant-based textured protein preparation technologies mainly employ extrusion puffing processes. By controlling process parameters such as extruder temperature, pressure, and screw speed, the molecular structure of plant proteins is altered to form a fibrous texture similar to fish paste. Traditional preparation monitoring methods rely on manual experience and offline testing. Operators adjust process parameters based on subjective judgments such as product appearance and feel, and quality testing is conducted through sampling and testing, with results lagging behind the production process.

[0003] However, existing technologies have significant shortcomings: First, there is a lack of real-time monitoring methods for changes in protein molecular structure during extrusion, making it impossible to accurately grasp the degree of protein denaturation and the process of fibrosis; second, the adjustment of process parameters mainly relies on operational experience, lacking scientific parameter correlation analysis and prediction models, resulting in poor product quality stability; finally, quality control adopts a post-processing inspection method, which cannot achieve real-time correction during the process, resulting in the waste of unqualified products.

[0004] Based on the above analysis of the current technological status, it can be inferred that the core problem facing existing technologies is the lack of an intelligent monitoring system based on data analysis. Specifically, the inability to acquire and effectively integrate multi-dimensional data on the extrusion state in real time makes it impossible to establish a precise mapping relationship between process parameters and product texture characteristics; furthermore, the lack of reliable predictive models makes it impossible to achieve early warning and accurate compensation for quality deviations; and finally, the lack of adaptive parameter control strategies makes it impossible to perform intelligent process optimization based on real-time quality status, severely restricting the industrialization and quality improvement of surimi-based protein products. Summary of the Invention

[0005] This application provides a data analysis-based method, system, and medium for monitoring the preparation of surimi protein, which addresses the lack of real-time data monitoring and intelligent parameter adjustment in existing surimi protein preparation processes, thereby improving product quality stability and the level of automation control in the production process.

[0006] Firstly, this application provides a data analysis-based monitoring method for the preparation of surimi-like protein. The method includes: real-time detection and processing of plant protein melt within an extruder using multi-point temperature sensors and a near-infrared spectrometer to obtain an extrusion state data matrix containing temperature distribution and protein denaturation degree; based on the extrusion state data matrix, performing correlation analysis on screw speed, feed rate, and moisture addition parameters using a weighted fusion algorithm to obtain a multi-parameter coupling coefficient table; combining the multi-parameter coupling coefficient table with elasticity and hardness data detected by a texture analyzer, and modeling the formation law of surimi-like fiber structure using a regression analysis algorithm to obtain a mapping relationship model between texture characteristics and process parameters; based on the mapping relationship model, performing prediction calculation on the deviation value of texture characteristics in the current preparation process using a deviation compensation algorithm to obtain a surimi-like quality deviation early warning index; and based on the numerical range of the surimi-like quality deviation early warning index, performing gradient correction processing on the extrusion temperature and screw speed using a graded control strategy to obtain a real-time process parameter adjustment instruction set.

[0007] Optionally, the real-time detection and processing of the vegetable protein melt inside the extruder using multi-point temperature sensors and a near-infrared spectrometer yields an extrusion state data matrix containing temperature distribution and protein denaturation degree, including:

[0008] Thermocouple temperature sensors are installed in the feeding section, compression section, shearing section, cooling section and discharge section of the extruder barrel. The temperature data of each section is collected and processed at high frequency to obtain a multi-segment temperature distribution vector.

[0009] The multiple temperature distribution vectors are arranged and stored according to time series to obtain a temperature change time series database;

[0010] The absorption peak intensities of the first and second amide bands were scanned and detected using a near-infrared spectrometer to obtain spectral data on changes in protein secondary structure.

[0011] Based on the relative changes in the content of helical and folded structures in the spectral data, the degree of protein denaturation is quantitatively calculated to obtain a percentage value of the degree of denaturation.

[0012] The multiple temperature distribution vectors and the percentage values ​​of the degree of deformation are combined and arranged in matrix form to obtain an extrusion state data matrix containing temperature and deformation information.

[0013] Optionally, based on the extrusion state data matrix, a weighted fusion algorithm is used to perform correlation analysis on the screw speed, feed rate, and moisture addition parameters to obtain a multi-parameter coupling relationship coefficient table, including:

[0014] The temperature distribution vector and the percentage value of the degree of deformation in the extrusion state data matrix are segmented according to the time window to obtain the time-segmented extrusion state sub-matrix.

[0015] The screw speed, feed rate and moisture addition amount parameters are collected and processed synchronously based on encoders and flow meters to obtain a real-time monitoring dataset of process parameters.

[0016] The real-time monitoring dataset of process parameters and the time-segmented extrusion status submatrix are matched according to timestamps to obtain a parameter synchronization association data table.

[0017] Based on the parameter synchronization correlation data table, the degree of mutual influence between each process parameter is quantitatively analyzed and processed by the Pearson correlation coefficient calculation method to obtain the parameter correlation numerical matrix.

[0018] The correlation matrix between the parameters is weighted and calibrated according to the influence intensity to obtain a multi-parameter coupling relationship coefficient table containing coupling strength and influence direction.

[0019] Optionally, the step of combining the multi-parameter coupling coefficient table with the elasticity and hardness data detected by the texture analyzer, and modeling the formation law of the simulated surimi fiber structure through regression analysis algorithm, to obtain a mapping relationship model between texture properties and process parameters, includes:

[0020] Based on the texture analyzer, the fish paste-like sample was subjected to compression elasticity test and hardness puncture test to obtain a texture property dataset including elastic recovery rate and maximum puncture force.

[0021] The texture property dataset and the multi-parameter coupling coefficient table are matched according to sample batches to obtain the correlation data table between texture properties and coupling coefficients.

[0022] Based on the aforementioned associated data table, the functional relationship between the elastic recovery rate and the coupling coefficient of the process parameters is fitted and calculated using a multiple linear regression algorithm to obtain an elastic parameter prediction model.

[0023] Based on the aforementioned associated data table, a hardness parameter prediction model is obtained by fitting the functional relationship between the maximum puncture force and the coupling coefficient of the process parameters using a multinomial regression algorithm.

[0024] By integrating the elastic parameter prediction model and the hardness parameter prediction model into a mathematical framework, a mapping relationship model describing the quantitative relationship between textural properties and process parameters is obtained.

[0025] Optionally, the step of fitting and calculating the functional relationship between the elastic recovery rate and the coupling coefficient of the process parameters using a multiple linear regression algorithm based on the associated data table to obtain an elastic parameter prediction model includes:

[0026] The elastic recovery rate values ​​in the associated data table are extracted as dependent variables to obtain the elastic recovery rate target data vector.

[0027] Based on the coupling coefficient of the process parameters in the associated data table as independent variables, a matrix construction process is performed to obtain an independent variable feature matrix containing screw speed coefficient, feed rate coefficient and moisture addition coefficient.

[0028] The feature matrix of the independent variables is processed by least squares to obtain the regression coefficient vector of the influence of each process parameter on the elastic recovery rate.

[0029] Based on the regression coefficient vector, the model fitting accuracy is verified and calculated using residual analysis to obtain model evaluation indicators including the coefficient of determination and standard error.

[0030] The regression coefficient vector and model evaluation index are encapsulated into mathematical expressions to obtain an elastic parameter prediction model for predicting the elastic recovery rate.

[0031] Optionally, based on the mapping relationship model, the deviation value of the textural properties in the current preparation process is predicted and calculated using a deviation compensation algorithm to obtain a fish paste quality deviation early warning index, including:

[0032] The real-time process parameters of the current preparation process are input into the mapping relationship model to perform textural property prediction calculations, and the current textural property prediction results, including elasticity prediction values ​​and hardness prediction values, are obtained.

[0033] The texture analysis instrument is used to perform real-time texture detection on the prepared fish paste sample to obtain the measured texture properties, including the measured elasticity value and the measured hardness value.

[0034] The measured results of the current texture properties and the predicted results of the current texture properties are processed by numerical difference calculation to obtain a texture property deviation value data group including elasticity deviation and hardness deviation.

[0035] Based on the data set of deviation values ​​of the texture characteristics, the degree of deviation is comprehensively evaluated and calculated using a weighted average algorithm to obtain a comprehensive deviation value that reflects the overall quality fluctuation.

[0036] The comprehensive deviation value is classified and judged according to a preset threshold range to obtain a fish paste quality deviation warning index that includes slight deviation, moderate deviation and severe deviation.

[0037] Optionally, based on the numerical range of the simulated surimi quality deviation early warning index, the extrusion temperature and screw speed are graded and corrected using a graded control strategy to obtain a real-time process parameter adjustment instruction set, including:

[0038] The simulated surimi quality deviation early warning index is classified into three levels: slight deviation, moderate deviation and severe deviation, to obtain a three-level deviation classification result.

[0039] Based on the three-level deviation classification results, the extrusion temperature adjustment range is gradient-set to obtain a temperature correction gradient table including small adjustment, medium adjustment and large adjustment.

[0040] Based on the three-level deviation classification results, the screw speed adjustment range is subjected to gradient setting processing to obtain a speed correction gradient table including low-speed adjustment, medium-speed adjustment and high-speed adjustment;

[0041] The temperature correction gradient table and the speed correction gradient table are processed according to the current deviation level to obtain the specific combination of adjustment parameters corresponding to the current quality deviation.

[0042] The specific combination of adjustment parameters is converted into digital signal processing that can be executed by the control system to obtain a real-time process parameter adjustment instruction set containing temperature adjustment instructions and speed adjustment instructions.

[0043] Secondly, this application provides a data analysis-based monitoring system for the preparation of surimi protein, the data analysis-based monitoring system for the preparation of surimi protein comprising:

[0044] The detection module is used to perform real-time detection and processing of the vegetable protein melt in the extruder using multi-point temperature sensors and a near-infrared spectrometer, and to obtain an extrusion state data matrix containing temperature distribution and protein denaturation degree.

[0045] The analysis module is used to perform correlation analysis on the screw speed, feed rate and moisture addition amount parameters based on the extrusion state data matrix using a weighted fusion algorithm, and obtain a multi-parameter coupling relationship coefficient table.

[0046] The modeling module is used to combine the multi-parameter coupling coefficient table with the elasticity and hardness data detected by the texture analyzer, and to model the formation law of the imitation surimi fiber structure through regression analysis algorithm to obtain the mapping relationship model between texture properties and process parameters.

[0047] The prediction module is used to predict and calculate the deviation value of the textural properties in the current preparation process based on the mapping relationship model and through the deviation compensation algorithm, so as to obtain the early warning index of the quality deviation of the surimi.

[0048] The correction module is used to perform gradient correction processing on the extrusion temperature and screw speed through a graded control strategy based on the numerical range of the pre-warning index for the quality deviation of the surimi-like material, so as to obtain a set of real-time process parameter adjustment instructions.

[0049] Thirdly, a data analysis-based monitoring device for the preparation of surimi protein is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the data analysis-based monitoring device for the preparation of surimi protein to execute the aforementioned data analysis-based monitoring method for the preparation of surimi protein.

[0050] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the above-described data analysis-based monitoring method for the preparation of surimi protein.

[0051] The technical solution provided in this application achieves comprehensive real-time monitoring of the state of plant protein melt within the extruder through the combined detection and processing of multi-point temperature sensors and near-infrared spectrometers, effectively solving the problems of insufficient monitoring accuracy and information loss in traditional single-sensor monitoring. The construction of the extrusion state data matrix organically integrates temperature distribution and protein denaturation information, providing a state characterization basis for subsequent data analysis. The weighted fusion algorithm analyzes the correlation between screw speed, feed rate, and moisture addition parameters, accurately identifying the complex coupling relationships between multiple process parameters. The generated multi-parameter coupling coefficient table quantitatively describes the degree and direction of mutual influence between parameters, overcoming the lack of scientific basis in traditional empirical adjustments. The regression analysis algorithm, through modeling the formation law of mimicking surimi fiber structure, establishes a precise mathematical mapping relationship between textural properties and process parameters. This model can accurately predict the textural properties of products under different process conditions, significantly improving the predictability and accuracy of quality control. The deviation compensation algorithm predicts and calculates deviations in textural properties, enabling early identification and quantitative assessment of quality deviations. The graded control strategy provides gradient correction for extrusion temperature and screw speed, automatically selecting the appropriate adjustment range and direction based on different levels of quality deviation. This ensures the accuracy and timeliness of parameter adjustment, avoiding the subjectivity and lag of manual adjustment.

[0052] The weighted fusion algorithm effectively eliminates dimensional differences and noise interference between data from different types of sensors through a reasonable weight allocation mechanism, significantly improving the accuracy and reliability of parameter correlation analysis. The regression analysis algorithm, employing a combination of multiple linear and polynomial fitting, can simultaneously handle the complex linear and nonlinear relationships between textural properties and process parameters. The established mapping model exhibits good prediction accuracy and generalization ability, providing a powerful mathematical tool for quality prediction in plant protein preparation processes. The deviation compensation algorithm, through residual analysis and statistical evaluation techniques, ensures the statistical significance and engineering applicability of the prediction results, making the identification of quality deviations more accurate and timely. The graded control strategy, through gradient parameter settings and intelligent adjustment logic, achieves differentiated and refined control strategies, significantly improving the automation level and response speed of process control, and providing an intelligent monitoring solution for the industrial production of surimi-like protein products. Attached Figure Description

[0053] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a schematic diagram of an embodiment of the data analysis-based monitoring method for the preparation of surimi protein in this application.

[0055] Figure 2 This is a schematic diagram of an embodiment of the data analysis-based monitoring system for the preparation of surimi protein in this application.

[0056] Figure 3 This is a schematic block diagram of the structure of the mimic fish surimi protein preparation monitoring device based on data analysis in an embodiment of the present invention. Detailed Implementation

[0057] This application provides a data analysis-based method, system, and medium for monitoring the preparation of surimi protein. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0058] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the data analysis-based monitoring method for the preparation of surimi protein in this application includes:

[0059] Step S101: The vegetable protein melt in the extruder is detected and processed in real time by using a multi-point temperature sensor and a near-infrared spectrometer to obtain an extrusion state data matrix containing temperature distribution and protein denaturation degree.

[0060] Step S102: Based on the extrusion state data matrix, the screw speed, feed rate and moisture addition amount parameters are analyzed and processed by a weighted fusion algorithm to obtain a multi-parameter coupling relationship coefficient table;

[0061] Step S103: Combine the multi-parameter coupling coefficient table with the elasticity and hardness data detected by the texture analyzer, and use regression analysis algorithm to model the formation law of the imitation surimi fiber structure to obtain the mapping relationship model between texture properties and process parameters.

[0062] Step S104: Based on the mapping relationship model, the deviation value of the textural characteristics in the current preparation process is predicted and calculated using the deviation compensation algorithm to obtain the early warning index of the quality deviation of the surimi.

[0063] Step S105: Based on the numerical range of the early warning index for the quality deviation of the surimi, the extrusion temperature and screw speed are graded and corrected using a graded control strategy to obtain a real-time process parameter adjustment instruction set.

[0064] It is understood that the executing entity of this application can be a data analysis-based monitoring system for the preparation of surimi protein, or it can be a terminal or a server; no specific limitation is made here. This application's embodiments use a server as an example for illustration.

[0065] Specifically, multi-point temperature sensors collect temperature data at the feeding, compression, shearing, cooling, and discharge sections of the extruder barrel, forming a multi-segment temperature distribution vector reflecting the heat transfer process. Simultaneously, a near-infrared spectrometer quantifies the degree of transformation from helical to folded structures in protein molecules by detecting changes in the absorption peak intensity of the first and second amide bands, calculating the percentage of denaturation. These data are arranged in a matrix to form an extrusion state data matrix containing temperature and denaturation information, directly reflecting the physicochemical state changes of plant proteins under high temperature and high pressure.

[0066] A weighted fusion algorithm is employed to handle the complex correlations between process parameters. First, the extrusion state data matrix is ​​segmented according to a preset time window to obtain time-segmented extrusion state sub-matrices. Then, screw speed, feed rate, and moisture addition are synchronously collected via encoders and flow meters to form a real-time monitoring dataset for process parameters. The Pearson correlation coefficient calculation method is used to quantify the degree of mutual influence between various process parameters. By calculating the ratio of the covariance to the standard deviation between each pair of parameters, a numerical value reflecting the strength of the linear correlation is obtained. The weight allocation process assigns influence weights to different parameter pairs based on the strength of the correlation, ultimately generating a multi-parameter coupling relationship coefficient table containing coupling strength and influence direction.

[0067] A quantitative mapping relationship between textural properties and process parameters was established using regression analysis algorithms. A texture analyzer was used to perform compressive elasticity and hardness / puncture tests on the simulated surimi sample, obtaining data on elastic recovery rate and maximum puncture force. A multiple linear regression algorithm used the elastic recovery rate as the dependent variable and the coupling coefficient of the process parameters as the independent variable, calculating the influence coefficient of each parameter on elasticity using the least squares method. A multinomial regression algorithm handled the nonlinear relationship between maximum puncture force and process parameters, considering the interaction terms and higher-order terms between parameters. The two prediction models were integrated through a mathematical framework to form a mapping relationship model that can describe the quantitative relationship between textural properties and process parameters.

[0068] Based on a mapping relationship model, the textural deviation of the current preparation process is predicted. Real-time process parameters are input into the model to calculate the predicted elasticity and hardness values. Simultaneously, a texture analyzer detects the actual sample to obtain the measured elasticity and hardness values. A deviation compensation algorithm calculates the difference between the predicted and measured values ​​to obtain the elasticity and hardness deviation. A weighted average algorithm assigns weights according to the importance of different textural parameters, calculates the comprehensive deviation value, and classifies it according to a preset threshold range, generating quality deviation warning indicators for slight, moderate, and severe deviations.

[0069] Based on the deviation warning indicators, a graded control strategy is implemented, with three deviation levels corresponding to different adjustment ranges. The temperature correction gradient table and the speed correction gradient table are pre-set with specific numerical ranges for small, medium, and large adjustments. When a moderate elastic deviation is detected, the control algorithm selects a medium-amplitude adjustment parameter from the temperature correction gradient table and a corresponding speed adjustment parameter from the speed correction gradient table, forming a specific combination of adjustment parameters. This is ultimately converted into temperature and speed adjustment commands that the control system can execute.

[0070] In one specific embodiment, the process of performing step S101 may specifically include the following steps:

[0071] Thermocouple temperature sensors are installed in the feeding section, compression section, shearing section, cooling section and discharge section of the extruder barrel. The temperature data of each section is collected and processed at high frequency to obtain a multi-segment temperature distribution vector.

[0072] Multiple temperature distribution vectors are arranged and stored according to time series to obtain a temperature change time series database;

[0073] The absorption peak intensities of the first and second amide bands were scanned and detected using a near-infrared spectrometer to obtain spectral data on changes in protein secondary structure.

[0074] Based on the relative changes in the content of helical and folded structures in the spectral data, the degree of protein denaturation is quantitatively calculated to obtain a percentage value of the degree of denaturation.

[0075] The multiple temperature distribution vectors and the percentage values ​​of the degree of deformation are combined and arranged in matrix form to obtain an extrusion state data matrix containing temperature and deformation information.

[0076] Specifically, the thermocouple temperature sensor works based on the thermoelectric effect of the electromotive force generated when two different metal materials change temperature. Installing these sensors at five key locations within the extruder barrel enables monitoring of the temperature of the plant protein melt. The feeding section, located at the very front of the extruder, primarily monitors the initial temperature of raw materials such as soybean protein and pea protein as they enter the extruder; this temperature is typically low and changes gradually. The compression section, located in the screw's compression zone, experiences intense compression from the screw, converting mechanical energy into heat and causing a rapid temperature rise; the temperature sensor in this section captures this abrupt temperature change. The shearing section is the area with the strongest shear force in the screw design; protein molecules undergo chain breakage and rearrangement under high shear stress, generating significant frictional heat, causing the temperature in this section to reach the peak of the entire extrusion process. The cooling section uses an external cooling device to lower the material temperature, providing a suitable temperature environment for protein molecule reorganization; the temperature sensor in this section monitors the cooling rate and the final stable temperature. The discharge section, located at the end of the extruder, has a temperature sensor that ensures the product reaches the preset temperature requirements when leaving the extruder.

[0077] High-frequency data acquisition refers to the continuous recording of temperature values ​​at a sampling frequency of ten times per second by temperature sensors. This high-frequency acquisition can capture minute fluctuations and instantaneous changes in temperature during the extrusion process. Each acquisition yields five temperature values ​​corresponding to five different locations, and these five values ​​are arranged in spatial order to form a multi-segment temperature distribution vector. This vector not only reflects the temperature state of each segment at a certain moment, but more importantly, it reflects the spatial distribution characteristics and heat transfer patterns within the extruder. The changing patterns of the temperature distribution vector can reveal the thermophysical state transitions of plant proteins at different processing stages.

[0078] The establishment of the temperature change time-series database involves systematically storing multiple temperature distribution vectors acquired each time in chronological order. The database employs a relational data structure, primarily containing a timestamp field, five temperature value fields, and relevant process condition fields. The timestamp field records the accurate acquisition time of each temperature vector, down to the millisecond level, ensuring data continuity. The five temperature value fields store real-time temperature data for the feeding, compression, shearing, cooling, and discharging sections, respectively. The process condition fields record parameters such as screw speed, feed rate, and moisture addition corresponding to the temperature data. Time-series sorting processing ensures data is stored in chronological order through the establishment of a time index, facilitating subsequent historical data queries, trend analysis, and anomaly detection.

[0079] The detection principle of near-infrared spectrometers is based on the theory of molecular vibrational absorption spectroscopy. When infrared light of a specific wavelength irradiates a plant protein melt, the chemical bonds in the protein molecules absorb infrared light energy corresponding to their vibrational frequencies. The first amide band is a characteristic absorption peak generated by the stretching vibration of the carbon-oxygen double bond in the amide bond of the protein molecule. Its wavelength position and intensity directly reflect the content and stability of the α-helix structure in the protein molecule. The second amide band is generated by the coupling of the bending vibration of the nitrogen-hydrogen bond and the stretching vibration of the carbon-nitrogen bond in the amide bond, mainly reflecting the changes in the β-sheet structure. Scanning detection and processing: The spectrometer's internal grating spectroscopic system decomposes the composite infrared light into monochromatic light of different wavelengths. The detector sequentially measures the light intensity at each wavelength, and the absorption intensity of each wavelength is calculated by comparing the difference between the incident light intensity and the transmitted light intensity, ultimately obtaining the absorption spectrum curve.

[0080] Quantitative calculation of protein denaturation is based on mathematical analysis of changes in secondary structure content. First, the initial secondary structure ratio of the native protein needs to be determined as a reference. The calculation process quantitatively assesses the degree of transformation from α-helix to β-sheet structure by analyzing the peak area changes of the first and second amide bands. Specifically, the calculation method compares the currently detected α-helix structure content with the initial content to calculate the proportion of structure loss, while simultaneously analyzing the increase in β-sheet structure. The total degree of structural change is obtained by combining these two changes. The percentage of denaturation is calculated using a formula; a higher value indicates a greater deviation of the protein's molecular structure from its native state and a more significant change in its functional properties.

[0081] Constructing the extrusion state data matrix is ​​a crucial step in data fusion. This matrix employs a six-column, multi-row two-dimensional array structure, where each column represents a specific state parameter and each row represents complete state information at a specific moment. The first five columns store the temperature values ​​for the feeding, compression, shearing, cooling, and discharging sections, respectively, while the sixth column stores the percentage of protein denaturation at the corresponding moment. The number of rows in the matrix is ​​determined by the data acquisition duration and sampling frequency; one row is added for each additional sampling period. This combination and arrangement process ensures complete temporal synchronization between temperature and denaturation data, using a unified timestamp for correspondence.

[0082] In one specific embodiment, the process of performing step S102 may specifically include the following steps:

[0083] The temperature distribution vector and the percentage of deformation in the extrusion state data matrix are segmented according to the time window to obtain the time-segmented extrusion state sub-matrix.

[0084] The screw speed, feed rate and moisture addition amount parameters are collected and processed synchronously based on encoders and flow meters to obtain a real-time monitoring dataset of process parameters.

[0085] The real-time monitoring dataset of process parameters and the sub-matrix of extrusion status in different time periods are matched according to timestamps to obtain a parameter synchronization association data table.

[0086] Based on the parameter synchronization correlation data table, the degree of mutual influence between various process parameters is quantitatively analyzed and processed by the Pearson correlation coefficient calculation method to obtain the parameter correlation numerical matrix.

[0087] The correlation matrix between parameters is weighted and calibrated according to the influence intensity to obtain a multi-parameter coupling relationship coefficient table containing coupling strength and influence direction.

[0088] Specifically, time window segmentation involves dividing the continuous extrusion state data matrix into multiple independent data segments according to preset time intervals for analysis. The time window is set based on the characteristics of the extrusion process and the needs of data analysis, typically selecting a time length that reflects process stability as the window size. The segmentation process begins at the start of the extrusion state data matrix and extracts data rows sequentially at fixed time intervals. The number of data rows within each time window is determined by the product of the sampling frequency and the window length. For example, with a sampling frequency of ten times per second and a time window of thirty seconds, each sub-matrix contains three hundred rows of data. The time-segmented extrusion state sub-matrix maintains the original six-column structure—five columns of temperature data and one column of deformation degree data—but the number of rows is significantly reduced compared to the original matrix. This segmentation method decomposes long-term continuous data into multiple short-time data blocks, facilitating the analysis of the changing characteristics and stability of the process state within different time periods.

[0089] The encoder synchronously acquires screw speed based on the photoelectric encoding principle. The encoder is mounted on the screw shaft and measures the speed by detecting changes in the shaft's rotation angle. The encoder's internal grating disk rotates synchronously with the screw shaft. A photoelectric sensor detects changes in the brightness of the grating stripes, converting the mechanical rotation signal into an electrical pulse signal. The speed value is obtained by calculating the number of pulses per unit time. The flow meter acquires the feed rate and moisture addition using the mass flow measurement principle. The mass flow meter measures the mass flow rate of material passing through the pipeline using the Coriolis effect, reflecting changes in the feed rate in real time. Moisture addition is measured using a dedicated moisture addition flow meter, installed on the moisture addition pipeline, measuring the mass of moisture added to the system per unit time. Synchronous acquisition and processing ensure that the data acquisition time for the three process parameters is completely consistent with the acquisition time for temperature and denaturation degree data. A unified time synchronization signal coordinates the sampling actions of each sensor. The real-time monitoring dataset for process parameters is organized in three columns, corresponding to screw speed, feed rate, and moisture addition, with the number of data rows synchronized with the acquisition frequency of temperature data.

[0090] The construction of the parameter synchronization and correlation data table achieves the correspondence between process parameter data and extrusion status data through timestamp matching technology. The timestamp matching process first adds a time stamp to each row of data in the real-time monitoring dataset of process parameters and the time-segmented extrusion status sub-matrix, with a time accuracy down to the millisecond level to ensure matching accuracy. The matching algorithm compares the timestamps of each row of data in the two datasets, merging data rows with the same timestamp or those within the allowable error range. The merged data table contains nine columns: the first five columns are temperature values ​​for each segment, the sixth column is the percentage of denaturation degree, and the last three columns are screw speed, feed rate, and moisture addition amount. Each row of data represents complete process status information at the same moment, including comprehensive information on temperature distribution, protein denaturation degree, and key process parameters. This synchronization and correlation processing ensures that the data foundation for subsequent correlation analysis has good temporal consistency and completeness.

[0091] The Pearson correlation coefficient method is used to quantify the linear correlation strength between various process parameters. This method is based on the covariance analysis theory in statistics. The calculation process first requires determining the parameter pairs to be analyzed, including the relationship between screw speed and temperature at different stages, the relationship between feed rate and temperature distribution, and the relationship between moisture addition and denaturation degree. For each pair of parameters, the Pearson correlation coefficient is calculated by solving the ratio of the product of the covariance and the standard deviation of each variable. The covariance reflects the common trend of the two variables, and the standard deviation reflects their respective dispersion. The ratio result varies between negative one and positive one. A correlation coefficient close to positive one indicates a strong positive correlation, that is, when one parameter increases, the other parameter tends to increase as well. A correlation coefficient close to negative one indicates a strong negative correlation, that is, when one parameter increases, the other parameter tends to decrease as well. A correlation coefficient close to zero indicates that there is no obvious linear correlation between the two parameters. The correlation coefficient matrix between parameters organizes the correlation coefficients of all parameter pairs in the form of a symmetric matrix. The rows and columns of the matrix represent different process parameters, and the matrix elements are the Pearson correlation coefficient values ​​of the corresponding parameter pairs.

[0092] The generation of the multi-parameter coupling coefficient table enables further quantitative analysis of the correlation matrix through weight allocation and coefficient calibration. Weight allocation determines the importance of different parameter pairs based on the absolute value of the correlation coefficients; the larger the absolute value of the correlation coefficient, the more significant the impact of the parameter pair on the entire process system, and the higher its weight. Weight allocation employs normalization to ensure that the sum of all weights equals one, avoiding excessively large or small weights that could affect subsequent analysis. Coefficient calibration converts the Pearson correlation coefficients into coupling strength coefficients suitable for engineering applications. The conversion process considers the numerical range of the correlation coefficients, the physical meaning of the process parameters, and the needs of practical applications. The coupling strength coefficients not only retain the numerical information of the original correlation coefficients but also add a clear indication of the direction of influence: positive values ​​indicate positive coupling influence, and negative values ​​indicate negative coupling influence. The multi-parameter coupling coefficient table organizes the coupling information of all parameter pairs in tabular form. The table includes fields such as parameter pair name, coupling strength coefficient, influence direction identifier, and weight coefficient, providing a quantitative theoretical basis for subsequent process optimization and control strategy formulation.

[0093] In one specific embodiment, the process of executing step S103 may specifically include the following steps:

[0094] Based on the texture analyzer, the fish paste-like sample was subjected to compression elasticity test and hardness puncture test to obtain a texture property dataset including elastic recovery rate and maximum puncture force.

[0095] The texture property dataset and the multi-parameter coupling coefficient table are matched according to the sample batch to obtain the correlation data table between texture properties and coupling coefficients;

[0096] Based on the associated data table, the functional relationship between the elastic recovery rate and the coupling coefficient of the process parameters is fitted and calculated using a multiple linear regression algorithm to obtain the elastic parameter prediction model.

[0097] Based on the associated data table, the coupling coefficient between the maximum puncture force and the process parameters is fitted using a multinomial regression algorithm to obtain a hardness parameter prediction model.

[0098] By integrating the elastic parameter prediction model and the hardness parameter prediction model into a mathematical framework, a mapping relationship model describing the quantitative relationship between textural properties and process parameters is obtained.

[0099] Specifically, the texture analyzer measures the compressive elasticity of the surimi-like samples using a standardized mechanical compression procedure. During the test, a cylindrical probe compresses the sample downwards at a constant speed to a preset deformation, then returns it to its initial position at the same speed. The entire process records the relationship between force and displacement. The compressive elasticity test employs a double compression cycle: the first compression measures the initial mechanical response of the sample, and the second compression assesses the sample's elastic recovery ability. The elastic recovery rate is calculated by dividing the height recovered during the second compression by the deformation amount during the first compression; this value directly reflects the elastic characteristics and texture quality of the surimi-like product. The hardness puncture test uses a sharp conical probe to penetrate the sample surface at a constant speed, measuring the maximum resistance during penetration. The maximum puncture force reflects the internal structural strength and fiber network density of the sample, and is an important indicator for evaluating the chewiness and texture of the surimi-like product. The texture characteristic dataset is organized in a two-column table format: the first column shows the elastic recovery rate, and the second column shows the maximum puncture force. Each row of data corresponds to complete texture characteristic information for a specific sample. Sample batch matching is achieved by establishing a sample coding system to accurately correlate textural property data with process parameter data. Each surimi-like sample is assigned a unique batch number during preparation, which records key information such as preparation time, raw material batch, and process conditions. Each sample in the textural property dataset is matched with its corresponding multi-parameter coupling coefficient table using its batch number, ensuring an accurate correspondence between textural test results and the coupling state of specific process parameters. The matching process is implemented using database query technology; the query algorithm searches for the corresponding coupling coefficient data in the multi-parameter coupling coefficient table based on the sample batch number. The associated data table merges the textural property data and coupling coefficient data into a unified data structure, containing the sample batch number, elastic recovery rate, maximum puncture force, and coupling strength coefficients between various process parameters. This association processing provides a data foundation for establishing a quantitative relationship between textural properties and process parameters.

[0100] Multiple linear regression is used to establish a mathematical model of the relationship between elastic recovery rate and the coupling coefficients of process parameters. The algorithm first establishes a linear regression equation with elastic recovery rate as the dependent variable and the coupling coefficients of process parameters as independent variables. The independent variables include multiple coupling parameters such as the coupling coefficients of screw speed and temperature, feed rate and degree of denaturation, and moisture addition and temperature at different stages. Multiple linear regression solves for the regression coefficients using the least squares method, which determines the optimal regression parameters by minimizing the sum of squared errors between actual and predicted values. The calculation process establishes a system of normal equations, and matrix operations are used to solve for the regression coefficients and constant terms for each independent variable. The regression coefficients reflect the degree and direction of influence of each coupling parameter on the elastic recovery rate; positive coefficients indicate a positive influence, and negative coefficients indicate a negative influence. The absolute value of the coefficient reflects the intensity of the influence. The elastic parameter prediction model expresses the quantitative relationship between elastic recovery rate and the coupling coefficients of process parameters in the form of a linear equation. This model can predict the elastic characteristics of the product based on the given coupling state of process parameters.

[0101] The polynomial regression algorithm handles the nonlinear relationship between maximum puncture force and the coupling coefficient of process parameters. This algorithm considers the interaction and higher-order effects among process parameters. The polynomial regression model includes first-order terms, second-order terms, and interaction terms. The first-order terms reflect the direct linear influence of each coupling parameter, the second-order terms reflect the nonlinear effect of a single parameter, and the interaction terms reflect the synergistic effect between different parameters. The algorithm achieves polynomial regression by expanding the independent variable matrix. The expanded matrix not only includes the original coupling coefficient variables but also the squared terms of each variable and the interaction terms formed by pairwise multiplication. Polynomial regression also uses the least squares method to solve for the regression parameters, but the number of parameters increases significantly, and the computational complexity increases accordingly. The regression calculation process establishes an expanded system of normal equations and uses matrix inversion to obtain the regression coefficients of all polynomial terms. The hardness parameter prediction model describes the complex nonlinear relationship between maximum puncture force and the coupling coefficient of process parameters in the form of polynomial equations. This model can more accurately capture the influence of process parameter changes on the hardness characteristics of the product.

[0102] The mathematical framework of the mapping relationship model is integrated by combining the elasticity parameter prediction model and the hardness parameter prediction model into a unified prediction system. The integration process establishes a multi-output prediction model structure, which uses the coupling coefficient of process parameters as a unified input variable and outputs predicted values ​​for two textural properties: elastic recovery rate and maximum puncture force. The mathematical framework expresses the integrated model in the form of vector functions, where the input is a vector containing all coupling coefficients, and the output is a vector containing the predicted values ​​of the two textural properties. The integration process also includes unified calibration of model parameters and consistency verification of prediction results, ensuring the coordination and reliability of the two sub-models within the unified framework. The mapping relationship model can not only independently predict elasticity and hardness properties but also analyze the interrelationships and synergistic changes between the two textural properties. This model provides a scientific theoretical tool for quality control and process optimization in the preparation of surimi protein-like materials. By inputting specific process parameter coupling states, it can accurately predict the textural properties of the product, guiding parameter adjustment and quality control decisions during production.

[0103] In one specific embodiment, the process of performing the fitting calculation of the functional relationship between the elastic recovery rate and the coupling coefficient of the process parameters using a multiple linear regression algorithm can specifically include the following steps:

[0104] The elastic recovery rate values ​​in the associated data table are extracted as dependent variables to obtain the elastic recovery rate target data vector.

[0105] The coupling coefficients of process parameters in the associated data table are used as independent variables to construct a matrix, resulting in a feature matrix of independent variables including screw speed coefficient, feed rate coefficient, and moisture addition coefficient.

[0106] The feature matrix of the independent variables is processed by least squares to obtain the regression coefficient vector of the influence of each process parameter on the elastic recovery rate.

[0107] Based on the regression coefficient vector, the model fitting accuracy is verified and calculated using residual analysis to obtain model evaluation indicators including the coefficient of determination and standard error.

[0108] By encapsulating the regression coefficient vector and model evaluation index into mathematical expressions, an elastic parameter prediction model for predicting elastic recovery rate is obtained.

[0109] Specifically, the extraction of the elastic recovery rate target data vector involves selecting elastic recovery rate values ​​from a specific column of a related data table and reorganizing them into a vector form. The related data table contains complete information on multiple sample batches, with each row corresponding to the textural properties and process parameter coupling coefficients of a single sample. The extraction process sequentially reads all values ​​from the elastic recovery rate column of the table through database queries or array indexing operations. These values ​​are arranged in chronological or batch order to form a one-dimensional array structure. The length of the elastic recovery rate target data vector is equal to the total number of samples, and each element in the vector represents the elastic recovery rate measurement result of a specific sample. This vector serves as the dependent variable in a multiple linear regression analysis, and its numerical changes reflect the differences in the elastic properties of the surimi-like product under different process conditions. The quality of the target data vector directly affects the accuracy of the regression model; therefore, the extraction process needs to ensure data integrity and accuracy, eliminating outliers and missing values ​​that could interfere with the analysis results.

[0110] The construction of the independent variable feature matrix is ​​based on the process parameter coupling coefficient data in the association data table. This matrix organizes the coupling relationships of the main process parameters affecting the elastic recovery rate into a standard matrix form. The matrix construction process first identifies the types of coupling coefficients closely related to elastic properties, mainly including the coupling coefficients of screw speed and shear zone temperature, feed rate and compression zone temperature, and moisture addition and protein denaturation degree. The number of rows in the matrix equals the total number of samples, and the number of columns equals the number of selected coupling coefficient types, typically three columns corresponding to the three main process parameter coupling relationships. Each element of the matrix represents the coefficient value of a specific sample under a specific process parameter coupling relationship; these values ​​reflect the strength and direction of the interaction between process parameters. The construction of the independent variable feature matrix also includes data standardization, using Z-score standardization or maximum / minimum standardization to eliminate differences in dimensions and numerical ranges between different coupling coefficients, ensuring that each independent variable has a considerable weight in the regression analysis.

[0111] The least squares method is the core algorithm for solving multiple linear regression parameters. This method determines the optimal regression coefficients by minimizing the sum of squared errors between predicted and observed values. The calculation process first establishes a system of normal equations. The matrix form of the normal equations is that the transpose of the independent variable characteristic matrix multiplied by the independent variable characteristic matrix equals the transpose of the independent variable characteristic matrix multiplied by the target data vector. The solution process obtains the regression coefficient vector through matrix inversion. Specifically, it is calculated as the inverse of the product of the transpose of the independent variable characteristic matrix and the independent variable characteristic matrix, multiplied by the product of the transpose of the independent variable characteristic matrix and the target data vector. The regression coefficient vector contains three elements, corresponding to the influence of the screw speed coefficient, feed rate coefficient, and moisture addition coefficient on the elastic recovery rate. Positive values ​​indicate a positive correlation between the coupling coefficient and the elastic recovery rate, while negative values ​​indicate a negative correlation. The absolute value reflects the intensity of the influence. The least squares calculation also generates a constant term, which reflects the baseline level of the elastic recovery rate when all process parameter coupling coefficients are zero.

[0112] Residual analysis assesses the fitting accuracy and reliability of a regression model by comparing the differences between model predictions and actual observations. The residual calculation process multiplies the regression coefficient vector by the feature matrix of the independent variables and adds a constant term to obtain the predicted elastic recovery rate for each sample. Then, the difference between the predicted value and the corresponding actual value in the target data vector is calculated. The coefficient of determination is determined by the ratio of the total sum of squares to the residual sum of squares. The total sum of squares reflects the total variability of the target variable, while the residual sum of squares reflects the variability that the model cannot explain. The ratio reflects the model's ability to explain data variation. The coefficient of determination ranges from zero to one; the closer the value is to one, the better the model fit and the higher the proportion of variation it can explain. The standard error is calculated by taking the square root of the residual sum of squares divided by the degrees of freedom. The degrees of freedom equal the number of samples minus the number of regression parameters. The standard error reflects the average deviation of the model's predictions. Model evaluation indicators also include the significance test results of each regression coefficient, using a t-test to determine the statistical significance of the influence of each process parameter coupling coefficient on the elastic recovery rate.

[0113] The mathematical expression of the elastic parameter prediction model encapsulates the regression coefficient vector, constant term, and model evaluation index into a predictive model structure. The encapsulation process establishes a standard linear regression equation form, where the left side of the equation represents the predicted elastic recovery rate, and the right side is a linear combination of the products of the coupling coefficients of each process parameter and their corresponding regression coefficients, plus a constant term. The mathematical expression clearly defines the physical meaning and numerical range of each symbol, ensuring that users can correctly understand and apply the predictive function. The encapsulation process also includes explanations of the model's applicable conditions, clarifying the effective range of values ​​for the process parameter coupling coefficients and the confidence interval for prediction accuracy. The elastic parameter prediction model provides a prediction interface in function form, with the input being the numerical values ​​of the three process parameter coupling coefficients, and the output being the corresponding predicted elastic recovery rate and prediction interval. This model provides a quantitative predictive tool for controlling the elastic properties of the surimi protein preparation process. Production personnel can quickly predict the elastic quality of the product based on the current coupling state of the process parameters and adjust the process parameters in a timely manner to ensure that the product quality meets the expected requirements.

[0114] In one specific embodiment, the process of executing step S104 may specifically include the following steps:

[0115] The real-time process parameters in the current preparation process are input into the mapping relationship model for textural property prediction calculation, and the current textural property prediction results including elasticity prediction value and hardness prediction value are obtained.

[0116] The texture analysis instrument is used to perform real-time texture detection on the prepared fish paste sample to obtain the measured texture properties, including the measured elasticity value and the measured hardness value.

[0117] The numerical difference between the current measured results of the texture properties and the current predicted results of the texture properties is calculated to obtain a set of texture property deviation values ​​including elasticity deviation and hardness deviation.

[0118] Based on the deviation data set of texture characteristics, the degree of deviation is comprehensively evaluated and calculated using a weighted average algorithm to obtain a comprehensive deviation value that reflects the overall quality fluctuation.

[0119] The comprehensive deviation value is classified and judged according to a preset threshold range to obtain a quality deviation warning index for surimi imitation. It includes slight deviation, moderate deviation and severe deviation.

[0120] Specifically, the input processing of real-time process parameters involves acquiring key parameters such as screw speed, feed rate, moisture addition, temperature at each stage, and protein denaturation level during the current preparation process through a data acquisition system. These real-time parameters are first processed by a weighted fusion algorithm to generate multi-parameter coupling coefficients for the current moment. These coupling coefficients reflect the interaction strength and direction of influence between various process parameters. The mapping relationship model receives these coupling coefficients as input variables and calculates the predicted values ​​of elastic recovery rate and maximum puncture force using built-in elastic parameter prediction models and hardness parameter prediction models, respectively. The elastic parameter prediction uses multiple linear regression calculation, multiplying the screw speed coefficient, feed rate coefficient, and moisture addition coefficient with the corresponding regression coefficients and summing the results to obtain the elasticity prediction value. The hardness parameter prediction uses polynomial regression calculation, considering the influence of the first, second, and interaction terms of the process parameter coupling coefficients, and obtaining the hardness prediction value through complex nonlinear calculations. The current textural property prediction results are output in the form of binary data pairs, including the predicted values ​​of elastic recovery rate and maximum puncture force calculated based on the current process state. These predicted values ​​represent the theoretical textural properties that the surimi-like product should possess under the current process conditions.

[0121] The texture analyzer performs real-time detection and processing, using an automated mechanical testing program to measure the texture properties of the prepared surimi-like samples in real time. The detection process includes automatic sample sampling, standardized pretreatment, and execution of standard testing procedures. Elasticity testing employs a dual compression cycle test. The probe first compresses the sample at a constant speed to a preset deformation amount, recording the force change during compression. Then, the probe returns to its initial position, and the same compression cycle is repeated. The measured elastic value is obtained by calculating the ratio of the height recovered by the sample in the second compression cycle to the initial deformation amount; this value directly reflects the sample's elastic recovery ability after deformation under stress. Hardness testing uses a puncture test method. A sharp, conical probe penetrates the sample surface at a constant speed, and the measurement system records the maximum resistance encountered during penetration. The measured hardness value is the maximum reaction force experienced by the probe during puncture, reflecting the strength and density of the internal fiber network of the sample. The current texture property measurement results are also output in binary data pairs, including the measured elastic recovery rate and the measured maximum puncture force obtained through actual physical testing. The calculation of textural property deviation values ​​quantifies the degree of deviation between the current product quality and the expected target by comparing the numerical differences between the predicted and measured results. The numerical difference calculation process separately calculates the deviations of elasticity and hardness properties. Elasticity deviation equals the absolute value of the measured elasticity value minus the absolute value of the predicted elasticity value; hardness deviation equals the absolute value of the measured hardness value minus the absolute value of the predicted hardness value. The positive or negative sign of the deviation value reflects the direction of deviation: a positive value indicates that the measured value is higher than the predicted value, and a negative value indicates that the measured value is lower than the predicted value. The absolute value reflects the severity of the deviation. The textural property deviation value data group organizes elasticity and hardness deviations into a unified data structure, facilitating subsequent comprehensive analysis and processing. The deviation value calculation also considers the impact of measurement errors and model prediction errors, distinguishing between normal fluctuations and abnormal deviations by setting reasonable deviation thresholds, avoiding erroneous quality judgments due to minor measurement errors.

[0122] The weighted average algorithm comprehensively evaluates the deviation of different textural properties based on the importance of each textural parameter to the overall product quality, assigning different weight coefficients. The weighting considers the impact of elasticity and hardness on the taste quality and consumer acceptance of the surimi-like product. Generally, elasticity has a slightly higher weight than hardness because elasticity more directly affects the chewiness and mouthfeel. The comprehensive evaluation calculation multiplies the elasticity deviation by an elasticity weight coefficient and the hardness deviation by a hardness weight coefficient, then sums the two weighted results to obtain the comprehensive deviation value. The weighted average algorithm also considers the influence of the deviation direction. When the deviation directions of two textural properties are the same, the comprehensive deviation value is larger; when the deviation directions are opposite, they may cancel each other out. The comprehensive deviation value reflects the severity of the current product's overall textural characteristics deviating from the expected target and is an important basis for judging the product quality status and determining whether process adjustments are needed.

[0123] The grading and determination of the quality deviation early warning indicators for surimi-like products determine the current product quality level by comparing the overall deviation value with a preset threshold range. The threshold range is set based on statistical analysis of historical production data and product quality standard requirements, typically using the statistical characteristics of a normal distribution to determine the numerical boundaries for each level. Slight deviation corresponds to an overall deviation value within the normal fluctuation range, indicating that the product quality basically meets expectations, with only acceptable minor differences. Moderate deviation corresponds to an overall deviation value exceeding the normal range but not reaching a severe level, indicating a significant deviation in product quality that requires attention but not immediate adjustment. Severe deviation corresponds to an overall deviation value significantly exceeding the expected range, indicating a serious product quality problem requiring immediate corrective action. The grading and determination process automatically determines the current product quality level by comparing the overall deviation value with each threshold. The early warning indicators are output with standardized level labels, providing production managers with intuitive quality status information.

[0124] In one specific embodiment, the process of executing step S105 may specifically include the following steps:

[0125] The quality deviation warning indicators of the simulated surimi were classified into three levels: slight deviation, moderate deviation and severe deviation, resulting in a three-level deviation classification result.

[0126] Based on the three-level deviation classification results, the extrusion temperature adjustment range is gradient-set to obtain a temperature correction gradient table including small, medium and large adjustment.

[0127] Based on the three-level deviation classification results, the screw speed adjustment range is subjected to gradient setting processing to obtain a speed correction gradient table including low-speed adjustment, medium-speed adjustment and high-speed adjustment;

[0128] The temperature correction gradient table and the speed correction gradient table are processed according to the current deviation level to obtain the specific combination of adjustment parameters corresponding to the current quality deviation.

[0129] The specific combination of adjustment parameters is converted into digital signal processing that can be executed by the control system, resulting in a real-time process parameter adjustment instruction set that includes temperature adjustment instructions and speed adjustment instructions.

[0130] Specifically, the three-level deviation classification process converts continuous quality deviation warning indicators into discrete level categories by establishing standardized level determination rules. The classification process first determines the numerical boundaries of each level based on the statistical distribution characteristics of the overall deviation values. A slight deviation level corresponds to an overall deviation value within one standard deviation of the mean; a moderate deviation level corresponds to a value within one to two standard deviations; and a severe deviation level corresponds to a value exceeding two standard deviations. The classification algorithm automatically identifies the level category through a logical judgment structure. When the overall deviation value is input into the system, the algorithm sequentially compares the value with the threshold values ​​of each level to determine its category. The three-level deviation classification results are output in an enumerated form, containing the level identifier and corresponding numerical range information. The classification process also considers the impact of the deviation direction, distinguishing between positive and negative deviations. Positive deviation indicates that the product's textural characteristics are too strong, requiring a reduction in relevant parameters; negative deviation indicates that the textural characteristics are insufficient, requiring an increase in relevant parameters. The classification results provide a decision-making basis for subsequent parameter adjustment strategy selection. The gradient setting of extrusion temperature adjustment is based on heat transfer theory and protein denaturation kinetics. Different levels of quality deviation correspond to different degrees of temperature adjustment requirements. The temperature correction gradient table is constructed considering the influence of temperature changes during extrusion on protein molecular structure and product texture. Small adjustments correspond to slight deviations, typically ranging from 2% to 5% of the target temperature, suitable for situations where product quality is basically acceptable but has minor deviations. Medium adjustments correspond to moderate deviations, ranging from 5% to 10% of the target temperature, suitable for situations where product quality significantly deviates from expectations but is still within acceptable limits. Large adjustments correspond to severe deviations, exceeding 10% of the target temperature, suitable for situations where product quality is severely unacceptable and requires significant correction. The temperature correction gradient table also distinguishes between different strategies for heating and cooling adjustments. Heating adjustments are achieved by increasing heating power or decreasing cooling intensity, while cooling adjustments are achieved by decreasing heating power or increasing cooling intensity. The gradient table organizes the specific temperature adjustment values ​​for each level in tabular form, providing a standardized adjustment reference for the automatic control system.

[0131] The screw speed adjustment gradient is based on mechanical shearing theory and the protein network formation mechanism. Screw speed changes directly affect the shear strength and mixing degree of the material. The screw speed correction gradient table considers the influence of screw speed on protein molecule breakage and recombination and fiber network formation. Low-speed adjustment corresponds to a slight deviation, with an adjustment range typically between 3% and 8% of the target speed, allowing for fine-tuning of the product texture. Medium-speed adjustment corresponds to a moderate deviation, with an adjustment range between 8% and 15% of the target speed, suitable for situations requiring significant changes in shear conditions to correct texture deviations. High-speed adjustment corresponds to a severe deviation, with an adjustment range exceeding 15% of the target speed, suitable for situations where the product texture deviates significantly and requires substantial changes in processing conditions. The screw speed correction gradient table also distinguishes between the different effects of acceleration and deceleration adjustments. Acceleration enhances shearing and promotes protein network formation, while deceleration reduces shear strength to avoid excessive damage to protein structure. The gradient table records the screw speed adjustment range and recommended adjustment step size for each level, ensuring the gradual and controllable nature of the adjustment process.

[0132] The parameter selection process uses a lookup algorithm to select appropriate adjustment parameters from the temperature correction gradient table and the speed correction gradient table based on the current quality deviation level. The selection algorithm first identifies the current deviation level category, then searches the gradient tables for the corresponding adjustment parameter range. Parameter selection also considers the influence of the deviation direction; when a positive deviation is detected, a strategy of decreasing the parameter is selected, and when a negative deviation is detected, a strategy of increasing the parameter is selected. Specific adjustment parameter combinations include the value, direction, and execution time of temperature adjustment, as well as the value, direction, and execution time of speed adjustment. The generation of parameter combinations also considers the coordination of temperature and speed adjustments to avoid the two parameters canceling each other out. The selection process uses an optimization algorithm to determine the optimal parameter combination. By evaluating the expected impact of different combinations on the improvement of textural properties, the parameter settings that can most effectively correct the current quality deviation are selected. The adjustment parameter combination is output in the form of structured data, including parameter identifier, adjustment value, adjustment direction, and execution priority. Digital signal conversion processing converts the abstract adjustment parameter combination into a standardized instruction format that the control system can recognize and execute. The conversion process first transforms temperature regulation parameters into heater power regulation signals and cooling system control signals. Temperature regulation commands include control parameters such as target temperature setpoint, regulation rate, and holding time. Speed ​​regulation parameters are converted into drive motor speed control signals, with speed regulation commands including target speed setpoint, acceleration limit, and operating mode. Signal conversion employs standard industrial control protocols to ensure that regulation commands can be correctly received and executed by the PLC control system or DCS distributed control system. The real-time process parameter regulation command set organizes all control commands in data packets, including fields such as command type, execution priority, timestamp, and checksum. Command set generation also includes a safety check mechanism to verify that regulation parameters are within the equipment's safe operating range, preventing over-regulation from damaging the equipment. The regulation command set is transmitted to the control system via a communication interface. The control system automatically executes the corresponding parameter adjustment operations based on the command content, achieving real-time optimized control of the extrusion process.

[0133] The above describes the data analysis-based monitoring method for the preparation of surimi protein in this application. The following describes the data analysis-based monitoring system for the preparation of surimi protein in this application. Please refer to [link to relevant documentation]. Figure 2 One embodiment of the data analysis-based monitoring system for the preparation of surimi protein in this application includes:

[0134] The detection module is used to perform real-time detection and processing of the vegetable protein melt in the extruder using multi-point temperature sensors and a near-infrared spectrometer, and to obtain an extrusion state data matrix containing temperature distribution and protein denaturation degree.

[0135] The analysis module is used to perform correlation analysis on screw speed, feed rate and moisture addition amount parameters based on the extrusion state data matrix and through a weighted fusion algorithm to obtain a multi-parameter coupling relationship coefficient table;

[0136] The modeling module combines the multi-parameter coupling coefficient table with the elasticity and hardness data detected by the texture analyzer, and uses a regression analysis algorithm to model the formation law of the imitation surimi fiber structure, thereby obtaining a mapping relationship model between texture properties and process parameters.

[0137] The prediction module is used to predict and calculate the deviation of textural properties in the current preparation process based on the mapping relationship model and the deviation compensation algorithm, so as to obtain the early warning index of fish paste quality deviation.

[0138] The correction module is used to perform gradient correction processing on the extrusion temperature and screw speed through a graded control strategy based on the numerical range of the early warning index for the quality deviation of the surimi, so as to obtain a set of real-time process parameter adjustment instructions.

[0139] above Figure 2 The data analysis-based mimicry protein preparation monitoring system in this embodiment of the invention is described in detail from the perspective of modular functional entities. The data analysis-based mimicry protein preparation monitoring device in this embodiment of the invention is described in detail from the perspective of hardware processing.

[0140] Reference Figure 3 This invention also provides a data analysis-based monitoring device for the preparation of surimi protein. This data analysis-based monitoring device can be a server, and its internal structure can be as follows: Figure 3 As shown, the data analysis-based mimicry protein preparation monitoring device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor, designed as a computer, provides computational and control capabilities. The memory of the data analysis-based mimicry protein preparation monitoring device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the data analysis-based mimicry protein preparation monitoring device stores the data corresponding to this embodiment. The network interface of the data analysis-based mimicry protein preparation monitoring device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0141] Those skilled in the art will understand that Figure 3The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the data analysis-based monitoring device for the preparation of surimi protein.

[0142] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the data analysis-based mimicry protein preparation monitoring method.

[0143] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0144] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a data analysis-based mimicry protein preparation monitoring device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0145] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data analysis-based monitoring method for preparing fish-mimicking surimi proteins, characterized by, The method comprises: Real-time detection and processing of plant protein melt in the extruder by multi-point temperature sensor and near-infrared spectrometer to obtain an extrusion state data matrix containing temperature distribution and protein denaturation degree; According to the extrusion state data matrix, the correlation analysis and processing of the screw rotation speed, the feeding speed and the water addition amount parameters are carried out by a weighted fusion algorithm to obtain a multi-parameter coupling relationship coefficient table; The multi-parameter coupling relationship coefficient table is combined with the elasticity and hardness data detected by the texture analyzer, and the modeling processing of the fiber structure formation law of the surimi-like product is carried out by a regression analysis algorithm to obtain a mapping relationship model of texture characteristics and process parameters, including: compression elasticity test and hardness penetration test of the surimi-like product sample based on the texture analyzer to obtain a texture characteristic data set containing elasticity recovery rate and maximum penetration force; the texture characteristic data set and the multi-parameter coupling relationship coefficient table are matched according to the sample batch to obtain a correlation data table of texture characteristics and coupling coefficients; according to the correlation data table, the function relationship between the elasticity recovery rate and the process parameter coupling coefficient is fitted and calculated by a multiple linear regression algorithm to obtain an elasticity parameter prediction model; based on the correlation data table, the function relationship between the maximum penetration force and the process parameter coupling coefficient is fitted by a polynomial regression algorithm to obtain a hardness parameter prediction model; the elasticity parameter prediction model and the hardness parameter prediction model are integrated in a mathematical framework to obtain a mapping relationship model describing the quantitative relationship between texture characteristics and process parameters; Based on the mapping relationship model, the deviation value of the texture characteristics in the current preparation process is predicted and calculated by a deviation compensation algorithm to obtain a surimi-like product quality deviation early warning index; According to the numerical range of the surimi-like product quality deviation early warning index, the extrusion temperature and the screw rotation speed are gradient corrected by a hierarchical control strategy to obtain a real-time process parameter adjustment instruction set.

2. The data analysis-based surimi protein preparation monitoring method according to claim 1, characterized by, The real-time detection and processing of plant protein melt in the extruder by multi-point temperature sensor and near-infrared spectrometer to obtain an extrusion state data matrix containing temperature distribution and protein denaturation degree comprises: Thermocouple temperature sensors are arranged at the feeding section, compression section, shearing section, cooling section and discharge section of the extruder barrel to collect high-frequency temperature data of each section and obtain a multi-section temperature distribution vector; The multi-section temperature distribution vector is arranged and stored in time sequence to obtain a temperature change time sequence database; The absorption peak intensity of the amide first band and the amide second band is scanned and detected by the near-infrared spectrometer to obtain protein secondary structure change spectrum data; According to the relative content changes of the helical structure and the folded structure in the spectrum data, the protein denaturation degree is quantitatively calculated to obtain a denaturation degree percentage value; The multi-section temperature distribution vector and the denaturation degree percentage value are combined and arranged in matrix form to obtain an extrusion state data matrix containing temperature and denaturation information.

3. The data analysis-based surimi protein preparation monitoring method according to claim 1, characterized by, The multi-parameter coupling relationship coefficient table is obtained by correlativity analysis and processing of the screw rotation speed, the feeding speed and the moisture addition amount parameters through a weighted fusion algorithm according to the extrusion state data matrix, including: The temperature distribution vector and the denaturation percentage value in the extrusion state data matrix are segmented and intercepted according to a time window to obtain an extrusion state sub-matrix in a time period; The process parameter real-time monitoring data set is obtained by synchronous acquisition and processing of the screw rotation speed, the feeding speed and the moisture addition amount parameters based on an encoder and a flowmeter; The parameter synchronous correlation data table is obtained by corresponding matching processing of the process parameter real-time monitoring data set and the extrusion state sub-matrix in a time period according to a time stamp; The mutual influence degree between the process parameters is quantitatively analyzed and processed by a Pearson correlation coefficient calculation method according to the parameter synchronous correlation data table to obtain a parameter correlation value matrix; The multi-parameter coupling relationship coefficient table containing coupling strength and influence direction is obtained by weight distribution and coefficient calibration processing of the parameter correlation value matrix according to influence intensity.

4. The data analysis-based surimi protein preparation monitoring method according to claim 1, characterized by, The elastic parameter prediction model is obtained by fitting calculation and processing of the function relationship between the elastic recovery rate and the process parameter coupling coefficient through a multivariate linear regression algorithm according to the correlation data table, including: The elastic recovery rate value in the correlation data table is extracted as a dependent variable to obtain an elastic recovery rate target data vector; The process parameter coupling coefficient in the correlation data table is matrix constructed as an independent variable to obtain an independent variable feature matrix containing a screw rotation speed coefficient, a feeding speed coefficient and a moisture addition amount coefficient; The regression coefficient vector of the influence of each process parameter on the elastic recovery rate is obtained by least square method calculation processing of the independent variable feature matrix; The model fitting accuracy is verified and calculated by a residual analysis method according to the regression coefficient vector to obtain model evaluation indexes containing a determination coefficient and a standard error; The elastic parameter prediction model for predicting the elastic recovery rate is obtained by mathematical expression packaging processing of the regression coefficient vector and the model evaluation indexes.

5. The data analysis-based surimi protein preparation monitoring method according to claim 1, characterized by, The texture characteristic deviation early warning index is obtained by deviation compensation algorithm prediction calculation processing of the texture characteristic deviation value in the current preparation process based on the mapping relationship model, including: The real-time process parameters in the current preparation process are input into the mapping relationship model for texture characteristic prediction calculation processing to obtain a current texture characteristic prediction result containing an elastic prediction value and a hardness prediction value; The current prepared surimi imitation fish sample is detected in real time by a texture analyzer to obtain a current texture characteristic measured result containing a measured elastic value and a measured hardness value; The current texture characteristic measured result and the current texture characteristic prediction result are calculated by a numerical difference to obtain a texture characteristic deviation value data group containing an elastic deviation degree and a hardness deviation degree; The comprehensive deviation degree value reflecting the overall quality fluctuation is obtained by comprehensive evaluation calculation processing of the deviation degree through a weighted average algorithm according to the texture characteristic deviation value data group; The texture characteristic deviation early warning index is obtained by deviation compensation algorithm prediction calculation processing of the texture characteristic deviation value in the current preparation process based on the mapping relationship model, including: The real-time process parameters in the current preparation process are input into the mapping relationship model for texture characteristic prediction calculation processing to obtain a current texture characteristic prediction result containing an elastic prediction value and a hardness prediction value; The current prepared surimi imitation fish sample is detected in real time by a texture analyzer to obtain a current texture characteristic measured result containing a measured elastic value and a measured hardness value; The current texture characteristic measured result and the current texture characteristic prediction result are calculated by a numerical difference to obtain a texture characteristic deviation value data group containing an elastic deviation degree and a hardness deviation degree; The comprehensive deviation degree value reflecting the overall quality fluctuation is obtained by comprehensive evaluation calculation processing of the deviation degree through a weighted average algorithm according to the texture characteristic deviation value data group; The comprehensive deviation value is classified and determined according to a preset threshold range to obtain an imitation surimi quality deviation early warning index including slight deviation, moderate deviation and severe deviation.

6. The data analysis-based surimi protein preparation monitoring method according to claim 1, characterized by, According to the numerical range of the imitation surimi quality deviation early warning index, gradient correction processing is performed on the extrusion temperature and screw rotation speed through a hierarchical regulation strategy to obtain a real-time process parameter adjustment instruction set, including: The imitation surimi quality deviation early warning index is classified and processed according to the levels of slight deviation, moderate deviation and severe deviation to obtain a three-level deviation level classification result; According to the three-level deviation level classification result, gradient setting processing is performed on the extrusion temperature adjustment amplitude to obtain a temperature correction gradient table including small amplitude adjustment, medium amplitude adjustment and large amplitude adjustment; Based on the three-level deviation level classification result, gradient setting processing is performed on the screw rotation speed adjustment amplitude to obtain a rotation speed correction gradient table including low speed adjustment, medium speed adjustment and high speed adjustment; The temperature correction gradient table and the rotation speed correction gradient table are parameter selected according to the current deviation level to obtain a specific adjustment parameter combination corresponding to the current quality deviation; The specific adjustment parameter combination is converted into a digital signal processing executable by a control system to obtain a real-time process parameter adjustment instruction set including temperature adjustment instructions and rotation speed adjustment instructions.

7. A data analysis-based surimi protein production monitoring system, characterized by, A data analysis-based imitation surimi protein preparation monitoring system for implementing the data analysis-based imitation surimi protein preparation monitoring method according to any one of claims 1-6, the data analysis-based imitation surimi protein preparation monitoring system comprising: A detection module for performing real-time detection processing on the plant protein melt in the extruder through a multi-point temperature sensor and a near-infrared spectrometer to obtain an extrusion state data matrix including temperature distribution and protein denaturation degree; An analysis module for performing correlation analysis processing on the screw rotation speed, feed speed and water addition amount parameters through a weighted fusion algorithm based on the extrusion state data matrix to obtain a multi-parameter coupling relationship coefficient table; A modeling module for combining the multi-parameter coupling relationship coefficient table with the elasticity and hardness data detected by a texture analyzer to perform modeling processing on the imitation surimi fiber structure formation rule through a regression analysis algorithm to obtain a mapping relationship model of texture characteristics and process parameters; A prediction module for performing prediction calculation processing on the texture characteristic deviation value in the current preparation process through a deviation compensation algorithm based on the mapping relationship model to obtain an imitation surimi quality deviation early warning index; A correction module for performing gradient correction processing on the extrusion temperature and screw rotation speed through a hierarchical regulation strategy according to the numerical range of the imitation surimi quality deviation early warning index to obtain a real-time process parameter adjustment instruction set.

8. A data analysis-based surimi protein production monitoring device, characterized by, A memory and a processor, the memory stores a computer program executable on the processor, and the processor implements the data analysis-based imitation surimi protein preparation monitoring method according to any one of claims 1-6 when executing the computer program.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program causes the processor to execute the data analysis-based imitation surimi protein preparation monitoring method according to any one of claims 1-6 when the processor executes the computer program.

Images