Preparation control method, device and equipment for formula nutritious food and storage medium

Abstract

The invention provides a formula nutritious food preparation control method, device and equipment and a storage medium, and relates to the technical field of formula nutritious food preparation control. The method comprises the following steps: acquiring real-time particle size data in the preparation process of the formula nutritious food, and inputting the real-time particle size data into a trained osmotic pressure fitting model to obtain fitting osmotic pressure; wherein the osmotic pressure fitting is obtained based on sample characteristic particle size data corresponding to the formula nutritional food to be prepared and sample osmotic pressure training; determining a risk intervention control scheme according to the fitting osmotic pressure and a preset early warning condition; wherein process parameters in the risk intervention control scheme comprise one or more of prebiotic regulating quantity, lactose regulating quantity and homogenizing pressure; and adjusting process parameters according to the risk intervention control scheme. According to the method, the intestinal stress reaction of the product can be predicted in the production process so as to realize real-time infiltration reduction, the intestinal stress risk of the formula nutritious food caused by improper infiltration pressure is avoided, and the preparation accuracy of the formula nutritious food and the product safety are improved.

Description

Technical Field

[0001] This invention relates to the field of formulation nutritional food preparation control technology, and in particular to a formulation nutritional food preparation control method, apparatus, equipment and storage medium. Background Technology

[0002] During infancy and early childhood, the intestinal function is not yet fully developed, making it more sensitive to the osmotic pressure of exogenous foods. Ingesting foods with high osmotic pressure can easily trigger intestinal stress responses, accelerate intestinal peristalsis, and lead to clinical symptoms such as diarrhea and vomiting. The degree to which the intestines adapt to food can be reflected to some extent by indicators such as the rate of intestinal motility promotion after eating; this rate is negatively correlated with intestinal comfort.

[0003] In the research and development and production of infant formula, osmotic pressure control is a key factor affecting the product's intestinal adaptability. Current technologies still face several significant bottlenecks, which can be summarized in the following five aspects: First, electrolyte composition often uses fixed ratios, failing to dynamically optimize according to the intestinal development of infants at different ages, thus lacking individualized adaptation capabilities; Second, the application ratio of deeply hydrolyzed proteins is not yet reasonably balanced. While reducing allergenicity, the increased number of short peptide chains significantly raises osmotic pressure, creating a technical contradiction between "desensitization" and "osmotic control"; Third, the addition of prebiotics is generally too low, failing to reach the concentration threshold for effectively promoting intestinal health, resulting in insufficient functional realization; Fourth, current assessments of intestinal discomfort caused by milk powder rely heavily on post-consumer feedback, leading to delayed identification, which not only affects user experience but also damages brand trust.

[0004] Currently, there is a lack of a scheme to predict the intestinal stress response of products during the production process in order to achieve real-time desaturation. Summary of the Invention

[0005] This invention provides a method, apparatus, equipment, and storage medium for controlling the preparation of formulated nutritional foods, in order to solve the problem of predicting intestinal stress response of products during the production process to achieve real-time osmosis reduction.

[0006] In a first aspect, embodiments of the present invention provide a method for controlling the preparation of formulated nutritional foods, comprising: Obtain real-time particle size data during the preparation of formulated nutritional foods; The real-time particle size data is input into the trained osmotic pressure fitting model to obtain the fitted osmotic pressure; wherein, the osmotic pressure fitting model is trained based on the sample characteristic particle size data and sample osmotic pressure corresponding to the nutritional food formula to be prepared; Based on the fitted osmotic pressure and preset early warning conditions, a risk intervention and control scheme is determined; wherein, the process parameters in the risk intervention and control scheme include one or more of the following: prebiotic adjustment amount, lactose adjustment amount, and homogenization pressure; Adjust process parameters according to the aforementioned risk intervention and control scheme.

[0007] In one possible implementation, the preset early warning conditions include preset low-risk early warning conditions, medium-risk early warning conditions, and high-risk early warning conditions; The risk intervention and control plan includes: a medium-risk risk intervention and control plan and a high-risk risk intervention and control plan; The process parameters corresponding to the medium-risk intervention and control scheme include: prebiotic regulation amount and lactose regulation amount; The process parameters corresponding to the high-risk intervention and control scheme include homogenization pressure.

[0008] In one possible implementation, determining the risk intervention and control scheme based on the fitted osmotic pressure and preset early warning conditions includes: When the fitted osmotic pressure meets the medium-risk warning condition, the first difference between the fitted osmotic pressure and the upper limit of the low-risk warning condition is calculated, and the amount of prebiotics added and the amount of lactose reduced are determined based on the first difference; wherein, the amount of prebiotics added and the amount of lactose reduced are the same; When the fitted osmotic pressure meets the high-risk warning condition, the second difference between the fitted osmotic pressure and the upper limit of the low-risk warning condition is calculated, and the homogeneous pressure reduction value is determined based on the second difference.

[0009] In one possible implementation, prior to acquiring the real-time particle size data during the preparation of the formulated nutritional food, the method further includes: Multiple formulation samples of the nutritional food to be prepared were obtained, and the particle size distribution parameters and measured osmotic pressure values ​​of each formulation sample were obtained. A multivariate nonlinear regression analysis method was used, with the particle size distribution parameters as independent variables and the osmotic pressure value as the dependent variable, to construct an initial fitting model; wherein the particle size distribution parameters include D10, D50, D90, and D10. 2 D50 2 and D10×D50; The initial fitting model is subjected to feature selection, five-step iterative optimization, cross-validation, and parameter optimization to obtain the osmotic pressure fitting model.

[0010] In one possible implementation, the feature filtering includes: Determine candidate particle size characteristics associated with osmotic pressure; wherein, the candidate particle size characteristics include: D10, D50, D90, and D10. 2 D50 2 and D10×D50; Based on the significance test, weakly correlated features are removed from the candidate particle size features to obtain the first candidate particle size features; Based on the variance inflation factor, the features with moderate collinearity in the first candidate particle size features are centered so that the variance inflation factor of the processed features is ≤5. The importance of each feature after centering is calculated using the random forest algorithm, and features with importance below a set threshold are removed. Compare the performance of models with and without D10×D50, and determine the final feature set based on model fit and information criterion values.

[0011] In one possible implementation, the five-step iterative optimization includes: Test three types of linear models: ordinary least squares, ridge regression, and Lasso regression. Compare the model fit, residual standard deviation, and overfitting to determine the basic model. For the final feature set, the model performance after introducing quadratic terms is tested respectively. Quadratic terms that improve the model fit of the basic model and reduce the information criterion value are retained to form a model that combines linear and quadratic terms. Candidate interaction terms are introduced using forward stepwise regression. The significance of the interaction terms is tested, and interaction terms that improve the model fit and reduce the information criterion value are retained. We tested three regularization methods: L1 regularization, L2 regularization, and elastic network regularization. We determined the optimal regularization coefficients through grid search and selected the regularization method that minimized the validation set error and made all core feature coefficients significant.

[0012] In one possible implementation, the cross-validation includes: The stability of the model was verified by 10-fold cross-validation and leave-one-out method. Residual analysis was also performed to ensure that the residuals were normally distributed and free from systematic bias, so as to obtain the target particle size characteristics corresponding to the osmotic pressure fitting model.

[0013] In one possible implementation, the parameter optimization includes: Test 1st to 3rd order polynomial models, using validation set goodness of fit and information criterion as evaluation criteria, and determine 2nd order polynomial as model complexity; The optimal regularization coefficients are determined by a two-step search method using coarse and fine grids. After the coarse grid search locks in the range of coefficients, the fine grid search obtains the coefficient values ​​that minimize the error of the validation set, and the stability of the coefficients is verified under different data partition ratios. Construct a hyperparameter combination matrix that combines the polynomial order and regularization coefficients, calculate the validation set goodness of fit and information criterion value for each combination, select the optimal hyperparameter combination, and determine the osmotic pressure fitting model based on the optimal hyperparameter combination.

[0014] Secondly, embodiments of the present invention provide a control device for the preparation of formulated nutritional foods, comprising: The acquisition module is used to acquire real-time particle size data during the preparation of formulated nutritional foods; The osmotic pressure fitting module is used to input the real-time particle size data into the trained osmotic pressure fitting model to obtain the fitted osmotic pressure; wherein, the osmotic pressure fitting model is trained based on the sample characteristic particle size data and sample osmotic pressure corresponding to the nutritional food formula to be prepared; The determination module is used to determine a risk intervention and control scheme based on the fitted osmotic pressure and preset early warning conditions; wherein, the process parameters in the risk intervention and control scheme include one or more of the following: prebiotic adjustment amount, lactose adjustment amount, and homogenization pressure; The control module is used to adjust process parameters according to the risk intervention control scheme.

[0015] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect or any possible implementation thereof.

[0016] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect or any possible implementation thereof.

[0017] In this embodiment of the invention, real-time particle size data during the preparation of formulated nutritional foods is acquired and input into an osmotic pressure fitting model trained based on the characteristic particle size and osmotic pressure of the corresponding formulated nutritional food sample to obtain the fitted osmotic pressure. Then, combined with preset early warning conditions, a risk intervention and control scheme including process parameters such as prebiotic regulation amount, lactose regulation amount, or homogenization pressure is determined and the process parameters are adjusted. This enables real-time prediction of osmotic pressure and dynamic adjustment of the process during the preparation of formulated nutritional foods, avoiding the risk of intestinal stress in formulated nutritional foods caused by improper osmotic pressure, improving the accuracy of formulated nutritional food preparation and product safety, and solving the defects of traditional control methods that rely on delayed feedback. Attached Figure Description

[0018] Figure 1 This is an application scenario diagram of the formula nutritional food preparation control method provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the implementation of a method for controlling the preparation of formulated nutritional foods according to an embodiment of the present invention. Figure 3 This is a flowchart illustrating the implementation of a method for controlling the preparation of formulated nutritional foods according to another embodiment of the present invention. Figure 4 This is a schematic diagram of the structure of a formula nutritional food preparation control device provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0019] The preparation and control method of the formulated nutritional food provided in the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Figure 1 This is an application scenario diagram of the method for controlling the preparation of formulated nutritional foods according to an embodiment of the present invention. For example... Figure 1 As shown, it includes a laser particle size analyzer, a control terminal, and equipment for preparing formulated nutritional foods.

[0021] In the specific implementation process, a laser particle size analyzer is used to detect real-time particle size data during the preparation of formulated nutritional foods. The control terminal communicates with the laser particle size analyzer via wired or wireless means to acquire real-time particle size data. Based on the real-time particle size data and the osmotic pressure fitting model, the fitted osmotic pressure is obtained. Based on the fitted osmotic pressure and preset early warning conditions, a risk intervention and control plan is determined. Under the guidance of the risk intervention and control plan, the formulated nutritional food preparation equipment adjusts one or more process parameters, including prebiotic regulation amount, lactose regulation amount, and homogenization pressure, to achieve real-time osmotic reduction during the preparation of formulated nutritional foods and avoid the risk of intestinal stress in formulated nutritional foods caused by improper osmotic pressure.

[0022] Figure 1 In the illustrated embodiment, the control terminal is set up independently of the formulated nutritional food preparation equipment, facilitating remote control by staff. In other possible implementations, the laser particle size analyzer communicates directly with the controller installed within the formulated nutritional food preparation equipment system, eliminating the need for an additional control terminal.

[0023] Figure 2 This is a flowchart illustrating the implementation of a method for controlling the preparation of formulated nutritional foods according to an embodiment of the present invention. Figure 2 As shown, it includes the following steps: S201, Obtain real-time particle size data during the preparation of formulated nutritional foods.

[0024] The execution entity in various embodiments of the present invention can be a server, processor, microprocessor, or other device with data processing capabilities. In actual implementation, the specific implementation method of the execution entity can be selected according to actual needs. This embodiment does not impose any particular limitation on this; any device with data processing capabilities is acceptable. For ease of understanding, in subsequent embodiments, [the following will be used as an example]. Figure 1 The control terminal shown is the execution entity for explanation.

[0025] In actual implementation, an online laser particle size analyzer is installed to collect real-time particle size data of the finished formula nutritional food samples flowing through the pipeline. Optionally, this device is installed at a key production node after the finished infant formula nutritional food is mixed and before packaging, and is connected to the material transfer pipeline of the formula nutritional food production line.

[0026] To improve detection efficiency, the online laser particle size analyzer samples real-time particle size data at a set sampling frequency. For example, a sampling frequency of once every 5 minutes is used to detect characteristic particle size parameters (including D10 and D50) of the finished formulated nutritional food product. Furthermore, each detection generates multiple sets of parallel data. After the average value is calculated by the device's built-in data processing module, this average value is used as the real-time particle size data and synchronously transmitted to the central processing unit of the formulated nutritional food production control system. Optionally, 3 to 5 sets of parallel data are generated per detection.

[0027] S202, input the real-time particle size data into the trained osmotic pressure fitting model to obtain the fitted osmotic pressure; wherein, the osmotic pressure fitting model is trained based on the sample characteristic particle size data and sample osmotic pressure corresponding to the nutritional food formula to be prepared.

[0028] Formulated nutritional food production lines can be used to prepare formulated nutritional foods for different age groups. Therefore, the osmotic pressure fitting model differs for each age group. Taking infant formula milk powder for children aged 0-6 months as an example, the training process is based on different batches of laboratory samples of this formula milk powder: First, multiple groups of formula milk powder samples are prepared through laboratory simulation of the production process. The characteristic particle size data and corresponding measured osmotic pressure values ​​of each group of samples after reconstitution are measured. Then, a multivariate nonlinear regression analysis method is used, with the characteristic particle size data as the independent variable and the measured osmotic pressure value as the dependent variable. The model is constructed after feature selection, iterative optimization, and cross-validation.

[0029] In actual implementation, after receiving real-time particle size data, the control terminal automatically inputs the data into the aforementioned osmotic pressure fitting model. The model processes the real-time particle size data through its built-in calculation logic, outputs the corresponding fitted osmotic pressure value, and performs subsequent risk assessment based on the fitted osmotic pressure value.

[0030] S203. Based on the fitted osmotic pressure and preset early warning conditions, determine the risk intervention and control plan; wherein, the process parameters in the risk intervention and control plan include one or more of the following: prebiotic regulation amount, lactose regulation amount, and homogenization pressure.

[0031] In the control terminal, pre-set osmotic pressure warning conditions corresponding to the nutritional food formula to be prepared. Taking infant formula milk powder for 0-6 months as an example, the warning conditions are determined based on the osmotic pressure range of breast milk (260~300mOsmol / kg) and infant intestinal stress risk test data, with the fitted osmotic pressure value as the core judgment indicator.

[0032] The following explanation uses preset warning conditions, including low-risk, medium-risk, and high-risk warning conditions, as an example. If the fitted osmotic pressure value is in the low-risk range (≤290 mOsmol / kg), the current milk powder osmotic pressure is determined to meet intestinal tolerance requirements, and a routine production plan without adjusting process parameters is determined. If the fitted osmotic pressure value is in the medium-risk range (290~320 mOsmol / kg), a slight risk of intestinal stress is determined, and a medium-risk intervention and control plan including prebiotic and lactose regulation is determined, reducing osmotic pressure by adjusting the ratio of prebiotics to lactose. If the fitted osmotic pressure value is in the high-risk range (>320 mOsmol / kg), a high risk of intestinal stress is determined, and a high-risk intervention and control plan including homogenization pressure adjustment is determined, reducing osmotic pressure by adjusting homogenization process parameters to change the milk powder particle size distribution.

[0033] In other possible implementation methods, the preset early warning conditions can be further refined, for example, divided into safe conditions, low-risk early warning conditions, medium-risk early warning conditions, and high-risk early warning conditions.

[0034] S204, Adjust process parameters according to the risk intervention and control plan.

[0035] In the actual implementation process, the control terminal transforms the determined risk intervention and control plan into specific process control instructions and sends them to the corresponding formula nutritional food preparation equipment control system to adjust the process parameters.

[0036] In this embodiment, by acquiring real-time particle size data during the preparation of formulated nutritional foods, and inputting it into an osmotic pressure fitting model trained based on the characteristic particle size and osmotic pressure of the corresponding formulated nutritional food sample, a fitted osmotic pressure is obtained. Then, combined with preset early warning conditions, a risk intervention and control scheme including process parameters such as prebiotic regulation amount, lactose regulation amount, or homogenization pressure is determined and the process parameters are adjusted. This enables real-time prediction of osmotic pressure and dynamic adjustment of the process during the preparation of formulated nutritional foods, avoiding the risk of intestinal stress in formulated nutritional foods caused by improper osmotic pressure, improving the accuracy of formulated nutritional food preparation and product safety, and solving the defects of traditional control methods that rely on delayed feedback.

[0037] Figure 3 This is a flowchart illustrating the implementation of a method for controlling the preparation of formulated nutritional foods according to another embodiment of the present invention. Figure 3As shown. In one possible implementation, the preset early warning conditions include preset low-risk, medium-risk, and high-risk early warning conditions; Risk intervention and control plans include: medium-risk intervention and control plans and high-risk intervention and control plans; Among them, the process parameters corresponding to the medium-risk intervention and control plan include: prebiotic regulation amount and lactose regulation amount; The process parameters corresponding to the high-risk intervention and control plan include homogenization pressure.

[0038] In the specific implementation process, if the fitted osmotic pressure meets the low-risk warning conditions, the conventional production plan will be maintained. The control system of the formula nutritional food preparation equipment will maintain the current operating parameters of the prebiotic addition device, lactose addition device and homogenizer, and the production of formula nutritional food will continue according to the original process.

[0039] If the fitted osmotic pressure meets the medium-risk warning conditions, a medium-risk intervention and control plan is adopted. The prebiotic addition device increases the amount of prebiotic added per unit time according to the control command, and the lactose addition device simultaneously reduces the amount of lactose added per unit time to ensure the stability of total carbohydrate content. After adjustment, the real-time particle size data and fitted osmotic pressure value of the finished formula nutritional food are continuously monitored until the fitted osmotic pressure falls back to the low-risk range. If the fitted osmotic pressure meets the high-risk warning conditions, a high-risk intervention and control scheme is adopted. The homogenizer control system lowers the homogenization pressure according to the control command, adjusting the particle size distribution by changing the dispersion of the formulated nutritional food particles, thereby affecting the osmotic pressure. Optionally, during the adjustment process, real-time particle size data is collected at a set frequency (e.g., once every 2 minutes) and the fitted osmotic pressure is calculated until the fitted osmotic pressure drops to the low-risk range. Subsequently, the homogenization pressure is stabilized at the adjusted parameter value to ensure that the osmotic pressure of the subsequently produced formulated nutritional food meets the requirements.

[0040] In this embodiment, the method for controlling the preparation of formulated nutritional foods divides preset early warning conditions into three categories: low-risk, medium-risk, and high-risk. Correspondingly, it sets up a medium-risk intervention control scheme that includes prebiotic regulation and lactose regulation, and a high-risk intervention control scheme that includes homogenization pressure. This makes risk intervention more targeted, avoids the waste of resources or insufficient intervention that may be caused by a uniform intervention scheme, further refines the risk control levels in the preparation process of formulated nutritional foods, and improves the adaptability and effectiveness of process adjustments under different risk levels.

[0041] In one possible implementation, a risk intervention and control scheme is determined based on the fitted osmotic pressure and preset early warning conditions, including: When the fitted osmotic pressure meets the medium-risk warning condition, the first difference between the fitted osmotic pressure and the upper limit of the low-risk warning condition is calculated, and the amount of prebiotics added and the amount of lactose reduced are determined based on the first difference; wherein, the amount of prebiotics added and the amount of lactose reduced are the same. When the fitted osmotic pressure meets the high-risk warning condition, calculate the second difference between the fitted osmotic pressure and the upper limit of the low-risk warning condition, and determine the homogeneous pressure reduction value based on the second difference.

[0042] In the specific implementation process, the osmotic pressure range of the medium-risk warning condition is 290~320mOsmol / kg, and the upper limit of the low-risk warning condition (290mOsmol / kg) is used as the benchmark for calculating the first difference; the osmotic pressure range of the high-risk warning condition is >320mOsmol / kg, and the upper limit of the low-risk warning condition (290mOsmol / kg) is used as the benchmark for calculating the second difference.

[0043] The control terminal compares the fitted osmotic pressure value with the preset warning conditions to determine the warning level. If the fitted osmotic pressure value is in the range of 290~320 mOsmol / kg, it is determined to be a medium-risk warning; if the fitted osmotic pressure value is >320 mOsmol / kg, it is determined to be a high-risk warning. When a medium-risk warning is determined (e.g., the fitted osmotic pressure value is 305 mOsmol / kg), the control terminal automatically calculates the first difference. The amount of prebiotics added is calculated according to Δosmotic pressure = K1 × (prebiotic addition amount / kg). Where K1 is an adjustment coefficient, which is optional, K1 = -2.8 mOsmol / kg / kg.

[0044] The first difference (i.e., Δosmotic pressure) = fitted osmotic pressure value - upper limit of low-risk warning conditions, i.e., 305mOsmol / kg - 290mOsmol / kg = 15mOsmol / kg, so the amount of prebiotic added is 5.4kg. According to the system's built-in correlation logic, the amount of prebiotic added is the same as the amount of lactose reduced.

[0045] When a high-risk warning is issued (e.g., the fitted osmotic pressure value is 335 mOsmol / kg), the central processing unit automatically calculates the second difference. The homogeneous pressure reduction value is calculated based on Δosmotic pressure = K2 × (homogeneous pressure increment / bar), where K2 is an adjustment coefficient, optional, K2 = -2.5 mOsmol / kg / bar.

[0046] The second difference = fitted osmotic pressure value - upper limit of low-risk warning conditions, i.e. 335mOsmol / kg - 290mOsmol / kg = 45mOsmol / kg, then the homogeneous pressure increment is 18 bar.

[0047] In this embodiment, when the fitted osmotic pressure meets the medium-risk warning condition, the equivalent amount of prebiotic addition and lactose reduction is determined by calculating the first difference between the fitted osmotic pressure and the upper limit of the low-risk warning condition. When the high-risk warning condition is met, the homogenization pressure reduction value is determined by calculating the second difference between the fitted osmotic pressure and the upper limit of the low-risk warning condition. This directly links the adjustment amount of process parameters to the osmotic pressure deviation, avoiding the blindness of adjustment amounts and ensuring that the adjustment of prebiotics, lactose, and homogenization pressure can accurately match the osmotic pressure control requirements, thus improving the accuracy of risk intervention. Furthermore, when the fitted osmotic pressure meets the high-risk warning condition, determining the homogenization pressure reduction value based on the upper limit of the low-risk warning condition allows for the fastest possible adjustment of osmotic pressure to ensure that the osmotic pressure drops to a reasonable range.

[0048] Other possible implementation methods include determining risk intervention and control schemes based on the fitted osmotic pressure and preset early warning conditions, including: When the fitted osmotic pressure meets the medium-risk warning condition, the first difference between the fitted osmotic pressure and the upper limit of the low-risk warning condition is calculated, and the amount of prebiotics added and the amount of lactose reduced are determined based on the first difference; wherein, the amount of prebiotics added and the amount of lactose reduced are the same. When the fitted osmotic pressure meets the high-risk warning condition, calculate the second difference between the fitted osmotic pressure and the upper limit of the medium-risk warning condition, and determine the homogeneous pressure reduction value based on the second difference.

[0049] In this embodiment, when the fitted osmotic pressure meets the high-risk warning condition, determining the homogeneous pressure reduction value based on the upper limit of the medium-risk warning condition can pull the system back from the brink of danger, thereby updating the fitted osmotic pressure. Further adjustments to prebiotics and lactose are then determined based on the fitted osmotic pressure and the warning condition. Determining the homogeneous pressure reduction value based on the fitted osmotic pressure and the upper limit of the medium-risk warning condition makes the control process smoother and effectively avoids drastic parameter fluctuations.

[0050] The above mainly introduced how to adjust process parameters based on real-time particle size data. The following describes the training process of the osmotic pressure fitting model.

[0051] In one possible implementation, before acquiring real-time particle size data during the preparation of the formulated nutritional food, the method further includes: Multiple formulation samples of the nutritional food to be prepared were obtained, and the particle size distribution parameters and measured osmotic pressure values ​​of each formulation sample were obtained. A multivariate nonlinear regression analysis method was used, with particle size distribution parameters as independent variables and osmotic pressure as the dependent variable, to construct an initial fitting model; the particle size distribution parameters included D10, D50, D90, and D10. 2 D50 2 and D10×D50; The initial fitted model was subjected to feature selection, five-step iterative optimization, cross-validation, and parameter optimization to obtain the osmotic pressure fitting model.

[0052] In the specific implementation process, the osmotic pressure of the formulation samples was measured using a freezing point osmoremeter. Each formulation sample was tested in triplicate, and the average value was calculated. The characteristic particle size (including at least D10 and D50) of the formulation samples was measured using a laser particle size analyzer. Each formulation sample was tested in triplicate, and the average value was calculated. Calculating the average value improves the accuracy of the measurement results.

[0053] In this embodiment, before acquiring real-time particle size data, multiple formulation samples of the nutritional food to be prepared, along with their particle size distribution parameters and measured osmotic pressure values, are obtained. A multivariate nonlinear regression analysis is used to construct an initial fitting model with particle size distribution parameters as independent variables and osmotic pressure as the dependent variable. Through feature selection, five-step iterative optimization, cross-validation, and parameter optimization, an osmotic pressure fitting model is obtained. This ensures that the construction of the osmotic pressure fitting model is based on the actual sample data of the nutritional food to be prepared, and that the model's accuracy and reliability are improved through multiple optimization steps. This provides precise model support for subsequent real-time osmotic pressure prediction and reduces process adjustment deviations caused by model errors.

[0054] In one possible implementation, feature selection includes: Candidate particle size characteristics associated with osmotic pressure were identified; these candidate particle size characteristics included: D10, D50, D90, and D10. 2 D50 2 and D10×D50; Weakly correlated features were removed from the candidate particle size features based on significance tests to obtain the first candidate particle size features; Based on the variance inflation factor, the features with moderate collinearity in the first candidate particle size feature are centered so that the variance inflation factor of the processed feature is ≤5. The importance of each feature after centering is calculated using the random forest algorithm, and features with importance below a set threshold are removed. Compare the performance of models with and without D10×D50, and determine the final feature set based on model fit and information criterion values.

[0055] In this embodiment, the feature selection process first identifies candidate particle size features, including D10 and D50. Then, weakly correlated features are eliminated through significance testing to obtain the first candidate particle size features. Based on the variance inflation factor, features with moderate collinearity are centered to ensure feature independence. The random forest algorithm is used to eliminate features with importance below a set threshold. Finally, the model performance with and without D10×D50 is compared to determine the final feature set. This process can gradually select features that are closely related to osmotic pressure, have strong independence, and are effective in improving model performance, avoiding the interference of redundant or weakly correlated features on the model fitting accuracy and improving the accuracy of the osmotic pressure fitting model.

[0056] In one possible implementation, the five-step iterative optimization includes: Test three types of linear models: ordinary least squares, ridge regression, and Lasso regression. Compare the model fit, residual standard deviation, and overfitting to determine the basic model. For the final feature set, the model performance after introducing quadratic terms was tested respectively. Quadratic terms that improved the model fit of the basic model and reduced the information criterion value were retained to form a model that combines linear and quadratic terms. Candidate interaction terms are introduced using forward stepwise regression. The significance of the interaction terms is tested, and interaction terms that improve the model fit and reduce the information criterion value are retained. We tested three regularization methods: L1 regularization, L2 regularization, and elastic network regularization. We determined the optimal regularization coefficients through grid search and selected the regularization method that minimized the validation set error and made all core feature coefficients significant.

[0057] In this embodiment, the five-step iterative optimization process first determines the basic model by testing three types of linear models. Then, it tests the introduction of quadratic terms for the final feature set and retains the quadratic terms that improve model performance to form a model combining linear and quadratic terms. Next, it introduces and retains significant interaction terms through forward stepwise regression. Finally, it tests three types of regularization methods and determines the optimal regularization coefficient and method. This process can optimize the model structure and parameters step by step, gradually improve the model's ability to fit osmotic pressure, control the risk of overfitting, and ensure that the model has good generalization ability while maintaining good fit, thereby improving the reliability of the osmotic pressure fitting model.

[0058] In one possible implementation, cross-validation includes: The stability of the model was verified by 10-fold cross-validation and leave-one-out method. Residual analysis was also performed to ensure that the residuals were normally distributed and free from systematic bias, so as to obtain the target particle size characteristics corresponding to the osmotic pressure fitting model.

[0059] In this embodiment, the cross-validation process uses 10-fold cross-validation and leave-one-out cross-validation to verify the model's stability. Residual analysis is also performed to ensure that the residuals are normally distributed and free from systematic bias. This allows for verification of the model's performance consistency across different data subsets from various perspectives, avoiding instability issues caused by data partitioning bias. Furthermore, residual analysis eliminates the impact of systematic bias on model accuracy, ultimately determining the target particle size characteristics corresponding to the osmotic pressure fitting model and ensuring the model's stability and accuracy when used for real-time osmotic pressure prediction.

[0060] In one possible implementation, parameter optimization includes: Test 1st to 3rd order polynomial models, using validation set goodness of fit and information criterion as evaluation criteria, and determine 2nd order polynomial as model complexity; The optimal regularization coefficients are determined by a two-step search method using coarse and fine grids. After the coarse grid search locks in the range of coefficients, the fine grid search obtains the coefficient values ​​that minimize the error of the validation set, and the stability of the coefficients is verified under different data partition ratios. A hyperparameter combination matrix combining polynomial order and regularization coefficients is constructed. The goodness of fit of the validation set and the information criterion value of each combination are calculated. The optimal hyperparameter combination is selected to determine the osmotic pressure fitting model based on the optimal hyperparameter combination.

[0061] In this embodiment, the parameter optimization process first tests 1st to 3rd order polynomial models and determines the 2nd order polynomial as the model complexity using the validation set goodness of fit and information criterion values. Then, the optimal regularization coefficient is determined through a two-step search method using coarse and fine grids, and its stability is verified. Finally, a hyperparameter combination matrix is ​​constructed to select the optimal hyperparameter combination to determine the osmotic pressure fitting model. This method can accurately determine the polynomial complexity and regularization coefficient of the model, ensuring the model's fitting ability while avoiding overfitting or underfitting. Furthermore, the selection of hyperparameter combinations ensures that the chosen hyperparameter combination optimizes the model's performance, further improving the accuracy and generalization ability of the osmotic pressure fitting model and providing more reliable osmotic pressure prediction results for subsequent adjustments to the formulation and preparation process of nutritional foods.

[0062] The following describes the training process of the osmotic pressure fitting model in the above embodiments using a specific example: 1. Multi-round feature selection and interactive verification Based on theoretical assumptions (the influence of particle size distribution on osmotic pressure), candidate particle size features corresponding to the initial fitting model were determined, and six candidate particle size features were initially included: D10, D50, D90 (particle size parameter), D10 2 D50 2 (Quadratic term) and D10×D50 (interaction term) form the initial feature pool.

[0063] Round 1: Single Feature Significance Screening The t-test was used to analyze the linear correlation between each feature and osmotic pressure, and weakly correlated features with p>0.05 were removed. The p-value of D90 was 0.12 (not significant), and the Pearson correlation coefficient r with D50 was 0.91 (risk of multicollinearity), so it was removed; the remaining features (D10, D50, D10) 2 D50 2 If both D10×D50 satisfy p<0.01, they are retained for the next round.

[0064] Second round: Multicollinearity test Calculate the variance inflation factor (VIF) among features to determine the degree of collinearity: D10 and D10 2 The VIF was 8.7 (>5, indicating moderate collinearity), which was reduced to VIF 3.2 through "centering" (subtracting the mean from the eigenvalues); D50 and D50 2 The VIF was 9.1, which was also reduced to VIF 2.9 through centralization to ensure feature independence.

[0065] Round 3: Feature Importance Ranking and Redundancy Removal The Mean Decrease Accuracy (MDA) of features was calculated using a random forest model and ranked by importance: D50 2 (MDA=0.32)>D10×D50 (MDA=0.28)>D50 (MDA=0.21)>D10 (MDA=0.15)>D10 2 (MDA=0.04); Because of D10 2 The MDA value is extremely low (<0.05), and the model R after removal is also low. 2 It only decreased by 0.8% (from 72.9% to 72.1%), so D10 was removed. 2 Ultimately, the core features are retained: D10, D50, and D50. 2 D10×D50.

[0066] Fourth round: Validation of interaction items Compare the performance of models with and without interactive elements: Without D10×D50 interaction terms, model R 2 =72.1%, AIC=128.5; After adding the D10×D50 interactive item, R 2The efficiency was increased to 79.8%, and the AIC decreased to 116.2 (a decrease in AIC of ≥10 indicates that the interaction terms significantly improve the model's explanatory power). The final feature set {D10, D50, D50} was determined. 2 The numbers D10×D50 lay the foundation for subsequent model construction.

[0067] 2. Five-step iterative optimization and multi-dimensional cross-validation A five-step optimization method was adopted, consisting of "basic model → step-by-step upgrade → risk control → stability verification". The optimal direction was determined at each step through comparative experiments and error analysis. The specific parameter tuning process is as follows: Step 1: Optimization of the basic linear model (determining the baseline) Test three types of linear models (Ordinary Least Squares (OLS), Ridge Regression, and Lasso Regression) and compare key metrics: OLS model: R 2 =63.5%, residual standard deviation =18.2mOsmol / kg, indicating slight overfitting (training set R). 2 =68.3%, test set R 2 =63.5%, difference 4.8%); Ridge regression (λ=0.05): test set R 2 =64.2%, residual standard deviation =17.8mOsmol / kg, overfitting decreased (difference 3.2%); Lasso regression (λ=0.05): D10 features removed (coefficients compressed to 0), R 2 Performance decreased to 59.8%. Based on the above, ridge regression was chosen as the basic model, and R... 2 =64.2% is the initial baseline. Step 2: Introduce quadratic terms for optimization (to improve nonlinear fitting ability) For the retained features (D10, D50), test the quadratic terms (D10) one by one. 2 D50 2 The effect of introducing ) Add only D10 2 Model R 2 =65.7%, AIC=132.1 (limited improvement); only add D50 2 Model R 2 =72.1%, AIC=120.3 (R 2 (Increased by 7.9%, AIC significantly reduced); D10 added simultaneously. 2 +D50 2 Model R 2 =72.3%, AIC=122.5 (R 2 The improvement is slight, AIC increases, and redundancy exists. Based on the above, only the D50 is retained. 2 The quadratic term upgrades the model to "linear + D50". 2 "R" 2 =72.1%. Step 3: Optimize by introducing interaction terms (capturing the synergistic effect of variables) The stepwise forward regression method is used to introduce interaction terms, adding one candidate interaction term (D10×D50, D10×D50) at a time. 2 ), and test its significance: Introducing D10×D50: interaction term p-value = 0.003 (< 0.01), model R 2 =79.8%, AIC=116.2 (R 2 (Increased by 7.7%); continue to introduce D10×D50. 2 Interaction term p-value = 0.15 (>0.05), Model R 2 =80.1% (an increase of 0.3%), AIC=118.7 (an increase of 2.5%). Based on the above, only the significant interaction terms D10×D50 are retained, and the model R... 2 =79.8%. Step 4: Regularization parameter optimization (to control overfitting) Three regularization methods (L1, L2, and ElasticNet) were tested, and the optimal regularization coefficient λ was determined through grid search (search range: 10). -4 ~10 1 (Log-scale division into 20 nodes) L1 regularization: When λ=0.08, the validation set error is 15.2 mOsmol / kg, but the D10 coefficient is compressed to 0, resulting in the loss of feature information; Elastic network (α=0.5): When λ=0.06, the validation set error is 14.8 mOsmol / kg, but the model complexity increases (the number of parameters increases by 2); L2 regularization: When λ=0.1, the validation set error is 14.5 mOsmol / kg (minimum), and all core feature coefficients are significant (p<0.01), indicating optimal control of overfitting (training set R). 2 =82.5%, test set R 2 =80.3%, a difference of 2.2%); Based on the above, we choose L2 regularization with λ=0.1, and the model R... 2 =80.3%. Step 5: Cross-validation and stability verification (to ensure generalization ability) A dual validation strategy is used to test the model's stability: 10-fold cross-validation: Divide the dataset into training and validation sets in an 8:2 ratio, repeat 10 times, and calculate the average R-value.2 =85.1%±2.3% (standard deviation <3%, good stability); Leave-one-out validation (LOOCV): With a total of 120 samples, one sample is left as the test set each time. The final R-value of LOOCV is... 2 =84.7%, which is close to the result of 10-fold cross-validation (difference <0.5%). Residual analysis: The validation set residuals are normally distributed (Shapiro-Wilk test p=0.23>0.05), with no systematic bias (mean residual = 0.32mOsmol / kg, close to 0). The final model structure, summed up above, is: L2 regularization (λ=0.1) + linear terms (D10, D50) + quadratic terms (D50). 2 + Interactive items (D10×D50), test set R 2 =85.1%, meeting all performance requirements.

[0068] 3. Parameter Optimization: Grid Search and Fine-Grained Iteration The following is a detailed process for fine-grained optimization of three types of key hyperparameters: A. Polynomial order optimization (determining model complexity) Test 1st to 3rd order polynomial models with "validation set R" 2 The "+AIC value" is a dual evaluation standard: First-order polynomial (linear terms only): Validation set R 2 =64.2%, AIC=135.6 (underfit, unable to capture nonlinear relationships); 2nd order polynomial (including linear terms + quadratic terms + interaction terms): validation set R 2 =80.3%, AIC=116.2 (lowest AIC, balancing fit and complexity); 3rd order polynomial (including higher-order terms D50) 3 D10×D50 2 ): Validation set R 2 =81.5%, but the training set R 2 =92.3% (severe overfitting), AIC=120.8 (higher than order 2).

[0069] Based on the above, the order of the polynomial is determined to be 2. B. Optimization of regularization coefficient λ (balancing fitting and generalization) The optimal λ is determined using a two-step search method of "coarse mesh + fine mesh". Step 1: Coarse grid search (range 10) -4 ~10 1 (Logarithmic scale with 10 nodes): Calculate the RMSE of the validation set corresponding to each λ, and find that λ is within 10... -2 ~100 When within the range, RMSE shows a decreasing trend followed by stabilization, with the initial range locked at 0.01~1; Step 2: Fine-grid search (range 0.01~1, interval 0.01, 100 nodes in total): Plot the "λ-RMSE" curve. It was found that when λ=0.1, the validation set RMSE=14.5mOsmol / kg (minimum), and when λ>0.1, the RMSE increases (model underfitting), and when λ<0.1, the RMSE does not decrease significantly (overfitting risk increases). Step 3: Stability verification: Under different data partitions (training set / validation set = 7:3, 9:1), the RMSE fluctuation of the validation set when λ=0.1 is <0.5mOsmol / kg, and the parameter stability meets the standard; Based on the above, the regularization coefficient λ is determined to be 0.1. C. Validation of hyperparameter combinations (ensuring global optimality) Construct a combination matrix of "polynomial order (1~2) × regularization coefficient (0.05~0.2)", resulting in 2×16=32 combinations of hyperparameters. Calculate the validation set R for each combination. 2 With AIC: Optimal combination: 2nd order polynomial + λ = 0.1, validation set R 2 =80.3%, AIC=116.2 (higher than other combinations, such as 2nd order +λ=0.08 AIC=118.5, 1st order +λ=0.1 AIC=132.7). In summary, the optimal combination of hyperparameters is a second-order polynomial plus L2 regularization (λ=0.1). This combination exhibits consistent performance on the training, validation, and test sets, with no risk of overfitting or underfitting.

[0070] D. Final model equations (R) 2 =85.1%, p-value < 0.01) Osmotic pressure (mOsmol / kg) = 450.2 - 1025.5 × D50 + 1213.8 × D50 2 +500.6×D10×D50 Wherein, D10 represents the diameter of 10% of the particles in the particle size distribution, in μm; D50 represents the diameter of 50% of the particles in the particle size distribution, in μm; particle size (D50) is the main influencing factor: when D50 increases, the osmotic pressure decreases twice; the interaction term D10×D50 indicates that when small particle sizes D10 and D50 are combined, they can more sensitively affect the osmotic pressure.

[0071] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0072] In the specific implementation process, in order to improve the accuracy of the preparation control method for formulated nutritional foods and the safety of the products, it is necessary to ensure the accuracy of the osmotic pressure fitting model in fitting the osmotic pressure of the formulated nutritional foods. To verify that the osmotic pressure fitting model has a good fitting effect, the osmotic pressure fitting model provided in the above embodiments is validated.

[0073] Example 1: Model Validation Experiment Formula 1: Raw milk 64.6%, demineralized whey powder 19.1%, blended edible vegetable oil 8.0%, lactose 4.8%, whey protein powder 0.26%, galactooligosaccharides 1.6%, arachidonic acid oil powder 0.5%, compound minerals 0.4%, docosahexaenoic acid oil powder 0.4%, compound vitamins 0.2%, calcium carbonate 0.1%, choline chloride 0.04%.

[0074] Preparation process 1: The formula is mixed, homogenized at 170 bar, sterilized, concentrated, and spray-dried to prepare a semi-finished product. The semi-finished product is then physically mixed with docosahexaenoic acid and arachidonic acid to produce the finished product.

[0075] Formula 2: 64.5% raw milk, 19.4% demineralized whey powder, 8.6% blended edible vegetable oil, 4.9% lactose, 0.38% whey protein powder, 0.6% galactooligosaccharides, 0.5% arachidonic acid oil powder, 0.33% compound minerals, 0.4% docosahexaenoic acid oil powder, 0.2% compound vitamins, 0.1% calcium carbonate, 0.06% choline chloride, and 0.03% lactoferrin.

[0076] Preparation process 2: The formula is mixed, homogenized at 170 bar, sterilized, concentrated, and spray-dried to prepare a semi-finished product. The semi-finished product is then physically mixed with docosahexaenoic acid, arachidonic acid, and lactoferrin to produce the finished product.

[0077] Formula 3: 62.9% raw milk, 12.9% demineralized whey powder, 8.09% blended edible vegetable oil, 11.2% lactose, 0.55% whey protein powder, 1.6% galactooligosaccharides, 0.9% anhydrous butter, 0.5% arachidonic acid oil powder, 0.62% compound minerals, 0.4% docosahexaenoic acid oil powder, 0.16% compound vitamins, 0.1% calcium carbonate, 0.06% choline chloride, and 0.02% nucleotides.

[0078] Preparation process 3: The formula is mixed, homogenized at 190 bar, sterilized, concentrated, and spray-dried to prepare a semi-finished product. The semi-finished product is then physically mixed with docosahexaenoic acid, arachidonic acid, and nucleotides to produce the finished product.

[0079] The consistency of predictions was verified using the Bland-Altman analysis method:

[0080] For formulations 1, 2, and 3, the predicted values ​​were all within the range of actual values, with a deviation of 1.0 mOsmol / kg and a 95% confidence interval of -1.08 to 3.08 mOsmol / kg. This indicates that the model can accurately predict osmotic pressure based on particle size, confirming that the model has good predictive reliability.

[0081] Example 2: Risk Intervention Measures

[0082] To reduce the osmotic pressure of formulas 2 and 3, intervention measures are initiated. These measures include adjusting the formula ingredients (such as increasing prebiotics and reducing lactose in formula 2) and the production process (such as reducing homogenization pressure in formula 3). This can reduce the osmotic pressure of the milk powder, help optimize the osmotic pressure index of the milk powder, and reduce the risk of osmotic stress.

[0083] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0084] Figure 4 A schematic diagram of the structure of the formula nutritional food preparation control device provided in an embodiment of the present invention is shown. For ease of explanation, only the parts related to the embodiment of the present invention are shown, and are described in detail below: like Figure 4 As shown, the formula nutritional food preparation control device 4 includes: The acquisition module 401 is used to acquire real-time particle size data during the preparation process of formulated nutritional foods; The osmotic pressure fitting module 402 is used to input real-time particle size data into the trained osmotic pressure fitting model to obtain the fitted osmotic pressure; wherein, the osmotic pressure fitting model is trained based on the sample characteristic particle size data and sample osmotic pressure corresponding to the nutritional food formula to be prepared. The determination module 403 is used to determine the risk intervention and control scheme based on the fitted osmotic pressure and the preset early warning conditions; wherein, the process parameters in the risk intervention and control scheme include one or more of the following: prebiotic adjustment amount, lactose adjustment amount, and homogenization pressure; Control module 404 is used to adjust process parameters according to the risk intervention control scheme.

[0085] In one possible implementation, module 403 is specifically used for: When the fitted osmotic pressure meets the medium-risk warning condition, the first difference between the fitted osmotic pressure and the upper limit of the low-risk warning condition is calculated, and the amount of prebiotics added and the amount of lactose reduced are determined based on the first difference; wherein, the amount of prebiotics added and the amount of lactose reduced are the same. When the fitted osmotic pressure meets the high-risk warning condition, calculate the second difference between the fitted osmotic pressure and the upper limit of the low-risk warning condition, and determine the homogeneous pressure reduction value based on the second difference.

[0086] In one possible implementation, the formula nutritional food preparation control device 4 further includes a model training module for acquiring multiple formula samples of the formula nutritional food to be prepared, and acquiring the particle size distribution parameters and measured osmotic pressure values ​​of each formula sample; using a multivariate nonlinear regression analysis method, with the particle size distribution parameters as independent variables and the osmotic pressure value as the dependent variable, an initial fitting model is constructed; wherein the particle size distribution parameters include D10, D50, D90, and D10 2 D50 2 The initial fitted model is subjected to feature selection, five-step iterative optimization, cross-validation, and parameter optimization to obtain the osmotic pressure fitting model.

[0087] In this embodiment, by acquiring real-time particle size data during the preparation of formulated nutritional foods, and inputting it into an osmotic pressure fitting model trained based on the characteristic particle size and osmotic pressure of the corresponding formulated nutritional food sample, a fitted osmotic pressure is obtained. Then, combined with preset early warning conditions, a risk intervention and control scheme including process parameters such as prebiotic regulation amount, lactose regulation amount, or homogenization pressure is determined and the process parameters are adjusted. This enables real-time prediction of osmotic pressure and dynamic adjustment of the process during the preparation of formulated nutritional foods, avoiding the risk of intestinal stress in formulated nutritional foods caused by improper osmotic pressure, improving the accuracy of formulated nutritional food preparation and product safety, and solving the defects of traditional control methods that rely on delayed feedback.

[0088] Figure 5 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Figure 5 As shown, the electronic device 5 of this embodiment includes a processor 50 and a memory 51. The memory 51 stores a computer program 52. When the processor 50 executes the computer program 52, it implements the steps in the various method embodiments described above. Alternatively, when the processor 50 executes the computer program 52, it implements the functions of each module / unit in the various device embodiments described above.

[0089] For example, computer program 52 may be divided into one or more modules / units, which are stored in memory 51 and executed by processor 50 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 52 in electronic device 5.

[0090] Electronic device 5 may include, but is not limited to, processor 50 and memory 51. Those skilled in the art will understand that... Figure 5 This is merely an example of electronic device 5 and does not constitute a limitation on electronic device 5. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device 5 may also include input / output devices, network access devices, buses, etc.

[0091] The processor 50 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0092] The memory 51 can be an internal storage unit of the electronic device 5, such as a hard disk or RAM. The memory 51 can also be an external storage device of the electronic device 5, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory 51 can include both internal and external storage units of the electronic device 5. The memory 51 is used to store the computer program 52 and other programs and data required by the electronic device 5. The memory 51 can also be used to temporarily store data that has been output or will be output.

[0093] For the sake of simplicity and clarity, only the above-described functional modules / units are used as examples. In practical applications, the functions described above can be assigned to different functional modules / units as needed. These modules / units can be implemented in hardware, software, or a combination of both.

[0094] This invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the methods described in the above-described method embodiments.

[0095] This invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the methods described in the above-described method embodiments.

[0096] Computer programs include computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. Computer-readable media can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0097] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0098] The above-described 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, and should all be included within the protection scope of the present invention.

Claims

1. A method for controlling the preparation of a formulated nutritional food, characterized in that, include: Obtain real-time particle size data during the preparation of formulated nutritional foods; The real-time particle size data is input into the trained osmotic pressure fitting model to obtain the fitted osmotic pressure; wherein, the osmotic pressure fitting model is trained based on the sample characteristic particle size data and sample osmotic pressure corresponding to the nutritional food formula to be prepared; Based on the fitted osmotic pressure and preset early warning conditions, a risk intervention and control scheme is determined; wherein, the process parameters in the risk intervention and control scheme include one or more of the following: prebiotic adjustment amount, lactose adjustment amount, and homogenization pressure; Adjust process parameters according to the aforementioned risk intervention and control scheme.

2. The method for controlling the preparation of formulated nutritional foods according to claim 1, characterized in that, The preset early warning conditions include preset low-risk, medium-risk, and high-risk early warning conditions; The risk intervention and control plan includes: a medium-risk risk intervention and control plan and a high-risk risk intervention and control plan; The process parameters corresponding to the medium-risk intervention and control scheme include: prebiotic regulation amount and lactose regulation amount; The process parameters corresponding to the high-risk intervention and control scheme include homogenization pressure.

3. The method for controlling the preparation of formulated nutritional foods according to claim 2, characterized in that, The step of determining a risk intervention and control plan based on the fitted osmotic pressure and preset early warning conditions includes: When the fitted osmotic pressure meets the medium-risk warning condition, the first difference between the fitted osmotic pressure and the upper limit of the low-risk warning condition is calculated, and the amount of prebiotics added and the amount of lactose reduced are determined based on the first difference; wherein, the amount of prebiotics added and the amount of lactose reduced are the same; When the fitted osmotic pressure meets the high-risk warning condition, the second difference between the fitted osmotic pressure and the upper limit of the low-risk warning condition is calculated, and the homogeneous pressure reduction value is determined based on the second difference.

4. The method for controlling the preparation of formulated nutritional foods according to claim 1, characterized in that, Before obtaining real-time particle size data during the preparation of the formulated nutritional food, the method further includes: Multiple formulation samples of the nutritional food to be prepared were obtained, and the particle size distribution parameters and measured osmotic pressure values ​​of each formulation sample were obtained. A multivariate nonlinear regression analysis method was used, with the particle size distribution parameters as independent variables and the osmotic pressure value as the dependent variable, to construct an initial fitting model; wherein the particle size distribution parameters include D10, D50, D90, and D10. 2 D50 2 and D10×D50; The initial fitting model is subjected to feature selection, five-step iterative optimization, cross-validation, and parameter optimization to obtain the osmotic pressure fitting model.

5. The method for controlling the preparation of formulated nutritional foods according to claim 4, characterized in that, The feature selection includes: Determine candidate particle size characteristics associated with osmotic pressure; wherein, the candidate particle size characteristics include: D10, D50, D90, and D10. 2 D50 2 and D10×D50; Based on the significance test, weakly correlated features are removed from the candidate particle size features to obtain the first candidate particle size features; Based on the variance inflation factor, the features with moderate collinearity in the first candidate particle size features are centered so that the variance inflation factor of the processed features is ≤5. The importance of each feature after centering is calculated using the random forest algorithm, and features with importance below a set threshold are removed. Compare the performance of models with and without D10×D50, and determine the final feature set based on model fit and information criterion values.

6. The method for controlling the preparation of formulated nutritional foods according to claim 5, characterized in that, The five-step iterative optimization includes: Test three types of linear models: ordinary least squares, ridge regression, and Lasso regression. Compare the model fit, residual standard deviation, and overfitting to determine the basic model. For the final feature set, the model performance after introducing quadratic terms is tested respectively. Quadratic terms that improve the model fit of the basic model and reduce the information criterion value are retained to form a model that combines linear and quadratic terms. Candidate interaction terms are introduced using forward stepwise regression. The significance of the interaction terms is tested, and interaction terms that improve the model fit and reduce the information criterion value are retained. We tested three regularization methods: L1 regularization, L2 regularization, and elastic network regularization. We determined the optimal regularization coefficients through grid search and selected the regularization method that minimized the validation set error and made all core feature coefficients significant.

7. The method for controlling the preparation of formulated nutritional foods according to claim 6, characterized in that, The cross-validation includes: The stability of the model was verified by 10-fold cross-validation and leave-one-out method. Residual analysis was also performed to ensure that the residuals were normally distributed and free from systematic bias, so as to obtain the target particle size characteristics corresponding to the osmotic pressure fitting model.

8. A control device for preparing formulated nutritional foods, characterized in that, include: The acquisition module is used to acquire real-time particle size data during the preparation of formulated nutritional foods; The osmotic pressure fitting module is used to input the real-time particle size data into the trained osmotic pressure fitting model to obtain the fitted osmotic pressure; wherein, the osmotic pressure fitting model is trained based on the sample characteristic particle size data and sample osmotic pressure corresponding to the nutritional food formula to be prepared; The determination module is used to determine a risk intervention and control scheme based on the fitted osmotic pressure and preset early warning conditions; wherein, the process parameters in the risk intervention and control scheme include one or more of the following: prebiotic adjustment amount, lactose adjustment amount, and homogenization pressure; The control module is used to adjust process parameters according to the risk intervention control scheme.

9. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images