A method and system for intelligent ingredient formulation of infant formula based on target nutrient structure inversion

Abstract

This invention relates to a smart formula mixing method for infant formula, and more particularly to a smart formula mixing method and system for infant formula based on target nutrient structure inversion. The method establishes a target nutrient structure set and a candidate feeding node set, collects real-time composition parameters and exposure sensitivity parameters of raw materials, calculates the cumulative heat exposure, oxygen exposure, and shear exposure of each raw material at different feeding nodes, and determines the effective retention rate accordingly. A subset of allowable feeding nodes is then selected, and a segmented inversion solution is performed to determine the target addition amount and target feeding nodes. Dynamic corrections are made based on online detection parameters before any node is executed. This approach can improve the accuracy of target nutrient structure realization, active ingredient retention rate, reconstitution properties, batch consistency, and full-process traceability.

Description

Technical Field

[0001] This invention relates to a method for intelligent ingredient formulation of milk powder, and more particularly to an intelligent ingredient formulation method and system for infant formula milk powder based on the inversion of target nutrient structure. Background Technology

[0002] Infant formula is a special type of food developed to mimic the nutritional composition of breast milk and meet the growth and development needs of infants at different ages. Compared with regular milk powder, infant formula not only requires that basic nutrients such as protein, fat, lactose, vitamins, and minerals meet regulatory and formulation design requirements, but also increasingly emphasizes higher-level structural nutritional indicators such as the whey protein to casein ratio, long-chain polyunsaturated fatty acid composition, prebiotic system, fat structure characteristics, and retention rate of active ingredients. Furthermore, the industrial production of infant formula typically involves multiple continuous processes, including raw material standardization, wet mixing, homogenization, heat treatment, concentration, spray drying, fluidized bed post-processing, dry mixing, and packaging. There are clear material coupling and process transfer relationships between these processes. Against this backdrop, establishing a stable, precise, and traceable ingredient control relationship between raw material fluctuations, process losses, and the final product target has become a crucial technical challenge in this field.

[0003] In existing technologies, automated control of the infant formula production process has emerged with a basic logic of target product standards—raw material information—online detection—automatic feeding. For example, patent document CN105145838A discloses a children's formula milk powder and its production method. Its technical approach involves: after fresh milk and other raw materials are added to a container, a computer accurately calculates the batches, different compositions, qualities, and weights of raw materials, and an automatic online detection system automatically detects the mixed raw materials again. The raw material data and detection data are then compared with the target product standard data. Subsequently, a computer-controlled automatic feeding system adds various auxiliary raw materials and additives to ensure the raw materials meet the standard requirements. Simultaneously, this document also computerizes processes such as raw material pretreatment, raw material standardization, raw material mixing, raw material homogenization, and sterilization. Therefore, existing technologies have recognized that in infant formula production, automated calculations and online detection can achieve closed-loop regulation of the mixing process, thereby improving the mixing accuracy and production consistency under different batch raw material conditions.

[0004] However, while the aforementioned existing technologies have solved the problem of automated ingredient mixing based on target standards, their core control still mainly revolves around the comparison between the total composition of raw materials and the target standards, placing more emphasis on the control of traditional nutritional indicators. For the structural nutritional targets that are increasingly important in infant formula, such as the whey protein to casein ratio, the structural distribution of specific fatty acids in triglycerides, the proportion of prebiotic combinations, and the effective retention level of active nutrients in the final product, existing technologies typically lack more detailed dynamic solution mechanisms. Especially in actual production, different raw materials not only have different compositional parameters, but also exhibit significant differences in their sensitivity to temperature, oxygen exposure, and shear conditions; the effective contribution of the same raw material added at different process stages will also differ in the final product. Simply relying on calculations between the total composition of raw materials and the target components often fails to simultaneously meet the comprehensive requirements of achieving nutritional standards, preserving active components, and adapting to the process.

[0005] On the other hand, existing technologies have also provided some improvement ideas for handling certain unstable, reactive, or heat-sensitive components in infant formula. For example, patent document CN103504025A discloses a wet preparation process for infant formula. This document proposes that for some reactive and unstable substances such as vitamins, minerals, vegetable oils, and oils in other nutrients, microencapsulation technology can be used to encapsulate them first, followed by mixing and packaging under certain production conditions. Simultaneously, raw materials such as fresh milk, whey powder, lactose, whey protein, and oligosaccharides are first mixed, then homogenized under high pressure, concentrated, and spray-dried to obtain a base powder. The aforementioned encapsulated substances are then mixed evenly with the base powder, and finally, nitrogen-filled quantitative packaging is performed. The advantage of this document is that by appropriately separating the reactive and unstable components from the base powder preparation process and supplementing it with microencapsulation, adverse reactions between components can be reduced or prevented, improving the stability, uniformity, and nutritional activity retention of the finished product.

[0006] However, technical solutions like CN103504025A are essentially still fixed process routes pre-set. That is, which raw materials need to be encapsulated, which are added in the wet process stage, and which are added in the subsequent mixing stage are usually determined in advance by the formulation process personnel based on experience or a pre-set process route. Once in the production process, facing fluctuations in the composition of different batches of raw materials, changes in thermal history under different production loads, and changes in oxygen exposure and shear strength under different equipment conditions, existing technologies generally lack an intelligent mechanism that can assess the exposure risk of each raw material at different candidate feeding points in real time, calculate its effective retention level, and further dynamically allocate feeding points and compensate for the amount added accordingly. In other words, although existing technologies have recognized that certain sensitive components are not suitable for direct entry into the entire high-temperature wet process stage, they have not yet truly established a systematic, quantifiable, and online-correctable ingredient control chain connecting raw material sensitivity—process exposure—effective retention rate—feeding path—target final nutrient structure.

[0007] In the current industrial production of infant formula, the sources of raw materials are becoming increasingly diverse. In addition to traditional basic raw materials such as skim milk, whey powder, lactose, and vegetable oil, various functional or active raw materials are also introduced, including structured lipids, DHA / ARA microcapsules, lactoferrin, osteopontin, prebiotics, probiotics, multivitamins, and multiminerals. These raw materials differ significantly in terms of physicochemical properties, thermal stability, oxidation sensitivity, shear tolerance, and solubility and dispersion performance. If a fixed formula and fixed feeding points are still used for ingredient preparation, on the one hand, some raw materials are prone to degradation under subsequent high-temperature, oxygen-rich, or high-shear conditions, leading to a deviation between the actual nutritional structure of the final product and the design value; on the other hand, simply increasing the amount of raw materials to compensate for process losses may cause instability in the upstream process, poor local solubility and dispersion, instability of the emulsion system, increased spray drying load, and even a decrease in the reconstitution properties, particle structure, and shelf stability of the finished product. Therefore, how to comprehensively consider the real-time composition of raw materials, the sensitive characteristics of raw material exposure, the environmental conditions of each process node, and the target nutritional structure of the final product in the actual production process, and to carry out intelligent inversion solution and dynamic ingredient control by grouping, segmenting, and node, is still a problem that existing technologies have not fully solved.

[0008] In summary, while existing technologies have disclosed automated ingredient control methods and segmented addition or encapsulation methods for unstable components in infant formula production, neither has addressed the following key technical issues: How to quantitatively assess the effective retention rate of raw materials based on the target nutritional structure and the heat, oxygen, and shear exposure characteristics of different raw materials at different candidate feeding points during infant formula production; and how to perform segmented inversion solutions and dynamic feeding control for basic milk phase raw materials, fat structure regulating raw materials, and heat-sensitive active raw materials, respectively, under multiple constraints such as achieving target nutritional standards, structural nutritional shaping, activity retention, and process adaptation. To address these issues, it is still necessary to propose a new intelligent ingredient control method and system for infant formula based on target nutritional structure inversion to further improve the accuracy, adaptability, and consistency of final product quality in infant formula ingredient control. Summary of the Invention

[0009] The technical objective of this invention is to address the problem that existing infant formula production processes mainly rely on fixed formula lists, fixed feeding paths, and experience-based compensation for ingredient control, making it difficult to simultaneously consider batch fluctuations of raw materials, differences in the retention of sensitive active components at different process stages, and the achievement of the target nutritional structure of the final product. This invention provides an intelligent ingredient formulation method and system for infant formula based on target nutritional structure inversion, enabling coordinated control of basic nutritional indicators, structural nutritional indicators, and process constraint indicators. This improves the accuracy of ingredient calculations, the matching of feeding paths, the retention rate of active ingredients, and the consistency and traceability of the final product quality.

[0010] Firstly, in order to achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0011] A smart ingredient formulation method for infant formula milk powder based on target nutrient structure inversion includes the following steps:

[0012] S1. Establish a target nutrient structure set and establish a candidate feeding node set corresponding to the target nutrient structure set;

[0013] S2. Obtain the real-time composition parameters and exposure sensitivity parameters of each candidate raw material, and divide each candidate raw material into the basic emulsion raw material group, the fat structure regulation raw material group, and the thermosensitive active raw material group.

[0014] S3. For any candidate raw material, calculate the cumulative heat exposure, cumulative oxygen exposure and cumulative shear exposure of the raw material at each candidate feeding node based on the temperature trajectory, oxygen concentration trajectory and shear trajectory corresponding to each candidate feeding node on the production line.

[0015] S4. Based on the cumulative heat exposure, cumulative oxygen exposure, and cumulative shear exposure, calculate the effective retention rate of each raw material at each candidate feeding node;

[0016] S5. Based on the effective retention rate, the raw material group to which each raw material belongs, the process adaptation constraints of each candidate feeding node, and the target nutrient structure set, screen the allowable feeding node subset for each raw material.

[0017] S6. Within the subset of allowed feeding nodes, establish a piecewise inversion solution model to solve for the target addition amount and target feeding nodes for the basic emulsion raw material group, the fat structure regulating raw material group, and the thermosensitive active raw material group, respectively.

[0018] S7. Based on the solution results of the piecewise inversion solution model, control the liquid metering device, the powder metering device, and the feeding execution device corresponding to the target feeding node to feed materials.

[0019] As a further improvement, in step S1, the target nutrient structure set includes at least: a target basic nutrient index set, a target structural nutrient index set, and a target process constraint index set.

[0020] The target set of basic nutritional indicators includes at least the target protein content, target fat content, target lactose content, and target ash content;

[0021] The set of target structural nutritional indicators includes at least one or more of the following: target whey protein to casein mass ratio, target palmitic acid percentage at sn-2 position, target prebiotic ratio, and target long-chain polyunsaturated fatty acid ratio.

[0022] The set of target process constraint indicators includes at least one or more of the following: target osmotic pressure range, target upper limit of oxidation risk, target remedial index, and target lower limit of activity retention rate.

[0023] The candidate feeding node set includes at least a wet main feeding node, a node before homogenization, a node before concentration, a dry mixing node after spray drying, and a node before and after packaging.

[0024] As a further improvement, in step S1, the set of target basic nutritional indicators further includes one or more of the following: target energy density, target calcium content, target phosphorus content, target sodium content, and target potassium content; the set of target process constraint indicators further includes one or more of the following: target upper limit of moisture content, target upper limit of water activity, target particle size distribution index, and target upper limit of insoluble index.

[0025] As a further improvement, in step S2, the real-time composition parameters of each raw material include at least the protein content, fat content, lactose content, ash content, and structural characteristic parameters corresponding to the nutritional indicators of the target structure; the exposure sensitivity parameters of each raw material include at least the thermosensitive critical temperature, the oxygen-sensitive critical oxygen concentration, and the shear sensitivity threshold.

[0026] As a further improvement, in step S2, the basic milk phase raw material group includes at least one or more of skim milk, whole milk, whey, lactose solution, and mineral salt solution.

[0027] As a further improvement, in step S2, the fat structure adjustment raw material group includes at least one or more of light cream, vegetable oil, structured lipids and long-chain polyunsaturated fatty acid oils.

[0028] As a further improvement, in step S2, the thermosensitive active raw material group includes at least one or more of lactoferrin, osteopontin, probiotics, vitamin-like active substances, and microcapsule functional oils.

[0029] As a further improvement, in step S2, the real-time composition parameters of each raw material are obtained by one or more of an online near-infrared spectrometer, an online fat analyzer, an online protein analyzer, an online densitometer, and an online viscometer; wherein, the online near-infrared spectrometer is used to output estimated values ​​of protein content, fat content, lactose content, and ash content.

[0030] As a further improvement, in step S3, the cumulative amount of heat exposure... Calculate as follows:

[0031] ;

[0032] in, and They represent the first The first type of raw material in The times when each candidate feeding node enters and leaves the subsequent process stage; Indicates the first After each candidate feeding node, the material temperature along the process path changes over time. The change value; Indicates the first The thermosensitive critical temperature of a certain raw material.

[0033] As a further improvement, in step S3, the cumulative oxygen exposure... Calculate as follows:

[0034] ;

[0035] in, Indicates the first After each candidate feeding node, the oxygen concentration in the process path changes over time. The change value; Indicates the first The oxygen-sensitive critical oxygen concentration of a certain raw material.

[0036] As a further improvement, in step S3, the cumulative amount of shear exposure... Calculate as follows:

[0037] ;

[0038] in, Indicates the first The equivalent shear strength along the process path after each candidate feeding node varies with time. The change value; Indicates the first The shear sensitivity threshold of a certain raw material.

[0039] As a further improvement, in step S4, the effective retention rate is determined in the following way: the cumulative amount of heat exposure, cumulative amount of oxygen exposure, and cumulative amount of shear exposure of each raw material at each candidate feeding node are input into the retention rate function pre-established for the corresponding raw material to obtain the effective retention ratio of the raw material that can be retained in the final product after subsequent processes at the candidate feeding node; wherein, the retention rate function is a functional relationship, lookup table, or piecewise function obtained by fitting the active retention experimental data of the raw material under preset temperature, oxygen concentration, and shear conditions.

[0040] As a further improvement, in step S5, the raw materials in the thermosensitive active raw material group are screened only at candidate feeding nodes where their effective retention rate is not lower than the preset active retention threshold; the raw materials in the fatty acid structure regulating raw material group are screened only at candidate feeding nodes where the target palmitic acid ratio at the sn-2 position and the emulsion dispersion stability requirements are met.

[0041] As a further improvement, in step S5, the process adaptation constraints include at least one or more of the following constraints: solubility constraints, dispersion stability constraints, pre-homogenization emulsion formation constraints, spray drying heat exposure constraints, and dry mixing uniformity constraints; wherein, when the raw materials of the fat structure adjustment raw material group meet the pre-homogenization emulsion formation constraints, they are preferentially allocated to the wet process main batching node or the pre-homogenization node.

[0042] As a further improvement, in step S6, the piecewise inversion solution model must simultaneously satisfy at least the following conditions:

[0043] The set of target basic nutritional indicators for the final product meets the deviation limit requirements;

[0044] The target set of structural nutritional indicators for the final product meets the deviation limit requirements.

[0045] The set of target process constraints for the final product shall not be exceeded;

[0046] Each raw material can only be assigned to one or more candidate feeding nodes in its subset of allowed feeding nodes;

[0047] The solution results for the thermosensitive active raw material group, after being compensated and corrected for the corresponding effective retention rate, meet the target value requirements of the final product.

[0048] As a further improvement, in step S6, the piecewise inversion solution model uses the weighted result of the final product's basic nutritional index deviation term, final product's structural nutritional index deviation term, process constraint penalty term, insufficient activity retention penalty term, and raw material cost term as the optimization objective, and the minimum of this weighted result is the solution objective; wherein, the final product's basic nutritional index deviation term is used to represent the deviation between the predicted value and the target value of the final product's basic nutritional index; the final product's structural nutritional index deviation term is used to represent the deviation between the predicted value and the target value of the final product's structural nutritional index; the process constraint penalty term is used to represent the penalty value generated when the process constraint index exceeds the target range; the insufficient activity retention penalty term is used to represent the penalty value generated when the thermosensitive active raw material still fails to meet the final product's target value after compensation and correction; and the raw material cost term is used to represent the cost of adding each raw material.

[0049] As a further improvement, in step S6, the basic nutritional index deviation term is calculated as follows: the absolute deviation between the predicted value and the target value of each basic nutritional index involved in the calculation is multiplied by the corresponding weight coefficient, and the results are summed to obtain the basic nutritional index deviation term; wherein, the number of basic nutritional indicators involved in the calculation represents the total number of basic nutritional indicators included in the deviation calculation; the predicted value represents the prediction result of the corresponding basic nutritional index after piecewise inversion solution; the target value represents the set value of the basic nutritional index in the target nutritional structure set; and the weight coefficient represents the importance of the basic nutritional index in the deviation evaluation.

[0050] As a further improvement, in step S6, for the target raw material in the thermosensitive active raw material group, when it is allocated to the target candidate feeding node, its target feeding amount is compensated and corrected according to the target demand amount of the final product and the effective retention rate. The compensation and correction satisfy that: the result of multiplying the target feeding amount by the effective retention rate is not lower than the target demand amount of the corresponding effective component of the raw material in the final product; wherein, the target feeding amount represents the amount of the target raw material added at the target candidate feeding node; and the target demand amount of the final product represents the target amount that the corresponding effective component of the target raw material needs to achieve in the final product.

[0051] As a further improvement, in step S7, the current process parameters and online detection parameters are obtained before each candidate feeding node; and the method also includes steps S8 and S9; wherein...

[0052] Step S8: Before any candidate feeding node that has not yet been executed, based on the latest online detection parameters and current process parameters, recalculate the effective retention rate of the corresponding unfeeded raw materials under the remaining candidate feeding nodes, and determine whether the predicted nutritional structure of the final product still meets the requirements of the target nutritional structure set.

[0053] If the target is not met, steps S5 and S6 are re-executed only for raw materials for which the feeding node has not yet been executed, in order to correct their target addition amount and / or target feeding node.

[0054] Step S9: Output the final batching control command and bind and store the raw material batch, candidate feeding node selection results, effective retention rate, segmented inversion solution results, online detection parameters, and correction records.

[0055] As a further improvement, in step S7, the online detection parameters include at least online protein detection value, online fat detection value, online lactose detection value, online temperature detection value, online oxygen concentration detection value, and online viscosity detection value;

[0056] In step S8, steps S5 and S6 are re-executed for raw materials that have not yet been fed at a feeding node only if the number of remaining candidate feeding nodes is not less than the preset minimum number of correction nodes.

[0057] In step S8, the corrected target feeding node is only allowed to switch within the subset of allowed feeding nodes for the corresponding raw material, and cannot switch to a process node before a candidate feeding node that has already been executed.

[0058] Secondly, the present invention also provides an intelligent ingredient formulation system for infant formula milk powder based on target nutrient structure inversion. This system is used to implement the method described above, including:

[0059] The target trophic structure modeling module is used to establish a set of target trophic structures and a set of candidate feeding nodes;

[0060] The raw material parameter acquisition module is used to acquire the real-time composition parameters and exposure sensitivity parameters of each candidate raw material, and to classify them into basic emulsion raw material group, lipid structure regulation raw material group and thermosensitive active raw material group.

[0061] The exposure calculation module is used to calculate the cumulative heat exposure, cumulative oxygen exposure, and cumulative shear exposure of each raw material at each candidate feeding node;

[0062] The retention rate calculation module is used to calculate the effective retention rate based on the cumulative heat exposure, cumulative oxygen exposure, and cumulative shear exposure.

[0063] The node filtering module is used to filter the subset of allowed feeding nodes for each raw material;

[0064] The piecewise inversion solution module is used to solve for the target addition amount and target feeding node of each raw material;

[0065] The execution control module is used to control the liquid metering device, powder metering device, and various feeding execution devices to complete the feeding process.

[0066] Thirdly, the present invention also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implement the method.

[0067] Fourthly, the present invention also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the method.

[0068] This invention constructs a target nutrient structure set and quantitatively analyzes the basic emulsion raw materials, fat structure regulating raw materials, and thermosensitive active raw materials under different candidate feeding nodes based on their thermal exposure, oxygen exposure, and shear exposure conditions. It further calculates the effective retention rate of each raw material in different process paths and performs allowable feeding node screening, segmented inversion solution, and dynamic correction before the nodes based on this. This allows the ingredient control to move beyond traditional static formula calculation and fixed feeding sequence, and instead enables collaborative optimization of the basic nutrient indicators, structural nutrient indicators, and process constraint indicators of the final product. Compared to existing technologies, this invention can more accurately compensate for the loss of heat-sensitive, oxygen-sensitive, and shear-sensitive raw materials in subsequent processes, avoiding nutritional deviations and activity attenuation caused by batch fluctuations in raw materials or changes in process conditions. It improves the accuracy of achieving target structural nutritional indicators such as the ratio of whey protein to casein, the proportion of palmitic acid at the sn-2 position, the proportion of prebiotic combinations, and the ratio of long-chain polyunsaturated fatty acids. At the same time, by grouping and matching raw materials to more suitable feeding nodes, it can also improve the dissolution and dispersion effect, emulsion stability, fat structure formation effect, and the retention level of heat-sensitive active components, reducing unnecessary excessive addition and subsequent quality fluctuations. In this way, it improves the comprehensive performance of infant formula in terms of nutritional compliance, active ingredient retention rate, reconstitution properties, process adaptability, batch consistency, and full-process traceability, and has significant engineering application value and industrial promotion significance. Attached Figure Description

[0069] Figure 1 This is a schematic diagram of the overall process of an intelligent ingredient formulation method for infant formula milk powder based on target nutrient structure inversion according to the present invention.

[0070] Figure 2 This is a system structure block diagram of the intelligent batching system of the present invention.

[0071] Figure 3 This is a schematic diagram showing the relationship between the candidate material feeding node set and the production process path of the present invention.

[0072] Figure 4 This is a schematic diagram illustrating the calculation principle of the cumulative thermal exposure, oxygen exposure, and shear exposure of the raw materials of this invention under different candidate feeding nodes.

[0073] Figure 5 This is a schematic diagram of the piecewise inversion solution model of the present invention.

[0074] Figure 6 This is a schematic diagram of the node pre-correction control process of the present invention.

[0075] Figure 7 This is a schematic diagram illustrating the relationship between traceability binding and data storage in this invention.

[0076] Figure 8 This is a schematic diagram illustrating the deployment of the intelligent ingredient dispensing system of the present invention in an infant formula milk powder production line.

[0077] Figure 9 This is a comparison chart of the effective retention rates of different raw materials at different candidate feeding nodes in this invention.

[0078] Figure 10 This is a comparison chart showing the deviations of the key nutrient structure indicators between the embodiments and comparative examples of the present invention.

[0079] Figure 11 This is a comparison chart of the retention rates of active components between the embodiments and comparative examples of the present invention.

[0080] Figure 12 This is a comparison chart of the overall performance indicators of the embodiments and comparative examples of the present invention. Detailed Implementation

[0081] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of the present invention.

[0082] I. Terminology Explanation

[0083] To make the technical solution of this invention clearer, some terms used in this application are explained below. Unless otherwise stated, these terms have the following meanings in this application:

[0084] Target nutritional structure set: refers to the combination of basic nutritional indicators, structural nutritional indicators, and process constraint indicators that the final product is expected to achieve.

[0085] Candidate feeding nodes: refer to predefined process nodes on the production line that allow raw materials to enter the system, including wet main batching nodes, nodes before homogenization, nodes before concentration, dry mixing nodes after spray drying, and addition nodes before and after packaging, etc.

[0086] Basic milk phase raw material group: refers to the raw material group mainly used to construct the main composition of milk, such as skim milk, whole milk, whey liquid, lactose liquid, mineral salt liquid, etc.

[0087] Fat structure regulating raw material group: refers to raw material group mainly used to regulate fat content, fatty acid composition or fat structure characteristics, such as light cream, vegetable oil, structured lipids, DHA / ARA oils, etc.

[0088] Thermosensitive active ingredient groups: These refer to ingredient groups that are sensitive to heat, oxygen, or shear, and whose effective function depends on the amount retained in the final product, such as lactoferrin, osteopontin, probiotics, vitamins, and microencapsulated functional oils.

[0089] Cumulative exposure: refers to the cumulative amount of heat, oxygen, or shear exposure that a raw material experiences beyond its critical conditions in subsequent process paths after entering a candidate feeding node.

[0090] Effective retention rate: refers to the proportion of a raw material that, after entering the system through a designated feeding node and undergoing subsequent processes, can still be retained in the final product in an effective state.

[0091] Allowable feed node subset: refers to the set of feasible nodes selected from all candidate feed nodes for a given raw material, based on its exposure sensitivity parameters and process adaptation constraints.

[0092] Segmented inversion solution: refers to the process of grouping raw materials with different functional attributes and solving for the target addition amount and target feeding node under node constraints and retention rate compensation conditions.

[0093] Pre-node correction: refers to recalculating the effective retention rate and correcting the addition amount and control actions of the feeding node before a candidate feeding node is executed, using the latest online detection parameters and current process parameters.

[0094] II. Technical Problem and Overall Idea to be Solved by the Invention

[0095] In current infant formula production processes, the theoretical addition amounts of each ingredient are typically set first based on the formula's objectives, and then ingredient control is implemented according to a fixed process route and predetermined feeding points. While this method can generally ensure that the main nutritional components of the final product meet standards, it struggles to address the following complex issues: First, the composition parameters and quality status of raw materials vary significantly between different batches, leading to discrepancies in the actual results of the same theoretical formula across different batches; second, heat-sensitive, oxygen-sensitive, and shear-sensitive ingredients undergo different subsequent process exposures after entering the system at different feeding points, resulting in significantly different actual effective retention amounts; third, infant formula increasingly emphasizes structural nutritional objectives, such as the whey protein to casein ratio, etc. - The proportion of palmitic acid, the ratio of prebiotics, and the ratio of long-chain polyunsaturated fatty acids are all affected by the amount of raw materials added, as well as the feeding point and subsequent process conditions. Fourth, although there are ideas such as post-addition, dry mixing, and encapsulation in the existing process, most of them rely on experience and are difficult to dynamically adjust according to real-time operating conditions.

[0096] This invention addresses the aforementioned problems by proposing an intelligent ingredient formulation method and system for infant formula milk powder based on target nutrient structure inversion. Instead of simply calculating the theoretical feed amount based on the final product target value, this method first calculates the cumulative heat exposure, cumulative oxygen exposure, and cumulative shear exposure for different raw materials at different candidate feed points in their subsequent process paths. Then, it uses a retention rate function to obtain the effective retention rate of the raw material at that point. Further, it filters a subset of allowable feed points based on the effective retention rate, raw material functional attributes, and process adaptation constraints. On this basis, it establishes segmented inversion solution models for the basic milk phase raw material group, the fat structure regulating raw material group, and the thermosensitive active raw material group, respectively, to obtain the target addition amount and target feed point that meet the requirements of the target nutrient structure set. Finally, before each candidate feed point that has not yet been executed, it re-executes the node screening and segmented solution for the remaining un-feeded raw materials using the latest online detection parameters to achieve dynamic correction before the node.

[0097] Thus, this invention forms a clear innovative technical route: target nutrient structure modeling → real-time raw material parameter acquisition → candidate feeding node path modeling → cumulative exposure calculation → effective retention rate determination → allowed feeding node screening → segmented inversion solution → node pre-correction execution → traceability binding storage.

[0098] III. System Structure of the Invention

[0099] Combination Figure 2 As shown, the intelligent ingredient dispensing system of the present invention includes at least the following modules:

[0100] 1. Target nutrient structure modeling module

[0101] This module receives product formulation design objectives, regulatory requirements, product stage requirements, and process performance requirements, and establishes a target nutrient structure set. The target nutrient structure set includes a set of target basic nutrient indicators. Target structural nutritional index set and the set of target process constraints In this embodiment, At least include the target protein content Target fat content Target lactose content Target ash content and target energy density ; At least include the target whey protein to casein mass ratio ,Target - Palmitic acid percentage Target prebiotic ratio Ratio with target long-chain polyunsaturated fatty acids ; At least include the target osmotic pressure range Target oxidation risk ceiling Target adjustment index Lower limit of target activity retention Upper limit of target moisture content and the upper limit of the target insoluble index .

[0102] 2. Raw material parameter acquisition module

[0103] This module connects to online near-infrared spectrometers, online fat analyzers, online protein analyzers, online densitometers, online viscometers, temperature sensors, oxygen concentration sensors, and other detection devices to collect the compositional parameters and exposure-sensitive parameters of each candidate raw material in real time. The compositional parameters include at least the... Protein content of the raw materials Fat content lactose content Ash content And parameters corresponding to structural nutrition objectives; exposure-sensitive parameters include at least the thermosensitive critical temperature. Oxygen-sensitive critical oxygen concentration Shear sensitivity threshold Meanwhile, this module automatically categorizes raw materials into three groups based on their properties: basic emulsion raw materials, fat structure regulating raw materials, and thermosensitive active raw materials.

[0104] 3. Exposure Calculation Module

[0105] This module calculates the cumulative heat exposure of any raw material at any candidate feeding node based on the process path data after each candidate feeding node. Cumulative oxygen exposure and cumulative shear exposure This module can access historical process databases, online process parameters, and equipment operating status information to construct real-time trajectories after each node.

[0106] 4. Retention Rate Calculation Module

[0107] This module calls the retention rate function corresponding to the raw material based on the exposure calculation results. The effective retention rate of raw materials at different feeding points was determined. Different retention rate functions can be used for different raw materials, such as experimental fitting functions, lookup tables, or piecewise functions.

[0108] 5. Node Filtering Module

[0109] This module is based on the effective retention rate. The raw material group and process compatibility constraints are considered when selecting a subset of allowed feeding nodes for each raw material from all candidate feeding nodes. For thermosensitive active raw material groups, feeding nodes are only allowed if their effective retention rate at a certain node is not lower than a threshold. Only when this condition is met will the node be included in the subset of nodes allowed to feed materials.

[0110] 6. Piecewise Inversion Solution Module

[0111] This module is one of the core modules of this invention. It solves for the target addition amount and target feeding node in stages according to the raw material group. The first stage solves for the basic emulsion raw material group to construct the basic nutritional framework; the second stage solves for the fat structure regulating raw material group to meet the requirements of fat structure and emulsion system; the third stage solves for the heat-sensitive active raw material group to meet the activity target of the final product with effective retention rate compensation.

[0112] 7. Execution Control Module

[0113] This module controls the actuators of liquid metering devices, powder metering devices, wet main ingredient feeding devices, homogenization pre-feeding devices, concentration pre-feeding devices, dry mixing feeding devices, and pre- and post-packaging addition devices, and implements feeding according to the results of the segmented inversion solution.

[0114] 8. Pre-node correction module

[0115] Before any candidate feeding node that has not yet been executed, this module recalculates the effective retention rate under the remaining candidate nodes based on the latest online detection parameters and current process parameters, and re-executes node screening and segmented inversion solution for the un-inputted raw materials to update its target addition amount and target feeding node.

[0116] 9. Traceability Storage Module

[0117] This module binds and stores raw material batch information, candidate feeding node selection results, exposure results, effective retention rate results, segmented solution results, online detection parameters, and correction records with the final product batch, forming a complete quality traceability chain.

[0118] IV. Specific Technical Route for Implementing the Method of the Invention

[0119] Combination Figures 1 to 8 The specific technical route of the method of the present invention is as follows:

[0120] First, a set of target nutritional structures for the final product and a set of candidate feed nodes available on the production line are established. Then, real-time composition and exposure sensitivity parameters of all candidate raw materials are collected, and the raw materials are grouped according to their functional attributes. Next, the system simulates the subsequent process path of each raw material after entering from each candidate feed node, calculates its cumulative exposure to heat, oxygen, and shear, and thereby obtains the effective retention rate at each node. Subsequently, combined with process adaptation constraints, a subset of allowable feed nodes for each raw material is selected. Then, using a piecewise inversion solution model, the target addition amount and target feed nodes are solved sequentially for the basic emulsion raw material group, the fat structure regulating raw material group, and the thermosensitive active raw material group. After the solution results are sent to the execution control module, each raw material is fed at the corresponding node. Whenever a candidate feed node has not yet been executed and the process status or online detection data changes, the system triggers a correction before that node, re-evaluates the effective retention rate of the remaining un-feeded raw materials at the remaining candidate nodes, and re-executes node screening and piecewise solution to ensure that the predicted nutritional structure of the final product still meets the target requirements. Finally, the actual material input results and the entire process data are linked into the database.

[0121] The following provides a detailed explanation of the specific implementation methods for each step.

[0122] (I) Step S1 – Establishment of the target trophic structure set and the candidate feeding node set

[0123] Step S1 primarily elevates the product design objective from a traditional single-component objective to a composite objective encompassing nutrition, structure, and process, and explicitly defines the feeder freedom of the production line as a finite set of candidate feeder nodes. Although this step belongs to the preparation stage, it provides fundamental support for the subsequent solution scope and technical effectiveness.

[0124] In this embodiment, the target trophic structure set is organized according to a three-layer structure.

[0125] The first layer is a set of target basic nutritional indicators. This set is used to describe the target values ​​for the final product in terms of common nutritional components. For example, the target protein content. Target fat content Target lactose content and target ash content It can be directly derived from formulation design documents, regulatory standards, or product technical standards; target energy density This is determined by the total energy target per unit mass of product. During project implementation, It is mainly used to control the amount of basic emulsion raw material group and some fat structure regulating raw material group added.

[0126] In this embodiment, The example deviation limits for each indicator are as follows: The allowable deviation is ±0.20 percentage points. The allowable deviation is ±0.25 percentage points. The allowable deviation is ±0.40 percentage points. The allowable deviation is ±0.10 percentage points. The permissible deviation is ±5 kcal / 100g. If the company's internal control standards are more stringent, this can also be adjusted. and The allowable deviations were tightened to ±0.10 percentage points and ±0.15 percentage points, respectively.

[0127] In this embodiment, The example deviation limits for each indicator are as follows: The allowable deviation is ±2.0 percentage points. The allowable deviation is ±2.0 percentage points. The allowable relative deviation is ±10%. The allowable relative deviation is ±10%. For example, when the target... When the ratio is 60:40, the final product test results are between 58:42 and 62:38, which can be considered to meet the structural protein ratio deviation limit requirements.

[0128] The third layer is the set of target process constraint indicators. This set describes the process performance or quality constraints that the final product should meet while satisfying nutrient structure targets. For example, target osmotic pressure range. Used to control the ionic and solute balance of the prepared liquid; target upper limit of oxidation risk. Used to limit the oxidation risk of highly unsaturated fatty acids and active ingredients; target reconstitution index Used to describe comprehensive properties such as wettability, dispersibility, and solubility; lower limit of target activity retention. Used to control the effective retention ratio of active ingredients such as lactoferrin, osteopontin, or probiotics in the final product.

[0129] In this embodiment, The example deviation limits for each indicator are as follows: wetting time is allowed to exceed the target value by no more than 2 seconds, insolubility index is allowed to exceed the target value by no more than 0.10 mL, peroxide value is allowed to exceed the upper limit of the target by no more than 0.02 g / 100 g, and moisture content is allowed to exceed the upper limit of the target by no more than 0.2 percentage points. The lower limit of target activity retention is also specified. Lactoferrin can be set to no less than 85%, osteopontin can be set to no less than 85%, probiotic survival rate can be set to no less than 80%, and the comprehensive retention rate of compound vitamins can be set to no less than 85%.

[0130] In parallel with the target trophic structure set, a candidate feeding node set is established. Figure 3 In this embodiment, the candidate feeding node set includes at least: wet main batching node, pre-homogenization node, pre-concentration node, post-spray drying dry mixing node, and pre- and post-packaging addition nodes. The purpose of this set is to transform the raw material feeding location from manually predetermined to a system-selectable set of process nodes, thereby providing selection space for subsequent node screening and segmented solution. It should be noted that in actual production lines, if additional intermediate nodes exist, such as fluidized bed spraying nodes, post-vacuum degassing nodes, or secondary dry mixing nodes, they can also be included in the candidate feeding node set.

[0131] Through step S1, this invention establishes a dual boundary for subsequent control: on the one hand, it defines what the product must achieve using the target nutrient structure set; on the other hand, it defines where the production line is allowed to achieve this using the candidate feed node set. This dual boundary setting is a key starting point that distinguishes this invention from simple static formula calculation.

[0132] (II) Step S2 – Real-time raw material parameter acquisition and raw material grouping

[0133] Step S2 is used to collect real-time composition parameters and exposure sensitivity parameters of the raw materials and complete the functional grouping of the raw materials. This step provides basic data for subsequent exposure calculation and segmented solution.

[0134] In this embodiment, for each candidate raw material, the system collects at least the following real-time composition parameters: protein content. Fat content lactose content Ash content For raw materials with special structural nutritional functions, their corresponding structural characteristic parameters are also collected. For example, the whey protein content of whey protein powder is collected, and the structural lipid content is collected. - The parameters for palmitic acid content were collected, along with the compositional ratios of prebiotic ingredients and the fatty acid composition of DHA / ARA oils. These parameters were obtained through a combination of online near-infrared spectroscopy, online fat analyzer, online protein analyzer, and rapid offline calibration results from the laboratory.

[0135] Regarding the exposure of sensitive parameters, this implementation method... The raw materials must at least define the thermosensitive critical temperature. Oxygen-sensitive critical oxygen concentration and shear sensitivity threshold .in, This refers to the critical temperature at which the key effective components in the raw material begin to show significant heat loss; This refers to the critical oxygen concentration at which the oxidation rate of the raw material increases significantly when the oxygen concentration exceeds this level. This refers to the threshold at which the activity or structure of the raw material is significantly damaged when the shear level is exceeded. These parameters can be derived from raw material supplier technical data, pre-established experimental databases, or internal company research results.

[0136] After parameter acquisition, the system groups the raw materials according to their functional attributes. In this embodiment, the raw materials are divided into three groups. The basic emulsion phase raw material group mainly serves to construct the milk matrix and basic nutritional framework; the fat structure regulating raw material group mainly serves to regulate fat content and fat structure; and the thermosensitive active raw material group mainly serves to provide specific active nutritional functions. This grouping is not a simple classification, but directly determines the order and constraints of subsequent segmented solutions. For example, the basic emulsion phase raw material group is solved first because it determines the basic nutrition and emulsion system of the product; the fat structure regulating raw material group is solved after the basic emulsion phase is formed to adapt to homogenization and emulsion stability requirements; and the thermosensitive active raw material group is solved last to maximize the use of effective retention information to select a more suitable feeding node.

[0137] Through step S2, the present invention transforms all raw materials from a single row of data in the formula table into dynamic objects with real-time composition, exposure sensitivity, and process role, creating conditions for subsequent node-based control.

[0138] (III) Steps S3 and S4 – Calculation of cumulative exposure and determination of effective retention rate

[0139] The main purpose of steps S3 and S4 is to transform the empirical understanding that adding different nodes will result in different retention effects into quantifiable, computable, and inverse-solution technical variables.

[0140] 1. Calculation of cumulative heat exposure

[0141] For the The raw material, when it is intended to be in the first After a candidate feed node is added, it will undergo the subsequent process path following that node. For example, if it is added at the wet main feed node, it will subsequently undergo homogenization, heat treatment, concentration, spray drying, fluidized bed treatment, dry mixing, and packaging; if it is added at the dry mixing node after spray drying, it will only undergo dry mixing and packaging. The system determines the process path based on the temperature trajectory along that path. Calculate the cumulative amount of heat exposure:

[0142] ;

[0143] in, Indicates the first The first type of raw material in Cumulative heat exposure at each candidate feeding node; and These represent the times when the raw material enters and leaves the subsequent path from this node, respectively. This represents a real-time temperature function along the subsequent process path; This represents the thermosensitive critical temperature of the raw material. The formula uses... This approach aims to accumulate only the portion of heat exposure exceeding the critical temperature, thus better reflecting the actual mechanism of activity loss.

[0144] 2. Calculation of cumulative oxygen exposure and cumulative shear exposure

[0145] Similarly, the cumulative oxygen exposure and cumulative shear exposure are calculated using the following methods:

[0146] ;

[0147] in, Indicates the first The first type of raw material in Cumulative oxygen exposure at each candidate feeding node; This represents the oxygen concentration function along the process path after this node; This indicates the oxygen-sensitive critical oxygen concentration of the raw material.

[0148] ;

[0149] in, Indicates the first The first type of raw material in Cumulative shear exposure at each candidate feeding node; This represents the equivalent shear strength function along the process path following this node; This indicates the shear sensitivity threshold of the raw material.

[0150] In practice, , and Data can be collected from online sensors on the production line, or predicted and generated from historical databases, process models, or digital twin modules. For certain equipment such as homogenizers, shear pumps, nozzles, or high-speed mixers, the equivalent shear strength can be estimated using equipment speed, pressure difference, flow rate, and viscosity.

[0151] 3. Calculation of effective retention rate

[0152] After obtaining the three types of cumulative exposure, the system targets the first... The raw material calls a pre-established retention rate function Calculate its in the first Effective retention rate under each candidate feeding node:

[0153] ;

[0154] in, Indicates the first The first type of raw material in The proportion of candidate feeding nodes that remain in an effective state in the final product after subsequent processes; This represents the retention function established for this raw material. This function can be obtained by fitting experimental data. For example, for lactoferrin, the function can be established through activity retention experiments under different combinations of heat and oxygen exposure; for probiotics, the function can be established through survival experiments under different temperature, oxygen concentration, and shear conditions; for structured lipids or DHA / ARA microcapsules, the function can be established through changes in structure retention rate or oxidation value.

[0155] The technical effect of steps S3 and S4 is that they concretize the previously empirical judgment of which raw material is suitable for pre-addition and which is suitable for post-addition into a retention rate matrix for different raw materials at different nodes. This matrix serves as the direct basis for subsequent screening of the subset of allowed feeding nodes and for compensatory feeding. In other words, this invention does not treat post-addition as a pre-set process, but rather bases the necessity and optimality of post-addition on the calculation of exposure integrals and retention rates.

[0156] (iv) Step S5 - Allowing the filtering of a subset of feeding nodes

[0157] Step S5 is used to select a subset of allowed feeding nodes for each raw material from all candidate feeding nodes. This step is the bridge connecting raw material properties and process selection, and it is also a key step in this invention to move from static formulation to dynamic node allocation.

[0158] In this embodiment, node selection considers at least three types of factors: first, the effective retention rate requirement; second, the raw material group attributes; and third, the process adaptation constraints.

[0159] First, the system sets an activity retention threshold for the raw materials in the thermosensitive active raw material group. Only when the effective retention rate of the raw material at a certain node... Not less than Only when this condition is met will the node be included in the subset of allowed feeding nodes. For example, if lactoferrin's effective retention rate is only 0.42 after being added at the wet master ingredient node, while the activity retention threshold of lactoferrin... If the retention threshold is 0.85, the wet-process main ingredient node will not be included in its allowed ingredient addition node subset; however, the effective retention rate of the spray-dried dry-mix node is 0.88, which is higher than 0.85, so this node can be included in its allowed ingredient addition node subset. For example, the activity retention threshold for lactoferrin is 0.85, the activity retention threshold for osteopontin is 0.85, the activity retention threshold for probiotics is 0.80, the overall retention threshold for multivitamins is 0.85, and the effective retention threshold for DHA / ARA microcapsule oil is 0.80.

[0160] Secondly, for the raw materials in the fatty acid structure-modifying raw material group, in addition to considering retention rate, the requirements for emulsion formation and homogenization stability are also emphasized. For example, if the structured lipid is added at the dry-mixing node after spray drying, although there is no significant heat loss, it is difficult to participate in wet homogenization and emulsion structure formation, which is not conducive to achieving the target. - The system requires a reasonable distribution of palmitic acid in the finished fat globules. Therefore, for this type of raw material, the system also requires that the nodes it enters meet the constraints of pre-homogenization emulsion formation and dispersion stability. In this embodiment, fat structure-modifying raw materials are preferentially screened at the wet-process main ingredient node or the pre-homogenization node.

[0161] Furthermore, for raw materials in the basic emulsion feedstock group, it is generally required that they enter the system in the wet process stage to ensure the formation of the basic emulsion, subsequent homogenization stability, and smooth concentration spray processing. Therefore, the allowable feedstock node subset for this type of raw material is generally concentrated at the wet main feedstock node, the pre-homogenization node, and the pre-concentration node.

[0162] Through step S5, each raw material forms a subset of permissible feeding nodes for the current batch and current process status, rather than a fixed pre- or post-feeding path. This subset selection result comes from both quantitatively calculated effective retention rates and process adaptation logic, thus balancing nutrient retention and processing feasibility.

[0163] (V) Step S6 – Piecewise inversion solution of the model

[0164] In existing technologies, even with automated batching control, most methods involve substituting all raw materials into a solution model to calculate a total addition amount, and then executing the process according to predetermined process nodes. This approach has significant limitations: different raw materials have different functional roles, process dependencies, and retention behaviors, and a uniform solution can easily lead to an overly coarse model, making it difficult to simultaneously satisfy multiple objectives related to nutrition, structure, and process. This invention, however, employs a grouping-segmentation-node inversion solution method.

[0165] 1. First stage: Solving the basic emulsion feedstock composition

[0166] In the first stage, the system only solves for the target addition amount and node allocation for the basic emulsion raw material group. The goal of this stage is to form a set of basic nutritional indicators. The required emulsion base. Since the base emulsion feedstock usually enters the wet process, the freedom of its node selection is relatively small, but its addition amount has a decisive impact on the subsequent system concentration, viscosity, homogenization adaptability and concentration spray load.

[0167] In this embodiment, the system uses the target protein content Target fat content Target lactose content Target ash content and target energy density As the primary constraint, the system calculates the required amounts of basic emulsion raw materials such as skim milk, whole milk, whey, and lactose solution by combining real-time composition parameters of the raw materials. If there are multiple alternative sources among the basic emulsion raw materials, the system will also consider process constraints to avoid excessive viscosity, excessive spray load, or excessive osmotic pressure caused by excessive solids.

[0168] During this phase, the deviation item of the basic nutritional index The internal weights can be set as follows: protein content deviation weight 0.30, fat content deviation weight 0.30, lactose content deviation weight 0.25, and ash content deviation weight 0.15. The system multiplies the absolute deviation between the predicted and target values ​​of each basic nutritional indicator by their corresponding weights and then sums them to obtain the basic nutritional indicator deviation item. If any basic nutritional indicator exceeds the deviation limit set in step S1, the system increases the corresponding penalty value and readjusts the amount of each ingredient added to the basic milk phase raw material group.

[0169] 2. Second Stage: Solving for the Raw Material Set for Fat Structure Regulation

[0170] In the second stage, based on the results of the first stage, the system solves for the raw material group for adjusting the fat structure. The core of this stage is to meet the target fat content. ,Target - Palmitic acid percentage and the ratio of long-chain polyunsaturated fatty acids It has structural nutritional indicators, while ensuring the stability of emulsion dispersion and homogeneity.

[0171] Unlike the first stage, this stage not only determines the amount to add, but also the most appropriate node for addition. For example, for cream and structured lipids, the system typically prioritizes distribution to the wet main ingredient node or the pre-homogenization node so that they can participate in the homogenization process to form a stable emulsion; for some microencapsulated oils, they may be added at the pre-concentration node or the dry-mixing node after spray drying to reduce oxidative exposure while meeting retention requirements.

[0172] During this stage, the structural nutrient index deviation item The internal weighting can be set as follows: the weighting for the deviation in the mass ratio of whey protein to casein is 0.30. - The weighting for deviations in palmitic acid ratio is 0.35, prebiotic ratio is 0.15, and DHA:ARA ratio is 0.20. These weights reflect the priority control of lipid and protein structure targets in this implementation method. If the target product emphasizes the ratio of long-chain polyunsaturated fatty acids such as DHA / ARA, the weighting for deviations in the DHA:ARA ratio can be increased to 0.25 to 0.30, and the weightings of other structural indicators can be reduced accordingly.

[0173] 3. Third stage: Solving for the thermosensitive active raw material group

[0174] The third stage is where the differences between this invention and existing technologies are most evident. The raw materials in the thermosensitive active material group are often most susceptible to process node influences. Their target feed quantity is not simply equal to the target demand of the final product, but should be compensated for based on the effective retention rate. Within the subset of allowed feed nodes, the system selects nodes for each thermosensitive active material that satisfy process constraints and have the highest possible retention rate, and calculates the compensated target feed quantity.

[0175] For the A type of thermosensitive active raw material, when it is allocated to the first When there are multiple candidate feeding nodes, the target feeding amount is... At least the following conditions must be met:

[0176] ;

[0177] in, This indicates the target feed amount of the raw material at this node; This indicates the effective retention rate at that node; This indicates the target quantity of the raw material in the final product. The meaning of the formula is: because the raw material is lost in subsequent processes, the actual amount fed in must be compensated for by the retention rate to still meet the target value of the final product.

[0178] In this embodiment, the system sets a penalty for insufficient activity retention of thermosensitive active raw materials. The internal weights for this penalty item can be: 0.35 for insufficient lactoferrin retention, 0.25 for insufficient osteopontin retention, 0.30 for insufficient probiotic survival rate, and 0.10 for insufficient retention of compound vitamins. When a certain active ingredient still cannot meet the target demand of the final product after effective retention rate compensation, the system adds a penalty item according to the corresponding weight and reselects the node or adjusts the target feed amount.

[0179] 4. Combine objective function and constraints

[0180] To simultaneously balance basic nutritional targets, structural objectives, preventing process constraints from being breached, activity retention, and cost control, this implementation method adopts the following comprehensive optimization objective function:

[0181] ;

[0182] in, To comprehensively optimize the objective function; The deviation item of the basic nutritional index; This refers to the deviation item of structural nutrition indicators; This is a penalty item for process constraints; Penalty for insufficient activity retention; This is a raw material cost item; , , , and These are the corresponding weighting coefficients, all of which are positive values.

[0183] In this embodiment, 0.30 is acceptable. 0.25 is acceptable. 0.20 is acceptable. 0.15 is acceptable. A value of 0.10 can be used. The sum of the above weighting coefficients is 1.00, which facilitates model normalization. For products that emphasize the retention of active factors, the weighting coefficient can be... Increase to 0.20 to 0.25; for cost-sensitive products, it can be increased to 0.20 to 0.25. Increase to 0.15 to 0.20.

[0184] The deviation from the basic nutritional indicators can be calculated as follows:

[0185] ;

[0186] in, The number of basic nutritional indicators used in the calculation; For the first Predicted values ​​of one basic nutritional indicator; For the first Target values ​​for each basic nutritional indicator; These are the corresponding weighting coefficients.

[0187] Similarly, the structural nutrient index deviation item It can also be obtained by weighted summation of the deviations between the predicted and target values ​​of each structural indicator. (Process constraint penalty term) Used to constrain targets such as osmotic pressure, water activity, insoluble index, and rehydration properties from being exceeded; a penalty for insufficient activity retention. This is used to penalize situations where the heat-sensitive active raw material still fails to meet the final product target even after compensation.

[0188] 5. The innovative effect of this step

[0189] The technical advantage of step S6 lies not in solving the problem using a single optimization model, but in explicitly incorporating the process adaptability, effective retention rate, and structural nutritional contribution of different raw materials into the solution process through piecewise inversion. Thus, the solution result is no longer just a formula table, but a three-dimensional result of raw materials, nodes, and addition amounts. Compared to existing technologies that only uniformly and automatically add materials according to target standards, this invention can more accurately reflect the true effective contribution of different raw materials at different nodes, thereby significantly improving the accuracy of achieving the target nutritional structure of the final product.

[0190] (vi) Step S7 - Feed materials according to the piecewise inversion solution results and collect online detection parameters before the nodes.

[0191] Step S7 is a crucial execution step that transforms the calculation results into production actions after completing steps S1 to S6. Following the aforementioned target nutrient structure modeling, raw material parameter acquisition, cumulative exposure calculation, effective retention rate determination, allowable feed node subset screening, and segmented inversion solution, the system has obtained the target addition amount, target feed node, and necessary compensation feed amount for each raw material. The task of step S7 is to control various metering devices and feed execution devices at each node to implement feed at the corresponding process nodes according to the solution results; simultaneously, it acquires the current process parameters and online detection parameters before each candidate feed node, providing real-time basis for the node pre-correction in the subsequent step S8.

[0192] 1. Overall execution logic of step S7

[0193] Combination Figure 2 and Figure 8 Step S7 can be specifically divided into the following sub-processes:

[0194] (1) Receive the execution instructions output by the piecewise inversion solution module;

[0195] (2) Assign tasks to each raw material according to the target feeding node;

[0196] (3) Use a liquid metering device or a powder metering device to complete the quantitative feeding;

[0197] (4) Control the feeding execution device of the corresponding node to complete the feeding within the specified window;

[0198] (5) Before each candidate feeding node that has not yet been executed, collect the current process parameters and online detection parameters;

[0199] (6) The execution data of this round is temporarily stored and transmitted to the node pre-correction module and the trace storage module.

[0200] In other words, step S7 is not simply adding materials according to the formula, but rather revolves around four questions: when to add each raw material, how much to add, from where to add it, and what the current operating conditions are before adding it. The key technology lies in this invention: it binds execution control and online status acquisition within the same process, ensuring that the material addition at each node is not an isolated event, but rather a stateful event that can be invoked by subsequent correction logic.

[0201] 2. Hardware composition and controlled object of the execution control module

[0202] In this embodiment, the execution control module includes at least:

[0203] Liquid metering device: used for continuous or intermittent quantitative metering of liquid raw materials such as skim emulsion, whole emulsion, whey liquid, lactose liquid, light cream, and some oil liquids;

[0204] Powder metering device: used for high-precision weighing of powdered or trace raw materials such as lactoferrin, osteopontin, probiotics, compound vitamins, microcapsule oil powder, and mineral salt powder.

[0205] Wet main ingredient feeding device: installed at the inlet of the main ingredient tank or main ingredient pipeline;

[0206] Pre-homogenization feeding device: A feeding branch installed before the homogenizer inlet;

[0207] Pre-concentration feeding device: a feeding port installed before the evaporator;

[0208] Dry mixing feeding device: installed before the base powder enters the mixing equipment after spray drying or during the mixing process;

[0209] Pre- and post-packaging addition device: installed at the end addition point before final packaging;

[0210] Controller: This can be a PLC, an industrial computer, or a control unit linked with MES / SCADA, used to receive the segmented inversion solution results and output control commands to each actuator.

[0211] In some implementations, the liquid metering device includes an electromagnetic flow meter, a mass flow meter, a servo control valve, and a liquid metering pump; the powder metering device includes a loss-in-weight balance, a micro-screw feeder, a powder buffer silo, and a pneumatic conveying unit. For trace amounts of high-value raw materials, such as lactoferrin, probiotics, or certain vitamins, a two-stage metering structure can also be used, i.e., coarse metering followed by fine metering, to improve the accuracy of addition.

[0212] 3. Segmented execution method

[0213] Since the present invention uses a piecewise inversion solution model in step S6, the execution of step S7 is also carried out in a piecewise manner, rather than issuing all raw materials at once as a single execution formula.

[0214] (1) Implementation of the basic emulsion raw material group

[0215] For the basic emulsion raw material group, such as skim emulsion, whole emulsion, whey, lactose solution, and some mineral salt solutions, the execution control module prioritizes feeding at the wet main ingredient node or the node before homogenization. These raw materials usually account for the majority of the total feed volume, and their role is to form the basic emulsion phase and the main nutritional framework.

[0216] For example, when the system receives an execution command to add 486.0 kg of skim emulsion at node J1, the controller sends control signals to the liquid metering pump and regulating valve on the skim emulsion branch, instructing them to deliver the corresponding volume or mass of skim emulsion into the main batching tank within a specified time window. The online flow meter provides real-time feedback on instantaneous and cumulative flow rates. When the target value is reached, the controller closes the regulating valve and records the execution result. Whole milk emulsion, whey liquid, and lactose liquid are processed in the same manner.

[0217] In some implementations, the execution of the basic emulsion raw material group requires not only accurate total amounts but also controlled order of addition. For example, a portion of the defatted emulsion can be introduced first as the main solvent, followed by lactose solution and mineral salt solution, to avoid sudden increases in local osmotic pressure; alternatively, whey can be introduced first, followed by whole milk emulsion, thereby optimizing system homogeneity. These sequential requirements can be issued by the execution control module based on preset logic or by the segmented solution module.

[0218] (2) Implementation of raw material group for fat structure adjustment

[0219] For lipid structure modifiers such as cream, OPO structured lipids, and DHA / ARA microcapsule oils, the process must consider both the node location and emulsion formation conditions. In particular, for cream and OPO structured lipids, if they are allocated to a node before homogenization, it is necessary to ensure that they have undergone preliminary dispersion and formed a homogenizable lipid phase before entering the homogenizer.

[0220] For example, if the system determines that 18.0 kg of OPO structured lipid is added at node J2, the controller will, before the main emulsion phase is established and the homogenizer starts main homogenization, control the pre-homogenization feeding device to introduce the OPO structured lipid into the feed branch at a set flow rate. Simultaneously, the branch temperature, flow rate, and viscosity are monitored to ensure that the OPO structured lipid forms a stable premix with the main emulsion phase before entering the homogenization valve. If the current branch temperature is lower than the preset minimum dispersion temperature, the execution control module can first start jacket heating or delay feeding until the process window is reached before execution.

[0221] For DHA / ARA microcapsule oils, the execution method differs depending on whether they are allocated to the pre-concentration stage or the dry-mixing stage:

[0222] When allocated to the pre-concentration node, it needs to be added before evaporation using a liquid or slurry metering device, and the residence time between addition and evaporation inlet should be controlled.

[0223] When the powder is allocated to the dry mixing node after spray drying, it is added by the powder metering device when the base powder enters the mixer, and the mixing time and mixing uniformity are recorded.

[0224] (3) Implementation of the thermosensitive active raw material group

[0225] For heat-sensitive active ingredients such as lactoferrin, osteopontin, probiotics, and multivitamins, the execution of step S7 is the part that requires the most precise control. These ingredients are typically added in small quantities, are of high value, and are highly sensitive to the process, thus requiring higher precision in metering, timing of material addition, and pre-execution conditions.

[0226] For example, when lactoferrin is assigned to node J4 (dry mixing node after spray drying), the system will activate the powder loss-in-weight scale and micro-screw feeder before the base powder enters the three-dimensional mixer, slowly adding the lactoferrin powder to the mixer according to the solution results; simultaneously, the system will record the current mixer speed, base powder flow rate, ambient temperature and humidity, and local oxygen concentration online. If probiotics are assigned to node J5 (addition node before and after packaging), the system will add them at the end of the pre-packaging stage using pneumatic conveying or small-dose loss-in-weight feeding, and control the time between addition and sealing within a preset upper limit to reduce exposure time.

[0227] 4. Collection of online detection parameters before nodes

[0228] A key difference between step S7 and traditional execution control is that current process parameters and online detection parameters must be collected before each candidate feeding node. These parameters are not only used to confirm whether the current node meets the execution conditions, but more importantly, they serve as the basis for recalculating the effective retention rate and determining whether corrections are needed in step S8.

[0229] In this embodiment, the online detection parameters before the node include at least: online protein detection values. Online body fat measurement values Online lactose testing value Online temperature detection value Online oxygen concentration detection value Online viscosity measurement value When necessary, it can also include online moisture detection values, online flow detection values, online solids detection values, etc.

[0230] The location of these parameters is related to the corresponding node. For example, at the J2 homogenization node, the focus is on collecting the feed temperature, fat content, viscosity, and local flow rate; at the J4 dry mixing node, the focus is on collecting the ambient oxygen concentration, powder temperature, powder flow rate, and mixer status; at the J5 pre- and post-packaging addition nodes, the focus is on collecting the headspace oxygen concentration, local ambient temperature and humidity, and addition cycle time.

[0231] 5. Generation of execution records

[0232] Step S7 generates a node execution record after each node execution. This record includes at least: raw material name and batch number; target feeding node number; target addition amount; actual addition amount; execution start and end time; online detection parameters before the node; equipment status parameters during execution; and whether it deviates from the preset window. These execution records are transmitted to step S8 as a correction reference and directly passed to step S9 for traceability binding. In this way, if a certain indicator is abnormal later, it is possible to trace back to which specific raw material, at which node, and under what process state it was added.

[0233] (vii) Step S8 – Recalculate the effective retention rate and correct the amount of raw materials not added and / or the feeding node before the node is executed.

[0234] Step S8 addresses the issue that the original operating conditions used in the solution process no longer match the current actual operating conditions. Without this step, even if the result obtained in step S6 is theoretically optimal, fluctuations in the process conditions may cause changes in the effective retention rate of subsequent raw materials at actual nodes, resulting in the final product deviating from the target nutritional structure. Therefore, the essence of step S8 is to perform a forward check before each candidate feeding node that has not yet been executed: determining whether the raw materials originally allocated to subsequent nodes should still be executed according to the original plan under the latest operating conditions; if not, then only the nodes for the parts that have not yet been fed are re-selected and the solution is re-executed.

[0235] 1. Triggering timing of step S8

[0236] Step S8 is not a unified correction after the entire process is completed, but is triggered before any candidate feeding node that has not yet been executed. In other words, the system retains a correction window for each subsequent node.

[0237] For example:

[0238] When J1 has finished executing but J2 has not yet executed, a correction can be triggered before J2.

[0239] When J2 has finished executing but J3 has not yet executed, a correction can be triggered before J3.

[0240] If J4 has not yet been executed, the raw materials originally allocated to J4 or J5 can be corrected again before J4.

[0241] Once a node has been completed, the raw materials cannot be reassigned to a node preceding that node.

[0242] This pre-node correction design allows the invention to retain dynamic adaptability while avoiding control chaos caused by unbounded backtracking.

[0243] 2. Input information for step S8

[0244] The input for step S8 includes at least the following: the latest online detection parameters and current process parameters collected in step S7; the actual feeding results of executed nodes; the set of remaining candidate nodes for nodes not yet executed; the current remaining target demand for raw materials not yet added; and the subset of allowed feeding nodes and the effective retention rate matrix obtained from the previous round of solution. The current remaining target demand for raw materials not yet added is not a simple copy of the initial target demand, but rather needs to be re-estimated based on the actual amount of raw materials added and the latest detection values. For example, if the actual fat content is detected to be lower than expected before J2, the remaining target demand for subsequent fat structure adjustment raw material groups may increase accordingly; if the protein structure indicators are detected to be close to the target before J4, the compensation amount for certain active proteins may not need to continue to maintain its original value.

[0245] 3. Recalculate the effective retention rate for the remaining nodes.

[0246] The first core action of step S8 is to recalculate the effective retention rate of un-added raw materials at the remaining candidate feeding nodes. This recalculation does not recalculate all historical nodes, but only for nodes that can still be selected from the current moment.

[0247] If the current time is before node J4, then for a certain thermosensitive active material, it is only necessary to recalculate its effective retention rate at nodes J4 and J5; J1 to J3 are meaningless because these nodes can no longer be executed.

[0248] For example, regarding lactoferrin, if it was originally planned to be added at node J4, but the current detection at node J4 shows: online temperature detection value... Rise; Online oxygen concentration detection value The current mixer environment dwell time may be prolonged; therefore, the system needs to be based on the latest... , Based on the predicted shearing conditions along the remaining paths, the exposure and effective retention of lactoferrin at nodes J4 and J5 were recalculated.

[0249] For the The first type of raw material in the remaining number Under each candidate node, the cumulative amount of heat exposure can be obtained again. Cumulative exposure ;

[0250] Cumulative shear exposure And from this, a new effective retention rate is obtained:

[0251] ;

[0252] in, Indicates the first The raw material, under the latest operating conditions, in the remaining... The new effective retention rate after the addition of each candidate feeding node.

[0253] 4. Reassessment of the predicted nutritional structure of the final product

[0254] The second core action of step S8 is to determine whether the predicted nutrient structure of the final product still meets the requirements of the target nutrient structure set based on the latest online detection parameters, the results of the execution, and the recalculated effective retention rate.

[0255] This judgment can be made using either a comprehensive deviation criterion or a component threshold criterion. In this embodiment, a comprehensive deviation criterion can be used:

[0256] ;

[0257] in, This is the overall deviation value; The number of indicators involved in the judgment; For the first One predicted nutrient structure index value; For the first Target nutritional structure index values; This is the deviation weighting coefficient for the corresponding indicator.

[0258] In one specific implementation, the deviations of each indicator can be normalized first, and then the overall deviation value can be calculated. .when Not greater than the preset allowable threshold When the predicted nutrient structure of the final product still meets the requirements of the target nutrient structure set, it is determined that the final product still meets the requirements of the target nutrient structure set; when Greater than If the current plan is deemed to have a deviation risk, steps S5 and S6 need to be re-executed for any raw materials not yet added. As an example, 1.00 is acceptable.

[0259] In the comprehensive deviation assessment, the comprehensive deviation weights of each indicator can be set as follows: 0.30 for basic nutritional indicators, 0.30 for structural nutritional indicators, 0.20 for process constraint indicators, and 0.20 for activity retention indicators. If the current batch is a batch under general nutritional structure control, the weights of basic and structural nutritional indicators can be appropriately increased; if the current batch is a fortified product containing highly active components, the weight of the activity retention indicator can be increased.

[0260] The predicted nutrient structure index values ​​here It is not a static setpoint, but is calculated from the following information: the actual addition result of the raw materials that have been added; the current online detection value; the expected contribution of the raw materials that have not been added if they were added according to the original planned nodes and amounts; and the recalculated retention rate compensation result.

[0261] 5. Target and Scope of Correction

[0262] This invention specifically emphasizes that the correction only applies to raw materials for which the feeding node has not yet been executed. Nodes that have already been executed and their feeding results are considered established facts and will not be rolled back. There are three reasons for this design:

[0263] It conforms to the actual production process, and the raw materials that have been fed into the system cannot be reversed or withdrawn;

[0264] Reduce control complexity to ensure that the problem-solving process is always confined to the remaining degrees of freedom;

[0265] This makes the technical solution in the claims more feasible for engineering implementation.

[0266] The objects to be corrected fall into two categories:

[0267] Target addition amount adjustment: Keep the original node unchanged, only adjust the remaining addition amount of raw materials that have not been added;

[0268] Target feeding node correction: Within the subset of allowed feeding nodes, switch the unfeeded raw materials from the original node to another remaining feasible node; adjust the addition amount simultaneously if necessary.

[0269] The revised target feeding node must meet two conditions:

[0270] (1) Switching is only allowed within a subset of the permitted feeding nodes for the raw material;

[0271] (2) You may not switch to a node that has already been executed before the current node.

[0272] In some implementations, a minimum correction node number condition can also be set. That is, only when the number of remaining candidate feeding nodes is not less than the preset minimum correction node number... The system will only perform node correction when the number of remaining nodes is too small; if there are too few remaining nodes, only the number of nodes to be added will be adjusted, and no more nodes will be switched, in order to avoid excessive disturbance at the end.

[0273] In this embodiment, Option 2 is acceptable. That is, when there are at least two remaining candidate feeding nodes, the system allows switching to the target feeding node that has not yet been fed raw materials; when there are fewer than two remaining candidate feeding nodes, the system only allows adjustment of the target addition amount and does not switch nodes. If the production line only has a wet main batching node and a dry mixing node after spray drying, then... Alternatively, a value of 1 can be selected, in which case the system will primarily perform adjustments to the added amount.

[0274] 6. Repeat steps S5 and S6.

[0275] Once it is confirmed that correction is needed, the system does not redo the entire solution from scratch, but only re-executes steps S5 and S6 for those materials that have not yet been fed in.

[0276] In other words: re-screen the subset of allowed feeding nodes among the remaining candidate nodes for the un-feeded raw materials; on the updated subset of allowed feeding nodes, perform a piecewise inversion solution in combination with the new retention rate matrix; output the corrected target addition amount and / or target feeding nodes.

[0277] For example, in an embodiment, if lactoferrin was originally planned to be added at node J4, but an increase in oxygen concentration is detected before J4, causing its effective retention rate at J4 to decrease from 0.87 to 0.82, while remaining at 0.92 at J5, the system will re-screen the remaining allowable node set for lactoferrin to {J4, J5}, and then recalculate the optimal addition method at these two nodes to meet the final product target. If the overall result shows that J5 is better, the target addition node for lactoferrin will be changed from J4 to J5, and the addition amount will be recalculated accordingly.

[0278] 7. Issuance and confirmation of the revised results

[0279] After the correction is completed, the pre-node correction module sends the updated execution instructions to the execution control module. Upon receiving the correction result, the execution control module will: overwrite the subsequent execution tasks for the corresponding unexecuted raw materials in the original plan; retain the difference record between the original plan version and the corrected version; and wait to execute according to the correction result when entering the next candidate node. If no correction is triggered before a certain node, the system will record the state before that node where a judgment was made but no correction was required. This record is also valuable for subsequent process analysis because it can indicate at which node a batch maintained a stable state.

[0280] (viii) Step S9 – Output the final batching control command and bind and store the results of the entire process.

[0281] Step S9 is the final step of the method of this invention. Its function is not only to output the final executable ingredient control instructions, but more importantly, to bind and store all key decision information, process data, and correction records generated during the batch ingredient preparation process, forming a complete data traceability chain. Although step S9 has a lower inventive contribution than steps S3 to S8, it plays an important supporting role in industrial applications and quality control, and is particularly suitable for products such as infant formula milk powder, which have extremely high requirements for safety, stability, and traceability.

[0282] 1. Formation of final ingredient control instructions

[0283] After the execution of step S7 and one or more pre-node corrections in step S8, the system finally forms a complete final batching control instruction set. This instruction set is not just a static formula table, but a structured control result containing the following: the final target addition amount of each raw material; the final target feeding node for each raw material; the execution order and time window for each node; the actual results of the executed parts; the content of the unexecuted parts that needs to be executed according to the corrected plan; the number of pre-node corrections, the triggering reasons, and the final adopted version.

[0284] 2. Bind the data range to storage

[0285] The binding storage described in step S9 is not simply saving the final result, but rather mapping key data from the entire ingredient control process to the final product batch. The binding objects include at least the following categories:

[0286] (1) Raw material batch information, including but not limited to: raw material name; raw material batch number; raw material supplier or source information; raw material incoming inspection results; raw material real-time composition parameters; raw material exposure sensitive parameters.

[0287] (2) Target setting information, including: the set of basic nutritional indicators for this batch of targets. This batch of target structural nutrient indicators set Set of target process constraints for this batch ; Set of candidate feeding nodes; Initial subset of allowed feeding nodes.

[0288] (3) Process calculation results, including: cumulative heat exposure of each raw material at each candidate node. Cumulative oxygen exposure Cumulative amount of shear exposure Effective retention rate matrix The results of the segmented inversion at each stage; the initial target addition amount and the target feeding node.

[0289] (4) Execution process information, including: online detection parameters before the node , , , , , ; Execution start / end time; Equipment status at the actual execution node; Whether it exceeds the process window.

[0290] (5) Correction information, including: correction trigger node; correction trigger reason; change in effective retention rate before and after correction; target addition amount before and after correction; target feeding node before and after correction; number of corrections; final adopted scheme number.

[0291] (6) Product batch information, including: finished product batch number; production line number; production time; packaging batch; factory inspection results.

[0292] The above information can be stored in the form of a relational database, batch database, or industrial big data platform. To ensure data alignment between different devices, different detection points, and different control modules, a unified timestamp mechanism can be used in some implementations to synchronize all acquisition and execution events at the millisecond level.

[0293] 3. Methods for implementing binding relationships

[0294] The binding in step S9 is not simply placing data in the same folder, but establishing explicit data association keys. Common binding relationships include: using the production batch number as the primary key to associate the target nutrient structure set, the final ingredient control instructions, and the final product test results; using the raw material batch number as the secondary key to associate the raw material composition parameters, exposure sensitivity parameters, and exposure and retention rate results at each node; and using the node event number as the event key to associate the online detection parameters before the node, the node execution results, and the node correction records.

[0295] Through this multi-key binding method, the system can trace back from any dimension during subsequent queries. For example, if a batch of finished products is found to have low lactoferrin retention, it can trace back to the corresponding raw material batch of lactoferrin, its correction process at node J4 or J5, and the oxygen concentration at that time; if an abnormal reconstitution property is found in a certain batch, it can trace back to whether the OPO structured lipids and cream were executed according to the predetermined nodes, as well as the viscosity and temperature state of the nodes before homogenization.

[0296] 4. Uses of traceability data

[0297] The traceability chain established in step S9 has at least three uses in this invention:

[0298] (1) Quality traceability. When a batch of products shows abnormalities in nutritional indicators, activity retention, reconstitution properties, or oxidative stability, the source of the problem can be accurately located through traceability data. For example, is it an abnormality in the raw material batch, an abnormality in the operating conditions at a certain stage, or a failure to trigger or insufficient correction before the stage?

[0299] (2) Process optimization and model update. Since step S9 saves the mapping relationship between actual operating conditions, effective retention rate results and final product performance, it can be used to update the raw material retention rate function in the future. This involves adjusting weighting coefficients or optimizing weighting parameters in the objective function. In other words, tracing back to the database is not only about recording the past, but also about providing the data foundation for optimizing the future.

[0300] (3) Compliance and Audit Support. Infant formula milk powder is a highly regulated product, and companies need to prove that the entire process from raw materials to finished products is controllable. The binding storage structure established in step S9 can provide companies with a complete batch-level production evidence chain, including raw material sources, online testing, node decisions, correction logic, and final execution results.

[0301] 5. Output method

[0302] The final batching control instruction output in step S9 can be implemented in several ways: output to the execution control module to continue driving the execution of unfinished nodes; output to the MES or ERP system as part of the work order for this batch; output as a batch report for process engineers, quality control personnel, or auditors to view; or output as an electronic traceability record for the product traceability platform to access.

[0303] In some implementations, the system can also automatically generate a batch batch control summary, which includes at least: whether a node correction occurred in this batch; the types and nodes of raw materials that were corrected; whether the final critical nutrient structure prediction met the standard; the total compensation amount of high-value raw materials in this batch; and whether there were any fluctuations in operating conditions that exceeded the preset warning threshold.

[0304] V. Specific Application Examples

[0305] The experimental logic for the application examples in this section is as follows: First, establish a connection with... Figure 8 The corresponding pilot-scale / laboratory scale-up production line shown includes a wet main ingredient mixing unit, a homogenization unit, a concentration unit, a spray drying unit, a dry mixing unit, and pre- and post-packaging additive units. Then, samples of the same Stage 1 infant formula milk powder product are prepared by comparing the method of this invention with existing fixed feeding path mixing methods. Finally, the obtained samples are tested and evaluated in terms of nutritional structure accuracy, retention rate of key active ingredients, reconstitution performance, oxidative stability, and batch-to-batch variability.

[0306] (a) Experimental materials and equipment

[0307] 1. Main raw materials

[0308] The following raw materials are used as ingredients in this embodiment:

[0309] (1) Basic milk phase raw materials: skim emulsion, whole milk emulsion, whey liquid, lactose solution;

[0310] (2) Raw materials for fat structure regulation: light cream, OPO structured lipids, DHA / ARA microcapsule oils;

[0311] (3) Heat-sensitive active ingredients: lactoferrin, osteopontin, probiotics, and compound vitamin premix;

[0312] (4) Prebiotic raw materials: GOS and FOS;

[0313] (5) Mineral salt system: calcium lactate, potassium hydrogen phosphate, sodium citrate, etc.

[0314] 2. Main Equipment

[0315] The following laboratory / pilot-scale equipment was used in this embodiment:

[0316] (1) Online near-infrared analyzer for detecting protein, fat, lactose and total solids;

[0317] (2) Online temperature sensor, online oxygen concentration sensor and online viscometer;

[0318] (3) Small high-pressure homogenizer, first-stage homogenization pressure 18MPa, second-stage homogenization pressure 4MPa;

[0319] (4) Small-scale triple-effect vacuum evaporation device;

[0320] (5) Experimental pressure spray drying tower;

[0321] (6) Experimental three-dimensional mixer;

[0322] (7) Liquid metering pump, powder loss-in-weight scale and micro-addition device.

[0323] 3. Testing Items and Methods

[0324] (1) Protein, fat, and lactose: determined according to conventional milk powder testing methods;

[0325] (2) Whey protein / casein ratio: determined by protein composition analysis;

[0326] (3) - The percentage of palmitic acid at position 1 was determined by lipase hydrolysis combined with gas chromatography.

[0327] (4) Lactoferrin retention rate: quantified by ELISA or HPLC;

[0328] (5) Probiotic viable count: determined by plate count method;

[0329] (6) Reconstitution properties: Wetting time, dissolution time, and insolubility index were measured;

[0330] (7) Oxidative stability: evaluated using peroxide value and malondialdehyde value;

[0331] (8) Batch consistency: Statistical analysis of the standard deviation of key indicators for three batches of parallel samples.

[0332] (II) Experimental Group Design

[0333] To verify the technical effects of the present invention, the following test group was set up.

[0334] 1. Example 1: The intelligent batching method of the present invention

[0335] according to Figure 1 , Figure 2 and Figure 6 The logic shown establishes a set of target trophic structures, and then... Figure 3 The candidate feeding node set shown performs exposure calculation, effective retention rate calculation, allowed feeding node screening, piecewise inversion solution, and node pre-correction for each raw material.

[0336] 2. Comparative Example 1: Fixed formula table + fixed feeding node method

[0337] The existing conventional infant formula milk powder formulation method is adopted, that is, the theoretical amount of each raw material is directly calculated according to the target formula. All basic milk phase raw materials and fat structure adjustment raw materials are added at the wet process main ingredient node. Lactoferrin, osteopontin, probiotics and compound vitamins are added uniformly at the dry mixing node after spray drying, without node correction according to the real-time process status.

[0338] 3. Comparative Example 2: Fixed Compensation Method

[0339] Based on Comparative Example 1, an additional 15% empirical compensation was added to the thermosensitive active raw material, but the effective retention rate at different nodes was still not calculated, and no pre-node correction was performed.

[0340] 4. Comparative Example 3: Only add later, without piecewise inversion.

[0341] Lactoferrin, probiotics, and multivitamins are all added at the point before and after packaging, but exposure calculations and effective retention rate determinations are not performed. The basic milk phase raw materials and fat structure adjustment raw materials are still added in a fixed theoretical ratio.

[0342] (III) Setting of target formulation and target nutrient structure

[0343] This embodiment uses a stage 1 infant formula milk powder as the target sample, and its target nutrient structure set is set as follows:

[0344] 1. Target basic nutritional indicator: Target protein content =12.50%, target fat content =26.00%, target lactose content =55.80%, target ash content =2.60%, target energy density =510kcal / 100g.

[0345] 2. Target structural nutritional indicator: Target whey protein to casein mass ratio =60:40, target - Palmitic acid percentage =42.0%, target prebiotic ratio =GOS:FOS=9:1, Target DHA:ARA Ratio =1:1.

[0346] 3. Target process constraints

[0347] Lower limit of target activity retention Lactoferrin retention rate ≥85%, probiotic survival rate ≥80%;

[0348] Target impulse control index Wetting time ≤ 18s, insolubility index ≤ 0.8mL;

[0349] Target oxidation risk ceiling Peroxide value ≤ 0.18g / 100g (based on the final product after 30 days of storage).

[0350] (iv) Specific operation process of Example 1

[0351] 1. Real-time acquisition of raw material composition parameters

[0352] The main parameters of the basic emulsion raw material and the fat structure adjustment raw material were detected by an online near-infrared analyzer. The detection results are shown in Table 1.

[0353] Table 1 Real-time composition parameters of raw materials

[0354]

[0355] 2. Setting of raw material exposure sensitivity parameters

[0356] Based on preliminary experimental data and supplier information, a database of exposure-sensitive parameters for raw materials was established, and the results are shown in Table 2.

[0357] Table 2. Sensitive parameters for exposure to key raw materials

[0358]

[0359] 3. Setting of candidate feeding nodes

[0360] Candidate feeding nodes include: node J1, wet main batching node; node J2, pre-homogenization node; node J3, pre-concentration node; node J4, dry mixing node after spray drying; and node J5, addition node before and after packaging.

[0361] 4. Calculation of exposure and determination of effective retention rate

[0362] according to Figure 4 The logic shown integrates the thermal exposure, oxygen exposure, and shear exposure of each raw material under different candidate nodes to obtain the effective retention rate matrix of each raw material. The results are shown in Table 3. Figure 9 .

[0363] Table 3 Effective Retention Rate of Key Raw Materials at Different Candidate Feeding Nodes

[0364]

[0365] As shown in Table 3, although the heat exposure is small when OPO structured lipids are added at nodes J4 or J5, they cannot effectively participate in the formation of the front-end emulsion structure, resulting in poor overall process adaptability. Lactoferrin and probiotics have significantly higher effective retention rates at nodes J4 and J5 than at the front-end nodes.

[0366] 5. Allow filtering of material feeding nodes subsets.

[0367] The system is based on the effective retention rate in Table 3 and Figure 3 The node process functions shown are evaluated, and the screening results are as follows:

[0368] OPO structural grease: Allowed feed node subset = {J1, J2}

[0369] DHA / ARA microcapsule oils: Allowable feed node subset = {J2, J3, J4, J5}

[0370] Lactoferrin: Allowed feed node subset = {J4, J5}

[0371] Osteopontin: Allowed feed node subset = {J4, J5}

[0372] Probiotics: Allowed feed node subset = {J4, J5}

[0373] Multivitamins: Allowed feed node subset = {J3, J4, J5}.

[0374] 6. Solving by piecewise inversion

[0375] based on Figure 5 The segmented solution logic shown in the figure first solves for the basic emulsion raw materials, then for the fatty acid structure-modifying raw materials, and finally for the thermosensitive active raw materials. The final target addition amount and target feeding node results are shown in Table 4.

[0376] Table 4 Target Addition Amount and Target Feeding Node for Example 1

[0377]

[0378] 7. Pre-node correction

[0379] In Example 1, when the system reaches node J4, the online oxygen concentration detection value... The increase of 0.8 individual integral percentage points compared to the initial solution resulted in a decrease in the effective retention rate of lactoferrin at node J4 from 0.87 to 0.82. (System data...) Figure 6 The process shown re-solved the problems for lactoferrin and multivitamins, and finally corrected the target feeding node for lactoferrin to J5, and adjusted the target addition amount from 1.72kg to 1.61kg; the multivitamin remained unchanged at J4, with only minor adjustments to the addition amount.

[0380] (v) Analysis of product testing results and technical effects

[0381] 1. Precision of final product nutritional structure

[0382] The samples prepared in Example 1 and each comparative example were tested, and the results are shown in Table 5.

[0383] Table 5 Comparison of Accuracy in Achieving Nutritional Structure

[0384]

[0385] As can be seen from Table 5 and Figure 10, Example 1 is closest to the target values ​​in both basal and structural nutrient indicators, especially... - The deviations in palmitic acid percentage, whey / casein ratio, and DHA:ARA ratio were significantly smaller than those in the respective pairs. This indicates that the present invention, by allocating OPO structured lipids and cream to appropriate nodes before homogenization and performing node-based allocation and retention compensation on DHA / ARA microcapsule oils, can more effectively achieve the target structural nutritional design.

[0386] 2. Retention rate of active components

[0387] The retention of lactoferrin, osteopontin, and probiotics in the final product was tested, and the results are shown in Table 6.

[0388] Table 6 Comparison of Retention Rates of Active Components

[0389]

[0390] From Table 6, Figure 11 It is evident that Example 1 demonstrates superior retention of lactoferrin, osteopontin, and probiotics compared to the comparative examples, particularly significantly better than Comparative Examples 1 and 2. Although Comparative Example 2 introduced a fixed compensation amount, its retention rate remained lower than that of Example 1, indicating that simply increasing the fixed addition amount cannot replace dynamic allocation and pre-node correction based on the effective retention rates of different candidate nodes. While Comparative Example 3 added all heat-sensitive components later, resulting in some indicators approaching those of Example 1, its basic milk phase and fat structure still relied on a fixed theoretical feed, which is detrimental to the overall nutritional structure and comprehensive process adaptation; therefore, its overall effect was still inferior to Example 1.

[0391] 3. Reconstitution and Powder Properties

[0392] The wetting time, dissolution time, and insolubility index of the samples were measured, and the results are shown in Table 7. Figure 12 .

[0393] Table 7 Comparison of reconstitution properties and powder performance

[0394]

[0395] Example 1 exhibits superior wettability and solubility, with a lower insolubility index. This result demonstrates that the present invention, in selecting raw material nodes, not only considers activity retention but also emulsion formation and front-end system stability, resulting in a more rational powder particle structure and more uniform surface properties, thus exhibiting better reconstitution performance.

[0396] 4. Oxidative stability

[0397] The oxidation index of the samples was detected after 30 days of storage following sample preparation. The results are shown in Table 8. Figure 12 .

[0398] Table 8 Comparison of Oxidation Stability

[0399]

[0400] Example 1 showed the lowest oxidation index, indicating that by implementing more reasonable node selection and exposure control for DHA / ARA microcapsule oils and active ingredients, the oxidation risk of the final product can be reduced.

[0401] 5. Batch consistency

[0402] Three parallel replicates were performed on Example 1 and Comparative Example 1 respectively, and the standard deviation of key indicators was detected. The results are shown in Table 9.

[0403] Table 9 Batch Consistency Comparison

[0404]

[0405] Table 9 shows that the present invention can significantly reduce the fluctuation of key indicators between different batches. This is because the present invention does not simply rely on a fixed formula table, but makes the batching process more adaptable and self-correcting through real-time parameter acquisition, effective retention rate calculation, and pre-node correction.

[0406] (vi) Conclusion on Technical Effectiveness

[0407] Combination Figures 4-12 As can be seen from the experimental results in Tables 5-9, the present invention has at least the following technical effects:

[0408] 1. Higher precision in achieving nutritional structure: This invention, through piecewise inversion solution and optimization of candidate feeding nodes, ensures that the final product achieves higher precision in basic nutritional indicators such as protein, fat, and lactose, as well as whey / casein ratio, etc. - The structural nutritional indicators, such as the proportion of palmitic acid and the DHA:ARA ratio, are closer to the design target.

[0409] 2. Higher retention rate of thermosensitive active ingredients: By modeling the exposure integral and effective retention rate of thermosensitive active raw materials such as lactoferrin, osteopontin, and probiotics, and performing corrections before the node, this invention can significantly improve the final product retention rate of active ingredients.

[0410] 3. Superior mixing performance and powder quality: Because the fat structure regulating raw materials are distributed to more suitable front-end nodes to participate in emulsion formation and homogenization processes, the samples of this invention exhibit better wettability, solubility and lower insolubility index.

[0411] 4. Lower oxidation risk and better batch consistency: This invention can reduce unnecessary oxygen exposure to highly unsaturated fats and active components, lower oxidation index, and at the same time control batch-to-batch differences through pre-node correction, thereby improving production stability.

[0412] The foregoing description of embodiments of the present invention, through which those skilled in the art are able to implement or use the present invention, will be readily apparent to those skilled in the art. Various modifications to these embodiments will be readily apparent to those skilled in the art. The general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novelty disclosed herein.

[0413] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0414] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0415] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0416] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0417] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0418] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0419] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

Claims

1. A method for intelligent ingredient formulation of infant formula milk powder based on target nutrient structure inversion, characterized in that, Includes the following steps: S1. Establish a target nutrient structure set and establish a candidate feeding node set corresponding to the target nutrient structure set; S2. Obtain the real-time composition parameters and exposure sensitivity parameters of each candidate raw material, and divide each candidate raw material into the basic emulsion raw material group, the fat structure regulation raw material group, and the thermosensitive active raw material group. S3. For any candidate raw material, calculate the cumulative heat exposure, cumulative oxygen exposure and cumulative shear exposure of the raw material at each candidate feeding node based on the temperature trajectory, oxygen concentration trajectory and shear trajectory corresponding to each candidate feeding node on the production line. S4. Based on the cumulative heat exposure, cumulative oxygen exposure, and cumulative shear exposure, calculate the effective retention rate of each raw material at each candidate feeding node; S5. Based on the effective retention rate, the raw material group to which each raw material belongs, the process adaptation constraints of each candidate feeding node, and the target nutrient structure set, screen the allowable feeding node subset for each raw material. S6. Within the subset of allowed feeding nodes, establish a piecewise inversion solution model to solve for the target addition amount and target feeding nodes for the basic emulsion raw material group, the fat structure regulating raw material group, and the thermosensitive active raw material group, respectively. S7. Based on the solution results of the piecewise inversion solution model, control the liquid metering device, the powder metering device, and the feeding execution device corresponding to the target feeding node to feed materials.

2. The intelligent ingredient dispensing method according to claim 1, characterized in that, In step S1, the target nutrient structure set includes at least: a target basic nutrient index set, a target structural nutrient index set, and a target process constraint index set; The target set of basic nutritional indicators includes at least the target protein content, target fat content, target lactose content, and target ash content; The set of target structural nutritional indicators includes at least one or more of the following: target whey protein to casein mass ratio, target palmitic acid percentage at sn-2 position, target prebiotic ratio, and target long-chain polyunsaturated fatty acid ratio. The set of target process constraint indicators includes at least one or more of the following: target osmotic pressure range, target upper limit of oxidation risk, target remedial index, and target lower limit of activity retention rate. The candidate feeding node set includes at least a wet main feeding node, a node before homogenization, a node before concentration, a dry mixing node after spray drying, and a node before and after packaging.

3. The intelligent ingredient dispensing method according to claim 1, characterized in that, In step S2, the real-time composition parameters of each raw material include at least the protein content, fat content, lactose content, ash content, and structural characteristic parameters corresponding to the nutritional indicators of the target structure; the exposure sensitivity parameters of each raw material include at least the thermosensitive critical temperature, the oxygen-sensitive critical oxygen concentration, and the shear sensitivity threshold. And / or, in step S2, the basic milk phase raw material group includes at least one or more of skim milk, whole milk, whey, lactose solution and mineral salt solution; And / or, in step S2, the fat structure adjustment raw material group includes at least one or more of light cream, vegetable oil, structured lipids and long-chain polyunsaturated fatty acid oils; And / or, in step S2, the thermosensitive active raw material group includes at least one or more of lactoferrin, osteopontin, probiotics, vitamin active substances and microcapsule functional oils; And / or, in step S2, the real-time composition parameters of each raw material are obtained by one or more of an online near-infrared spectrometer, an online fat analyzer, an online protein analyzer, an online densitometer, and an online viscometer; wherein, the online near-infrared spectrometer is used to output estimated values ​​of protein content, fat content, lactose content, and ash content.

4. The intelligent ingredient dispensing method according to claim 1, characterized in that, In step S3, the cumulative amount of heat exposure Calculate as follows: ； in, and They represent the first The first type of raw material in The times when each candidate feeding node enters and leaves the subsequent process stage; Indicates the first After each candidate feeding node, the material temperature along the process path changes over time. The change value; Indicates the first Thermosensitive critical temperature of the raw material; And / or, in step S3, the cumulative oxygen exposure Calculate as follows: ； in, Indicates the first After each candidate feeding node, the oxygen concentration in the process path changes over time. The change value; Indicates the first The oxygen-sensitive critical oxygen concentration of the raw material; And / or, in step S3, the cumulative amount of shear exposure Calculate as follows: ； in, Indicates the first The equivalent shear strength along the process path after each candidate feeding node varies with time. The change value; Indicates the first The shear sensitivity threshold of a certain raw material.

5. The intelligent ingredient dispensing method according to claim 1, characterized in that, In step S4, the effective retention rate is determined as follows: the cumulative heat exposure, cumulative oxygen exposure, and cumulative shear exposure of each raw material at each candidate feeding node are input into the retention rate function pre-established for the corresponding raw material to obtain the effective retention ratio of the raw material after subsequent processing at the candidate feeding node; wherein, the retention rate function is a functional relationship, lookup table, or piecewise function obtained by fitting the active retention experimental data of the raw material under preset temperature, oxygen concentration, and shear conditions; And / or, in step S5, the raw materials in the thermosensitive active raw material group are screened only at candidate feeding nodes where their effective retention rate is not lower than the preset active retention threshold; the raw materials in the fatty acid structure regulating raw material group are screened only at candidate feeding nodes where the target sn-2 position palmitic acid ratio and emulsion dispersion stability requirements are met. And / or, in step S5, the process adaptation constraints include at least one or more of the following constraints: solubility constraints, dispersion stability constraints, pre-homogenization emulsion formation constraints, spray drying heat exposure constraints, and dry mixing uniformity constraints; wherein, when the raw materials of the fat structure adjustment raw material group meet the pre-homogenization emulsion formation constraints, they are preferentially allocated to the wet process main batching node or the pre-homogenization node.

6. The intelligent ingredient dispensing method according to claim 1, characterized in that, In step S6, the piecewise inversion solution model must simultaneously satisfy at least the following conditions: a) The set of target basic nutritional indicators for the final product meets the deviation limit requirements; b) The set of target structural nutritional indicators for the final product meets the deviation limit requirements; c) The set of target process constraints for the final product is not exceeded; d) Each raw material can only be assigned to one or more candidate feeding nodes in its subset of allowed feeding nodes; e) The solution results of the thermosensitive active raw material group, after being compensated and corrected with the corresponding effective retention rate, meet the target value requirements of the final product. The piecewise inversion solution model uses the weighted result of the deviation of the basic nutritional index of the final product, the deviation of the structural nutritional index of the final product, the process constraint penalty, the insufficient activity retention penalty, and the raw material cost as the optimization objective, and takes the minimum of the weighted result as the solution objective.

7. The intelligent ingredient dispensing method according to claim 6, characterized in that, In step S6, the basic nutritional index deviation term is calculated as follows: the absolute deviation between the predicted value and the target value of each basic nutritional index involved in the calculation is multiplied by the corresponding weight coefficient, and the results are summed to obtain the basic nutritional index deviation term. And / or, in step S6, for the target raw material in the thermosensitive active raw material group, when it is assigned to the target candidate feeding node, its target feeding amount is compensated and corrected according to the target demand amount of the final product and the effective retention rate. The compensation and correction satisfy that: the result of multiplying the target feeding amount by the effective retention rate is not lower than the target demand amount of the corresponding effective component of the raw material in the final product.

8. The intelligent ingredient dispensing method according to claim 1, characterized in that, In step S7, the current process parameters and online detection parameters are obtained before each candidate feeding node; and the method also includes steps S8 and S9; wherein, Step S8: Before any candidate feeding node that has not yet been executed, based on the latest online detection parameters and current process parameters, recalculate the effective retention rate of the corresponding unfeeded raw materials under the remaining candidate feeding nodes, and determine whether the predicted nutritional structure of the final product still meets the requirements of the target nutritional structure set. If the target is not met, steps S5 and S6 are re-executed only for raw materials for which the feeding node has not yet been executed, in order to correct their target addition amount and / or target feeding node. Step S9: Output the final batching control command and bind and store the raw material batch, candidate feeding node selection results, effective retention rate, segmented inversion solution results, online detection parameters, and correction records.

9. The intelligent ingredient dispensing method according to claim 8, characterized in that, In step S7, the online detection parameters include at least online protein detection values, online fat detection values, online lactose detection values, online temperature detection values, online oxygen concentration detection values, and online viscosity detection values; In step S8, steps S5 and S6 are re-executed for raw materials that have not yet been fed at a feeding node only if the number of remaining candidate feeding nodes is not less than the preset minimum number of correction nodes. In step S8, the corrected target feeding node is only allowed to switch within the subset of allowed feeding nodes for the corresponding raw material, and cannot switch to a process node before a candidate feeding node that has already been executed.

10. An intelligent ingredient mixing system for infant formula milk powder based on target nutrient structure inversion, characterized in that, The system is used to implement the method according to any one of claims 1-9, comprising: The target trophic structure modeling module is used to establish a set of target trophic structures and a set of candidate feeding nodes; The raw material parameter acquisition module is used to acquire the real-time composition parameters and exposure sensitivity parameters of each candidate raw material, and to classify them into basic emulsion raw material group, lipid structure regulation raw material group and thermosensitive active raw material group. The exposure calculation module is used to calculate the cumulative heat exposure, cumulative oxygen exposure, and cumulative shear exposure of each raw material at each candidate feeding node; The retention rate calculation module is used to calculate the effective retention rate based on the cumulative heat exposure, cumulative oxygen exposure, and cumulative shear exposure. The node filtering module is used to filter the subset of allowed feeding nodes for each raw material; The piecewise inversion solution module is used to solve for the target addition amount and target feeding node of each raw material; The execution control module is used to control the liquid metering device, powder metering device, and various feeding execution devices to complete the feeding process.

Images