Method and system for optimizing complementary amino acid ratio of bone bean milk double protein

Abstract

The present application relates to the field of electric digital data processing, and particularly relates to a bone bean milk double protein amino acid complementary matching optimization method and system, the method comprising: constructing a lower envelope potency curve changing with bone protein proportion, determining a transition interval and an active limiting amino acid set according to limiting amino acid identification replacement and proximity conditions, and taking a local high value area as a candidate area when there is no transition; based on raw material protein content, equipment step distance and total quantity constraints, discretizing the candidate area and reverse checking node potency, obtaining effective nodes through closed verification; constructing a node potency confidence interval according to analysis uncertainty, and executing cross dominance to exclude dominated nodes; for nodes that cannot be uniquely dominated, rejudging by supplementing samples of the active limiting amino acid set, and outputting according to a conservative strategy when still unable to make a decision. The present application ensures stable feeding by tracking amino acid replacement, mapping equipment nodes, combining anti-disturbance cross dominance and limited feedback error suppression.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing. In particular, it relates to a method and system for optimizing the complementary amino acid ratio of bone marrow and soy milk dual proteins. Background Technology

[0002] The bone-soy-milk dual-protein amino acid complementary ratio is a method of maximizing the nutritional value of the mixed protein by scientifically adjusting the ratio of bone-derived and soy-derived proteins and utilizing the natural complementary advantages of their amino acid profiles to eliminate the "limiting amino acid" deficiencies of each. Since changes in the mixing ratio dynamically alter the types of deficient amino acids, the core of this ratio optimization lies in precisely capturing the optimal solution for nutritional value.

[0003] Existing methods for optimizing the ratio of bone bean protein are difficult to produce stable, accurate, and executable formulations during the conversion from theoretical calculations to industrial feeding. Specific shortcomings include: the use of a static evaluation model in the ratio optimization stage fails to dynamically track the changing trajectory of limiting amino acids during the continuous change of bone protein proportion, easily leading to misjudgments of potency and deviations from the theoretically optimal ratio in transition boundary regions; the lack of sufficient consideration of raw material purity differences, equipment metering step size, and total feeding capacity constraints in the engineering conversion stage, making direct conversion prone to deviations in actual protein proportion and non-closure of total loading; and the scalar evaluation of potency in the decision-making stage is highly susceptible to interference from inherent random errors in raw material testing, causing oscillations in feeding decisions between adjacent grids. Summary of the Invention

[0004] To address the limitations of existing bone marrow dual protein ratio optimization methods, such as static evaluation failing to track restrictive amino acid changes, engineering transformation failing to consider physical constraints leading to non-closed-loop conversions, and single scalars being easily affected by detection errors, making it difficult to output stable and executable industrial formulations, this invention provides solutions in the following aspects.

[0005] In the first aspect, the method for optimizing the amino acid complementarity ratio of bone and soy milk dual proteins includes: obtaining a set range of bone protein mass percentage; calculating the relative abundance score of essential amino acids under each percentage in the range of bone protein mass percentage based on the amino acid vectors and reference patterns of bone and soy proteins; and constructing a lower envelope valence curve that reflects the continuous change of the complementary shortcomings of dual proteins with the ratio. The replacement of limiting amino acid markers is analyzed based on the lower envelope valence curve to determine the transition boundary region. The limiting amino acids before and after the replacement, as well as the non-limiting amino acids that meet the approximation conditions, are recorded as the active limiting amino acid set. If no transition occurs, the local high value region is used as the candidate region. Within the candidate area, the theoretical ratio is discretized into actual physical feeding nodes by combining the purity of raw material protein, equipment metering step distance and total amount of single batch feeding, and the lower envelope effect value corresponding to each actual physical feeding node is calculated in reverse. Valid nodes that meet the constraints are screened out through closure verification. The raw material analysis uncertainty of the amino acid detection error is obtained, and the lower envelope effect value of each effective node is converted into a confidence interval. By comparing the upper and lower limits of the confidence intervals of each effective node, cross-domination judgment is performed to screen out dominated nodes whose lower confidence limit is lower than the upper confidence limit of other nodes. When there are multiple valid nodes and the optimal node cannot be uniquely determined by cross-dominance determination, supplementary sampling and testing are performed on the detection items corresponding to the active restriction amino acid set to update the confidence interval of the valid node and perform cross-dominance determination again; if a unique decision still cannot be made, the conservative strategy is adopted for output.

[0006] Preferably, the construction of the lower envelope potency curve that continuously varies with the proportion of bone protein includes: Based on the amino acid vectors of bone and soybean proteins and the reference pattern, the relative abundance score of each essential amino acid under each bone protein mass percentage is calculated. The relative abundance score is the ratio of the mixed amino acid content to the reference abundance benchmark value. The minimum relative abundance score for each of the bone protein mass percentages is taken as the lower envelope potency value to form a lower envelope potency curve that changes continuously with the bone protein mass percentage.

[0007] Preferably, the active limiting amino acid set includes: During the continuous change of the bone protein mass percentage, the limiting amino acid markers that form the lower envelope effect value are recorded for each percentage. When the restricted amino acid identifier corresponding to adjacent percentage points changes, the interval where the change occurs is located as the restricted amino acid transition boundary region, and the restricted amino acid identifiers before and after the change are included in the active restricted amino acid set; For any bone protein mass percentage, if the difference between the relative abundance score of an essential amino acid that is not identified as a restricted amino acid and the lower envelope effect value meets a preset relative proximity tolerance boundary, then the amino acid is identified as a candidate restricted amino acid and incorporated into the active restricted amino acid set.

[0008] Preferably, the determination criteria for the preset relative proximity tolerance boundary include: If the difference between the relative abundance score of an essential amino acid and the lower envelope value does not exceed the product of the relative proximity tolerance coefficient and the larger of the lower envelope value and the preset positive lower limit, then the amino acid is identified as a potential limiting amino acid that satisfies the relative proximity tolerance boundary.

[0009] Preferably, the confirmation conditions for the transition boundary region further include: If the difference between the relative abundance score of a potential limiting amino acid and the lower envelope value consistently satisfies the relative proximity tolerance boundary, and the relative abundance score difference between potential limiting amino acids shows a sign change or slope convergence, then the neighborhood is confirmed to belong to the transition boundary region, and the relevant amino acid is incorporated into the active limiting amino acid set.

[0010] Preferably, the method for determining the candidate region includes: Identify the local high-value region on the lower envelope titer curve. The local high-value region is the region where the lower envelope titer reaches a local peak and the limiting amino acid identifier of the lower envelope titer is stable and traceable within the region. The neighborhood around the ratio region is expanded according to the step size to form the candidate region, and the batch is marked as a batch without clearly defined amino acid replacement. The active restricted amino acid set is the set of restricted amino acid identifiers that form the lower envelope value within the candidate region.

[0011] Preferably, the reverse calculation of the lower envelope effect value corresponding to each actual physical feeding node includes: Obtain the protein content coefficients of bone phase raw materials and soybean phase raw materials, as well as the equivalent mass step distance of bone phase equipment, the equivalent mass step distance of soybean phase equipment, and the total feed amount constraint threshold for a single batch. Based on the theoretical bone protein content ratio in the candidate area and the total mass of the target protein in the current batch, the theoretical target bone protein content and the theoretical target soybean protein content are calculated, and the theoretical target protein content is converted into the theoretical physical feed mass using the bone phase raw material protein content coefficient and the soybean phase raw material protein content coefficient. Based on the equivalent mass step distance of the bone phase equipment and the equivalent mass step distance of the bean phase equipment, the theoretical physical feeding mass is rounded to obtain the number of steps of the bone phase equipment and the number of steps of the bean phase equipment, and the corresponding actual physical feeding mass is determined. The actual bone protein content and actual soybean protein content are calculated in reverse based on the actual physical feed mass and the protein content coefficient, and the proportion of actual bone protein content is calculated; the corresponding lower envelope potency is recalculated based on the proportion of actual bone protein content.

[0012] Preferably, the step of filtering out valid nodes that satisfy the constraints through closure verification includes: The difference between the sum of the actual physical feed mass and the single batch feed total quantity constraint threshold is calculated to obtain the total loading deviation. The difference between the actual bone protein mass percentage and the theoretical bone protein mass percentage is calculated to obtain the ratio mapping deviation; A node is defined as a valid node if the total loading deviation does not exceed the preset total closure tolerance, the ratio mapping deviation does not exceed the preset ratio mapping tolerance, and the limiting amino acid corresponding to the lower envelope effect value recalculated based on the actual bone protein content belongs to the set of active limiting amino acids.

[0013] Preferably, the process of performing cross-dominance determination includes: For each of the aforementioned valid nodes, the deviation is obtained by multiplying the lower envelope effect value by the raw material analysis uncertainty. The lower envelope effect value minus the deviation is used as the confidence lowerfimum, and the deviation is added to it as the confidence upperfimum. The valid nodes are compared pairwise. When the confidence lower bound of one node is greater than the confidence upper bound of another node, it is determined that the former node dominates the latter node, and the dominated node is eliminated.

[0014] Secondly, the bone marrow and soy milk dual protein amino acid complementary ratio optimization system includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned bone marrow and soy milk dual protein amino acid complementary ratio optimization method is implemented.

[0015] The present invention has the following effects: 1. This invention, by constructing a lower envelope potency curve that continuously changes with the proportion of bone protein, can accurately capture the replacement trajectory of limiting amino acids in the transition boundary region. By including the limiting amino acids before and after the replacement, as well as the unrestricted amino acids that meet the approximation conditions, into the active set, this invention effectively solves the problem of potency misjudgment caused by static assumptions in the transition boundary region, ensuring the accuracy and continuity of the theoretical optimal ratio calculation.

[0016] 2. By performing step-rounding and reverse calculation on the theoretical ratio and executing strict closure verification, this invention selects effective nodes that meet both equipment accuracy and total feed constraints, effectively solving the problems of actual protein ratio deviation and total loading non-closure caused by direct conversion, and ensuring the accurate implementation of the formula in industrial production.

[0017] 3. This invention constructs confidence intervals by introducing uncertainty in raw material analysis and uses cross-dominance decision logic to filter out dominated nodes. This effectively avoids the risk of oscillations in feeding decisions between adjacent grids caused by inherent random errors in raw material detection. Combined with a supplementary sampling and testing mechanism for active limiting amino acids, robust decision-making is further achieved in data ambiguity areas, thereby outputting a stable, reliable, and interference-resistant optimal feeding scheme. Attached Figure Description

[0018] Figure 1This is a flowchart of steps S1-S5 in the method for optimizing the amino acid complementary ratio of bone marrow and soy milk dual proteins in an embodiment of the present invention.

[0019] Figure 2 This is a structural block diagram of the bone marrow and soy milk dual protein amino acid complementary ratio optimization system according to an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0021] Reference Figure 1 The method for optimizing the amino acid complementarity ratio of bone marrow and soy milk double protein includes steps S1-S5, as follows: S1: Obtain the set range of bone protein mass percentage, calculate the relative abundance score of essential amino acids under each percentage in the range of bone protein mass percentage based on the amino acid vectors of bone and soybean protein and the reference pattern, and construct the lower envelope valence curve that reflects the continuous change of the complementary shortcomings of the two proteins with the ratio.

[0022] Obtain bone-derived protein amino acid vector and the amino acid vector of soy protein , of which element and These represent the first and second components of bone-derived protein and soy-derived protein, respectively. The content of each essential amino acid is expressed in milligrams per gram of protein; in this embodiment, the reference model adopted is the FAO / WHO 2013 revised essential amino acid reference model for children aged 3-10 years. All content vectors have been normalized to per gram of protein.

[0023] Set the bone protein content as follows: Defined as the ratio of bone protein mass to total mixed protein mass, with a value range of [value missing]. Dimensionless. Set the scan step size. This is an empirical value (e.g., 0.0005) used to control the accuracy of the mixing ratio. within the interval Perform equidistant sampling to obtain a series of sampling points. ( For any given percentage of bone protein mass... The first in the mixed protein The expected content of the essential amino acids was obtained by weighted summation of contributions from bone and soybean sources.

[0024] Among them, scanning step size Calibration was performed based on the relative standard deviation (RSD) of amino acid detection using liquid chromatography: First, 20 consecutive amino acid analyses were conducted on the same sample, and the RSD was calculated. Then, a precision threshold for the formulation was calculated by combining this with the maximum permissible error of ≤0.5% for nutritional potency. Finally, a baseline step size was defined based on the RSD interval. High-precision chromatography met the relative standard deviation requirement. Time calibration base step size The relative standard deviation of conventional testing equipment is within Interval time calibration Relative standard deviation Low-precision testing equipment is calibrated If the scan step size If the sampling step size is too small, the number of sampling points explodes, the computational load increases exponentially, the system lags, recipe output is delayed, and industrial real-time control becomes unusable; if the scanning step size is too small... If the sampling interval is too large, the discrete sampling interval will be too wide, and the amino acid transition boundary will easily fall between two sampling points, resulting in missed transitions, distortion of the lower envelope potency curve, and complete deviation of the optimal ratio.

[0025] Specifically, the relative abundance scores satisfy the following relationship: ; In the formula, This indicates the current percentage of bone protein. Next The relative abundance scores of essential amino acids, This represents the vector element representing the content of the i-th essential amino acid in bone-derived protein. This indicates the percentage of bone protein, that is, the ratio of bone protein to the total mixed protein, and its value ranges from [value missing]. , Indicates the first of the soybean protein The vector elements representing the content of essential amino acids. Represents the first in the reference mode vector Reference abundance benchmarks for essential amino acids. This indicates the current percentage of bone protein. Below, the proportionate contributions of bone and soy sources. The linear mixture content of amino acids. This reflects the actual weighted average concentration of a specific amino acid in the two-protein mixture system. When... When the amino acid content is exactly the same as the reference standard, it means that the amino acid content is exactly the same as the reference standard; when When this occurs, it indicates a relative deficiency of that amino acid, which may become a limiting amino acid; when... When the time is right, it indicates that the amino acid is relatively abundant.

[0026] The reference pattern vector is an ordered set of values ​​representing the ideal intake ratios of various essential amino acids for the target population. Each dimension within the vector corresponds to the standard reference proportion of each essential amino acid. The reference pattern vector can be flexibly selected based on the target population and nutritional standards. It can adopt the essential amino acid reference patterns for different age groups of children and adults published by FAO / WHO, or the amino acid balance standards stipulated by national standards and dietary guidelines. Different reference pattern vectors represent differentiated ideal amino acid ratio benchmarks, and the weight comparison relationship between various essential amino acids changes accordingly. If the reference pattern vector is changed, the benchmark for calculating the relative abundance of amino acids changes simultaneously, and the results of limiting amino acid determination and potency ranking will also change. If the original judgment thresholds are continued to be used, problems such as incorrect identification of limiting amino acids and missed detection of transition boundaries will occur. Therefore, the judgment thresholds related to potency assessment and transition boundary identification should be recalibrated accordingly. Threshold calibration follows a unified principle: keeping the interpolation error tolerance threshold, sampling base step size, and other underlying operational parameters unchanged, and refitting and determining the transition identification-related judgment thresholds suitable for this nutritional benchmark based on the amino acid score distribution range under the new reference pattern vector, ensuring that the sensitivity of limiting amino acid transition boundary identification remains stable.

[0027] According to the barrel effect principle, the overall nutritional complementarity value of a mixed protein is limited by the amino acid with the lowest relative abundance score. The reference pattern vector contains several essential amino acids corresponding to the target population. Therefore, under the same bone protein ratio, the minimum relative abundance score of all essential amino acids included in the evaluation range is extracted as the lower envelope value corresponding to that ratio.

[0028] Traverse all sampling points Calculate the corresponding lower envelope effect value. All coordinate points according to By connecting the components in ascending order, a lower envelope potency curve can be constructed, with the proportion of bone protein mass as the x-axis and the lower envelope potency value as the y-axis. This curve fully reflects the lower limit of the complementary potency of the dual-protein mixture system under different ratios, determined by the most scarce essential amino acid, and its overall variation with the proportion of bone protein.

[0029] This application uniformly adopts a piecewise linear interpolation method to connect discrete sampling points. A continuous lower envelope valence curve is formed. Piecewise linear interpolation is chosen because the relative abundance of essential amino acids in bone and bean mixed proteins changes approximately linearly with the ratio. This algorithm has low computational cost and is suitable for real-time industrial computing scenarios. To avoid the defect of missed detection caused by discrete sampling intervals where transition boundaries fall between two points, an intersection point supplementary sampling mechanism is implemented: for each interpolation segment, the intersection point of the relative abundance scores of the two restricting amino acids at both ends of the segment is solved. If the proportion of bone protein corresponding to the intersection point falls within the interval of the segment, the intersection point is added as a new encrypted sampling point to the sampling sequence, and the restricting amino acid identifier at the intersection point is recalculated to completely identify hidden transition boundaries. At the same time, an interpolation error evaluation mechanism is set: the absolute difference between the valence value of the encrypted fine sampling point and the interpolation curve is the interpolation error, and the preset interpolation error tolerance threshold is 0.003; if the average interpolation error of a single interpolation interval exceeds this threshold, the ratio scanning step size is automatically reduced and the sampling sequence is regenerated until the interpolation error meets the tolerance requirements.

[0030] The interpolation error tolerance threshold is determined based on the following: the allowable deviation of 0.3% in the nutritional value of mixed proteins is converted to a baseline threshold of 0.003; the reasonable range is 0.001 to 0.005; setting the threshold too small will frequently trigger encrypted sampling and significantly increase the computation time, while setting the threshold too large will lead to distortion of the interpolation curve and an increased probability of missed detection at transition boundaries.

[0031] This invention clearly defines the construction of valence curves using piecewise linear interpolation, and adds interpolation intersection supplementary sampling and interpolation error verification mechanisms to solve the problem of missed detection of transition boundaries caused by discrete sampling intervals, ensuring that all restrictive amino acid replacement intervals are completely identified, and further improving the accuracy of ratio optimization.

[0032] It should be noted that all the calculation formulas for the relative abundance scores of essential amino acids and the construction of the lower envelope potency curve are based on a pure protein mixture theory model, where... The ratio of bone protein to total pure protein is used only for evaluating amino acid complementarity at the nutritional level, corresponding to the upper nutritional calculation layer. The bone and bean-based raw materials used in production contain non-protein components such as moisture, carbohydrates, and impurities. These components do not contain essential amino acids and are not included in the amino acid abundance weighting calculation; therefore, impurity-related parameters do not need to be introduced during the nutritional calculation stage. This application sets up a lower industrial feeding layer to convert the pure protein ratio to the crude raw material feeding quality. The two-layer system establishes a bidirectional conversion relationship based on the raw material protein content coefficient. The complete forward and reverse mathematical conversion derivation is detailed in step S3 below.

[0033] S2: Analyze the replacement of limiting amino acid markers based on the lower envelope valence curve, determine the transition boundary region, and record the limiting amino acids before and after the replacement and the non-limiting amino acids that meet the approximation conditions as the active limiting amino acid set. If no transition occurs, the local high value region is taken as the candidate region.

[0034] By dynamically tracking the replacement trajectories of limiting amino acids, key sensitive regions (i.e., transition boundary regions) or stable high-value regions (i.e., candidate regions) in the formulation optimization process are precisely located, and an active limiting amino acid set is constructed to provide a precise target range for subsequent engineered feeding. Specifically, this includes the following sub-steps: After calculating each sampling point Corresponding lower envelope effect value Simultaneously, the system records the essential amino acid species sequence number that actually achieves the lowest global relative abundance score at that sampling point, and defines it as the limiting amino acid identifier for that point, denoted as... .

[0035] Due to the proportion of bone protein It is continuously changing; the system applies each pair of adjacent sampling points. and Comparative monitoring will be conducted. when This indicates that with slight changes in the ratio, the type of amino acid that dominates the nutritional value of the system has shifted. (The interval is missing from the original text.) The region is located as a boundary region restricting amino acid transitions. At this point, the system identifies the restricting amino acids before and after the change. and Included in the active limiting amino acid set This is to ensure that subsequent optimizations can cover this critical transition zone.

[0036] At routine sampling points where non-restricted amino acid labels change, focusing solely on a single minimum might overlook other amino acids that are already approaching the lower envelope boundary. Therefore, a relative proximity tolerance boundary mechanism is introduced: For each sampling point The following is not the current limiting amino acid Any essential amino acid Calculate the relative abundance score Value of lower envelope effect The difference Introduce a preset relative proximity tolerance coefficient Sum of positive lower bound boundary Determine whether an amino acid is a candidate limiting amino acid using the following formula: ; In the formula, Indicates the first At the sampling point, the first The difference between the relative abundance score of each unrestricted amino acid and its lower envelope value. This represents the relative proximity tolerance coefficient, with an empirical value of 0.05 (dimensionless), used to control the width of the tolerance range; Indicates the first The lower envelope value at each sampling point, i.e., the global minimum of the relative abundance score of all essential amino acids under the current ratio; An empirical value of 0.1 is taken as the preset positive lower limit parameter, and its physical effect is... When the numerical value is extremely small, the ratio is avoided from being distorted, thus ensuring the robustness of the judgment.

[0037] The relative proximity tolerance coefficient is used to screen candidate limiting amino acids that are adjacent to the shortest protein. It is calibrated according to the allowable deviation of 5% nutritional potency of FAO / WHO mixed protein, and the reasonable range is 0.02 to 0.08. It can be scaled and adjusted according to the nutritional standards of the applicable population. If the coefficient is too small, it will filter out key adjacent amino acids, incompletely identify the transition, and increase the misjudgment rate of potency. If the coefficient is too large, it will cause redundancy of the active limiting amino acid set, double the detection cost and calculation time, and distort the determination range of the transition interval.

[0038] Furthermore, the positive lower limit parameter To avoid logic failure in determining low potency ranges, the potency distribution is calibrated based on extreme ratios of bone marrow and bean protein, with a reasonable range of 0.05–0.2. This value can be adaptively adjusted according to the mixing ratio range and the degree of product fortification. Too small a value can lead to missed screening of candidate amino acids and failure of transition recognition under extremely low valence conditions. If the value is too large, it will raise the threshold for determining the intermediate valence range and cause the loss of amino acids that are sensitive to transitions.

[0039] When the above inequality holds, amino acids are identified. Amino acids that are sufficiently close to the lower envelope at the current ratio point are identified as candidate restricting amino acids and are incorporated into the active restricting amino acid set. .

[0040] To avoid instantaneous fluctuations in the active set caused by sampling noise in single-point threshold judgment, a secondary confirmation of the transition boundary region is further performed by incorporating local slope changes: Within a continuously scanned neighborhood, if the difference in the candidate limiting amino acid... The relative abundance scores of different candidate limiting amino acids do not exceed the aforementioned tolerance boundary; and the sign of the difference between relative abundance scores changes between adjacent sampling points within the continuous scanning neighborhood, i.e., a zero-crossing point changes from positive to negative. This indicates that the abundance relationships of the two amino acids substantially intersect, and this neighborhood essentially constitutes the intersection region of the limiting amino acid's role as a bottleneck. The system classifies this neighborhood into the transition boundary region and includes all candidate limiting amino acid identifiers involved in it into the active limiting amino acid set. The contiguous neighborhood used for confirmation must contain at least [number of elements]. One sampling point, Take 5 experience points.

[0041] In this application, the local slope is calculated using the first-order difference between adjacent sampling points, and the difference in relative abundance scores between adjacent sampling points is used to characterize the trend of change; the continuous neighborhood used for confirmation includes at least Each sampling point forms a decision window, requiring the sign reversal phenomenon to occur within this continuous window interval. This filters out isolated single-point random noise and prevents sampling noise from causing instantaneous fluctuations in the active set and resulting in misjudgments.

[0042] The number of sampling points in a continuous neighborhood is the minimum number of continuous sampling points required to filter sampling noise. Combined with the calibration of the equipment's proportion fluctuation range, a reasonable value range is 3–10. When raw material detection noise is high, the number of sampling points can be automatically increased to enhance filtering. Values ​​that are too small are easily affected by noise from single-point sampling and may lead to misjudgment of transition regions. Taking values ​​that are too high will smooth out narrow transition intervals, causing missed detection of transition boundaries and loss of the optimal ratio range.

[0043] If no change in the limiting amino acid marker is detected after the bone protein mass percentage parameter has been fully traversed within the range of zero to one, and no two or more essential amino acids are simultaneously within the tolerance boundary, then it is determined that there is no clear phenomenon of limiting amino acid replacement in the current dual-protein mixture system within the continuous ratio domain.

[0044] In this scenario, the system will instead identify locally high-value regions on the lower envelope potency curve as candidate regions: First-order difference calculations are performed on the lower envelope potency sequence, and zero-crossing points where the numerical change changes from positive to negative are identified as local maxima. Subsequently, the system checks whether the limiting amino acid identifiers of this local maxima and its consecutive sampling points before and after it remain consistent. If the identifiers are consistent, the limiting amino acid identifiers corresponding to the lower envelope potency value at this local peak are considered stable and traceable. Using the bone protein mass percentage corresponding to this peak as the center, and combining engineering parameters such as the equivalent mass step distance of the bone phase device, the equivalent mass step distance of the bean phase device, and the protein content coefficients of the bone and bean phases, the system estimates the equivalent width corresponding to the step distance of a single device on the bone protein mass percentage axis, and expands the candidate region to an interval extending twice the equivalent width to both sides of the peak. This batch will be marked as a batch without explicit limiting amino acid replacement, and its active limiting amino acid set will be the limiting amino acid identifiers that actually form the lower envelope potency value within this candidate region.

[0045] When a transition exists, the boundary region of the restricted amino acid transition is obtained, along with the set of all active restricted amino acids, including those before and after the transition and those that are close to the short plate. When no transition exists, candidate regions based on the local high value region of the lower envelope and the corresponding set of active restricted amino acids are obtained.

[0046] S3: Within the candidate region, the theoretical ratio is discretized into actual physical feeding nodes by combining the purity of raw material protein, equipment metering step distance and total amount of single batch feeding. The lower envelope effect value corresponding to each actual physical feeding node is calculated in reverse. Valid nodes that meet the constraints are screened out through closure verification.

[0047] By converting continuous proportioning intervals into discrete feeding instructions executable by actual production equipment, and considering the minimum equivalent mass step distance of the bone and bean phase feeding mechanisms, and the limitation of the total batch feeding amount by the rated capacity of the reactor, the system needs to generate a series of discrete feeding nodes within the candidate area. Through dual verification of material conservation and proportioning mapping, effective nodes that meet both equipment execution capabilities and nutritional attribution requirements are selected. Specifically, the following sub-steps are included: The protein content coefficients of bone-phase raw materials and soybean-phase raw materials are obtained, representing the mass proportion of the target protein in a unit mass of bone-phase or soybean-phase physical raw materials, respectively. In this embodiment, the protein content coefficients of bone-phase raw materials and soybean-phase raw materials are not fixed constants. A process for single-batch real-time correction, multi-batch sliding iterative update, and anomaly re-inspection is set up, and the specific execution method is as follows: Real-time calibration for each batch: Before each batch of feed is put into operation, parallel samples of bone raw materials and soybean raw materials are taken to conduct quantitative detection of crude protein. Based on the measured mass ratio of protein per unit raw material obtained from the detection, the protein content coefficient of the raw materials used in this batch is updated in real time. Multi-batch sliding iterative update: The system caches the corrected raw material protein content coefficients of 5 consecutive production batches, calculates the arithmetic mean of the 5 sets of coefficients as the initial benchmark coefficient for the next batch feeding calculation, and weakens the random error of single sampling detection by averaging multiple batches; Raw material anomaly re-inspection mechanism: Compare the measured coefficient of this batch with the sliding iteration benchmark coefficient. When the absolute deviation between the two exceeds the preset floating threshold plus or minus 0.10, the protein content of the raw materials in this batch is determined to be abnormal, triggering a second parallel sampling and retesting of the raw materials. The protein content coefficient of the raw materials is updated with the qualified retest result. If the retest deviation still exceeds the threshold, an abnormal raw material quality warning is output, and the proportion optimization calculation is suspended. In this embodiment, the empirical value of the protein content coefficient of bone phase raw materials is 0.80, and the empirical value of the protein content coefficient of soybean phase raw materials is 0.40. These are only the average empirical benchmark values ​​of conventional raw materials and are only used as temporary initial reference values ​​when there is no previous batch data and raw material testing has not yet been carried out. The above correction iteration process is performed on all formal production batches to dynamically update the protein content coefficient of the raw materials.

[0048] The complete conversion logic of the two-layer system is as follows: This application divides the overall system into an independent nutritional theoretical calculation layer and an industrial actual feeding conversion layer. The aforementioned calculations of the relative abundance and lower envelope potency of essential amino acids are all carried out at the nutritional theoretical calculation layer, and nutritional assessment is only performed on the pure protein mixture system. Non-protein components such as moisture, carbohydrates, and impurities in the raw materials do not contain essential amino acids and do not participate in the nutritional potency calculation. The two layers establish a bidirectional data mapping link based on the protein purity parameters of the raw materials. During the calculation, the total mass of mixed pure protein required for a single batch of product is first preset. Based on the theoretical bone protein ratio obtained from nutritional optimization, the masses of bone-derived pure protein and soybean-derived pure protein required to achieve the target ratio are calculated separately. Combining the proportion of effective protein per unit mass of bone raw materials and soybean raw materials, the required mass of pure protein is converted into the theoretical raw material feeding weight that is not limited by the feeding accuracy of the equipment. Since the automated feeding equipment can only output materials that are integer multiples of the minimum step weight, the theoretical feeding weight needs to be rounded down to obtain the actual physical feeding weight that meets the hardware constraints of the equipment. Based on the actual physical weight of the feed ingredients, and combined with the protein ratio of the raw materials, the actual bone protein and soybean protein content of the feed system are reconstructed to further calculate the actual bone protein ratio formed after the feed was administered. The actual bone protein ratio is then substituted back into the calculation rules for the relative abundance and lower envelope potency of all essential amino acids mentioned earlier to re-evaluate the amino acid complementarity effect of the feed scheme. Through a two-way calculation process—deriving the feed weight from the theoretical ratio and then correcting the nutritional potency from the actual feed weight—a complete calculation loop is established between theoretical nutritional calculations and actual feed input on the production line.

[0049] Simultaneously, the equivalent mass step distances of the bone phase equipment and the bean phase equipment are obtained, representing the increment of feed mass corresponding to a single stepping action of the feeding mechanism; and the total feed volume constraint threshold for a single batch. In this embodiment, the equivalent mass step distances of the bone and bean phase equipment are both taken as empirical values ​​of 0.1 kg. The equipment equivalent mass step distance represents the minimum feed increment of the feeding mechanism in a single step. Based on the factory metering accuracy of the equipment, it has been calibrated and measured multiple times by stepping weighing, and the reasonable value range is 0.01 kg to 1 kg. It is automatically updated as the reactor specifications are switched between small-scale, pilot-scale, and mass production. Setting the step distance less than the actual stepping of the equipment will lead to ratio deviation and failure of a large number of node verifications. Setting the step distance greater than the actual stepping of the equipment will artificially reduce the feed accuracy and lose the discrete ratio nodes with high nutritional value. The threshold for the total amount of material fed in a single batch is based on an empirical value of 100 kg. The maximum loading mass of raw materials in a single batch in the reactor is determined by converting the reactor's rated volume and material density, with a reasonable range of 10 kg to 2000 kg. This threshold is adjusted synchronously with the reactor specifications for small-scale, pilot-scale, and mass production. If the threshold is lower than the actual loading limit of the reactor, the total amount of material fed at all stages will exceed the limit, resulting in no usable formula. If the threshold is higher than the reactor's load-bearing limit, an overcapacity feeding scheme will be output, posing a safety hazard of overflow during production. Implementers can adjust the above parameters based on actual raw material testing reports and equipment parameters.

[0050] Within the candidate region, a series of discrete sampling points representing the theoretical bone protein content are generated according to a preset sampling step size. The preset sampling step size is an empirical value of 0.001. This parameter represents the discrete interval of the theoretical ratio within the candidate region, and the ratio mapping tolerance is used for calibration. A reasonable value range is 0.0005 to 0.005. For high-precision products or narrow candidate regions, the step size can be automatically reduced to densify the nodes. If the step size is too small, it will significantly increase the time spent on feed conversion and closure verification calculations. If the step size is too large, the discrete nodes will be sparse, skipping the optimal nutritional ratio within the candidate region. Its value should not exceed the subsequent ratio mapping tolerance, and can also be adjusted by the implementer according to the required grid density.

[0051] For each theoretical sampling point, the total mass of the target protein in the current batch is used as a benchmark. The theoretical target mass of bone protein is calculated by multiplying the theoretical bone protein mass ratio of the sampling point by the total mass of the target protein. The theoretical target mass of soy protein is calculated by multiplying the value after subtracting the theoretical bone protein mass ratio of the sampling point by the total mass of the target protein.

[0052] Next, considering the difference in raw material protein purity, the theoretical target mass of bone protein calculated above is divided by the raw material protein content coefficient of bone phase to convert it into the theoretical physical feed mass of bone phase; the theoretical target mass of soybean protein is divided by the raw material protein content coefficient of soybean phase to convert it into the theoretical physical feed mass of soybean phase.

[0053] Since the feeding equipment can only feed in integer multiples of the step distance, the theoretical physical feeding mass must be rounded up according to the step distance. Using the nearest rounding method, the theoretical physical feeding mass of the bone phase is divided by the equivalent mass step distance of the bone phase equipment, then 0.5 is added and rounded down to obtain the number of steps required by the bone phase equipment. Similarly, the theoretical physical feeding mass of the bean phase is divided by the equivalent mass step distance of the bean phase equipment, then 0.5 is added and rounded down to obtain the number of steps required by the bean phase equipment.

[0054] Subsequently, the number of steps executed by the bone phase equipment is multiplied by the equivalent mass step distance of the bone phase equipment to obtain the actual physical feeding mass of the bone phase; the number of steps executed by the bean phase equipment is multiplied by the equivalent mass step distance of the bean phase equipment to obtain the actual physical feeding mass of the bean phase.

[0055] Based on the actual physical feed mass and raw material protein content coefficient mentioned above, the actual protein content corresponding to this discrete node is calculated in reverse: the actual physical feed mass of the bone phase is multiplied by the protein content coefficient of the bone phase raw material to obtain the protein content contained in the actual input bone phase raw material; the actual physical feed mass of the bean phase is multiplied by the protein content coefficient of the bean phase raw material to obtain the protein content contained in the actual input bean phase raw material. The actual bone protein content ratio of this discrete node is further calculated: the actual bone protein content is divided by the sum of the actual bone protein content and the actual bean protein content to obtain the corrected actual bone protein content ratio. The corrected actual bone protein content ratio is used to replace the original theoretical sampling point as input for subsequent nutritional evaluation.

[0056] After obtaining the actual feed parameters and potency for each discrete node, a closure check is performed to filter out nodes that exceed acceptable boundaries due to step rounding or protein purity conversion. The closure check includes three conditions: Calculate the total loading deviation: This measures the degree of deviation between the actual physical total mass of the material fed to discrete nodes and the batch total quantity constraint. The calculation method is to take the sum of the actual physical mass of the bone phase and the actual physical mass of the bean phase, and subtract the single batch total quantity constraint threshold from the absolute value. If this absolute value is less than or equal to the preset total loading deviation tolerance threshold (e.g., 0.5 kg), then this check passes. The total loading deviation tolerance threshold is the maximum allowable difference between the total mass of the material fed and the batch threshold. Combined with the weighing sensor error and the safety margin calibration for step rounding deviation, a reasonable value range is 0.1 kg to 2 kg. The threshold is lowered for precision micro-scale equipment and raised for large-scale mass production coarse weighing equipment. A threshold that is too small will cause high-quality batching nodes to be rejected due to normal weighing deviations, while a threshold that is too large will cause the total material quantity to deviate significantly from the rated capacity of the reactor, leading to production process control issues.

[0057] Calculate the proportion mapping offset: This characterizes the offset of the actual bone protein mass percentage relative to the theoretical sampling point. It is calculated as the absolute value of the difference between the actual bone protein mass percentage and the theoretical sampling point percentage. If this absolute value is less than or equal to the preset proportion mapping tolerance (e.g., 0.005), the proportion mapping tolerance is the maximum allowable offset of the actual protein mass percentage relative to the theoretical percentage, calibrated by the upper limit of fluctuation in the nutritional value of the mixed protein. A reasonable range is 0.002–0.01. For infant and toddler products with stringent requirements, the tolerance is narrowed; for coarse meal replacement products, the tolerance is relaxed. Too small a tolerance will result in the elimination of all feasible production formulas due to normal equipment step-by-step offset. Too large a tolerance will cause the actual proportion to deviate from the optimal complementary range, significantly reducing the product's amino acid nutritional value. In this case, the verification passes.

[0058] Verify the consistency of limiting amino acid attribution: Determine whether the limiting amino acid identifier of the discrete node still belongs to the set of active limiting amino acids associated with the candidate region. If it does, it indicates that although the node has been rounded by the device, it is still within the limiting amino acid transition region of interest, and no essential nutritional deficiency has occurred.

[0059] A discrete node is retained as a valid node only if all three of the above conditions are met simultaneously. Each valid node stores the corresponding actual physical feed mass of the bone phase, the actual physical feed mass of the bean phase, the actual bone protein content, the lower envelope effect value, and the limiting amino acid identifier. After traversing all sampling points within the candidate region, the set of valid nodes is obtained.

[0060] S4: Obtain the raw material analysis uncertainty of the amino acid detection error, convert the lower envelope effect value of each effective node into a confidence interval, and perform cross-domination judgment by comparing the upper and lower limits of the confidence intervals of each effective node to screen out dominated nodes whose confidence lower limit is lower than the confidence upper limit of other nodes.

[0061] In the effective node set, since the difference in lower envelope effectiveness values ​​between adjacent discrete nodes is usually small, and liquid chromatography detection has inherent analytical uncertainty, if only a single detection value is used for selection, random errors may cause unnecessary switching of the output formulation between adjacent grids. Therefore, this step introduces a raw material analysis uncertainty factor, extending the deterministic effectiveness value to a confidence interval, and using interval cross-domination criteria to screen for statistically significant nodes with the best resistance to disturbances. Specifically, it includes the following sub-steps: First, a preset uncertainty factor for raw material analysis is obtained. In this embodiment, an empirical value of 0.03 is used. This factor is dimensionless and can be adjusted by the implementer based on instrument repeatability data. The uncertainty factor characterizes the random error in amino acid detection and is calibrated by the mean of the coefficients of variation of multiple batches of raw material detection. A reasonable range is 0.01–0.08. The adjustment factor is set down for high-precision chromatographic instruments and up for simple, rapid detection equipment. Too small an uncertainty factor will underestimate the detection error and cause frequent formula oscillations and switching. Too large an uncertainty factor will widen the confidence interval, causing cross-dominance screening to fail, requiring additional sample testing for each decision, and significantly increasing detection costs. For each node in the effective node set, the following operations are performed to construct its valence confidence interval: Multiplying the lower envelope efficacy value corresponding to the node by the uncertainty factor yields the deviation. Subtracting this deviation from the original efficacy value gives the confidence infimum, which represents the conservative nutritional complementarity valence boundary that the node may reach under the influence of analytical uncertainty. Adding the aforementioned deviation to the lower envelope efficacy value corresponding to the node gives the confidence supremum, which represents the higher nutritional complementarity valence boundary that the node may reach. The confidence infimum and the confidence supremum together constitute the valence confidence interval of the node.

[0062] Iterate through any two distinct nodes in the set of valid nodes, denoted as the first node and the second node, and compare them according to the following logic: When the confidence lower bound of the first node is strictly greater than the confidence upper bound of the second node, it indicates that even under the most unfavorable detection error conditions, the lower bound of the nutritional complementarity valence of the first node is still higher than the upper bound of the second node. In this case, the first node is considered to have a clear and conservative dominance over the second node.

[0063] Once it is determined that the first node dominates the second node, the dominated second node is permanently removed from the set of valid nodes and will no longer participate in subsequent comparisons. This process does not affect the information already recorded by the remaining nodes, such as the actual physical feed quality of the bone phase, the actual physical feed quality of the bean phase, the actual bone protein content, and the limiting amino acid label.

[0064] After all node pairs have completed the cross-domination determination, the remaining nodes not dominated by any node constitute the set of undominated nodes output in this step. Each node in the set of undominated nodes has no other strictly conservative, better node within the existing analysis uncertainty range. Based on the number of nodes in the set, the system outputs the following status: When the set contains only one node, it indicates that the node with the unique and preferred anti-disturbance formula within the uncertainty range has been selected, and the feeding parameters corresponding to that node are directly output.

[0065] When the set contains multiple nodes, it indicates that the confidence intervals of the remaining nodes overlap, making it impossible to determine a unique winner using conservative boundaries. The current step uses this multi-node set as the output state for subsequent manual decision-making or selection using other auxiliary rules.

[0066] S5: When there are multiple valid nodes and the optimal node cannot be uniquely determined by cross-dominance determination, supplementary sampling and testing are performed on the detection items corresponding to the active restriction amino acid set to update the confidence interval of the valid node and perform cross-dominance determination again; if a unique decision still cannot be made, output according to the conservative strategy.

[0067] This step aims to address the problem of overlapping confidence intervals for multiple candidate nodes due to detection uncertainty, making it impossible to directly determine a unique optimal solution. When the set of undominated nodes output by S4 contains two or more nodes, it indicates that, given the current data accuracy, the differences in the quality of these nodes are within the error range. Therefore, the system triggers a supplementary sampling and re-judgment process based on key amino acids; if the re-judgment still fails to distinguish them, a conservative output strategy based on safety boundaries is executed. Specifically, this includes the following sub-steps: Because multiple nodes with overlapping confidence intervals exist within the undominated node set, the original detection data alone cannot determine the unique optimal feeding node through conservative boundaries, thus triggering a supplementary sampling and re-evaluation process. To improve efficiency and reduce detection costs, the scope of supplementary sampling is not extended to all essential amino acid detection items, but is limited to the types of amino acids contained in the active restricted amino acid set associated with the current batch. Specifically, samples are resampled from the bone phase and bean phase raw materials of the current batch, and the amino acid liquid chromatography analysis method is the same as that used in the initial detection. This technique is well-known to those skilled in the art and will not be described in detail here. The content of each amino acid belonging to this set is re-measured to obtain updated detection values.

[0068] The system replaces the elements corresponding to each amino acid in the active limiting amino acid set within the initial detection data vector with the retested values, while the remaining elements in the vector retain their original detection values. For each node in the multi-node set, using the actual proportion of bone protein stored in that node as input, and based on the defined relative abundance score calculation method and lower envelope potency extraction method, the updated lower envelope potency value corresponding to that node is recalculated. On this basis, following the same uncertainty expansion method, the updated confidence interval of the node is calculated using the raw material analysis uncertainty factor. Specifically, the operation steps are as follows: Multiply the updated lower envelope effectiveness value by the uncertainty factor to obtain a new bias, and then subtract this bias from the updated effectiveness value to obtain the updated confidence lowerfimum.

[0069] Adding the new bias to the updated lower envelope value yields the updated confidence supremacy. The updated confidence supremacy and confidence supremacy together constitute the updated valence confidence interval for this node.

[0070] After obtaining the updated confidence intervals of all nodes, the system performs the cross-dominance determination between nodes again. The determination rule is exactly the same as the previous steps: for any two nodes, if the updated confidence lower bound of one node is strictly greater than the updated confidence upper bound of the other node, then the former dominates the latter, and the dominated node is removed from the set.

[0071] If, after this round of judgment, only one uncontrolled node remains in the multi-node set, the system will use that node as the final feeding strategy node and end the supplementary judgment process.

[0072] If, after supplementary sampling and reassessment, the supplementary test data still has uncertainty, or the test value does not change substantially due to sample abnormalities, and the updated confidence interval still overlaps between multiple nodes, making it impossible to filter to a unique node through the above dominance rules, i.e., a unique decision cannot be achieved, then the system stops further supplementary sampling feedback and instead executes a conservative output strategy.

[0073] The specific steps of the conservative output strategy are as follows: First, compare the updated confidence lower bounds of all currently undominated nodes and select the node with the highest confidence lower bound as the output node. The highest confidence lower bound indicates that, considering analytical uncertainty, the lower bound of nutrient complementarity at this node is most conservatively better than other nodes. If two or more nodes have the same highest confidence lower bound, further compare the total loading deviation of these nodes and select the node with the smallest total loading deviation as the output node. The definition of total loading deviation is consistent with the closure check in the previous steps; the smaller its value, the closer the actual total feed mass of the node is to the single-batch total constraint, and the smaller the material redundancy or deficit during feed execution.

[0074] The actual physical feed quality of the bone phase and the actual physical feed quality of the bean phase corresponding to the finally selected output node will be issued as the feed formula for this batch, and the decision path identifier of this node will be recorded as a conservative output after supplementary sampling and re-judgment. At the same time, complete information such as the active limiting amino acid set attribution, actual bone protein mass ratio, lower envelope titer, confidence interval, and total loading deviation of this node will be retained for production batch traceability.

[0075] Based on the above description of the implementation method for optimizing the proportioning process, the technical effect of the present invention is verified through comparative simulation experiments. The comparative verification examples are as follows: To objectively verify the technical effectiveness of the formulation optimization scheme of this invention compared with the traditional static single-point formulation calculation scheme, a single-variable control simulation experiment was set up. The experiment uniformly used the same amino acid detection data of bone protein isolate and soy protein isolate, and unified the formulation scanning range and raw material fluctuation range. The simulation process introduced raw material protein content fluctuations and random errors in amino acid detection that conform to the actual industrial level to simulate real production disturbances; in multiple independent simulation calculations, only the core formulation optimization algorithm was replaced, while the other boundary conditions remained the same.

[0076] The control group used a traditional static single-point ratio calculation scheme: nutritional potency was calculated only at discrete sampling points, without setting a curve interpolation and encrypted sampling mechanism, without a special identification process for amino acid transition boundaries, using a fixed raw material protein coefficient, and lacking a reverse verification mechanism for feed input; the experimental group used the complete ratio optimization process of this invention: including the construction of continuous lower envelope potency curves, interpolation and supplementary sampling to prevent omissions, batch adaptive correction of raw material protein coefficients, bidirectional conversion verification between pure protein nutrient layer and crude raw material feed layer, and node screening logic based on potency confidence intervals.

[0077] Simulation results were evaluated from multiple dimensions, including potency misjudgment within the transition interval, oscillation of the optimal ratio, deviation of the actual ratio from the theoretical optimal value, and the number of batches triggering amino acid retesting. Compared to the control group, the experimental group significantly reduced the possibility of potency misjudgment within the limiting amino acid transition interval, effectively suppressed oscillation of the optimal formula caused by random perturbations in detection, reduced the deviation of the actual feed ratio from the theoretical optimal ratio, and reduced the frequency of amino acid retesting under production conditions. This demonstrates that the present invention possesses significant technical advantages in terms of computational accuracy, continuous operational stability, and industrial applicability.

[0078] This invention also provides a system for optimizing the complementary amino acid ratio of bone marrow and soy milk dual proteins. For example... Figure 2 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the method for optimizing the complementary amino acid ratio of bone marrow and soy milk dual proteins according to the first aspect of the present invention. The system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, the setup and functions of which are known in the art and will not be described further here.

[0079] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for optimizing the complementary amino acid ratio of bone marrow and soy milk dual proteins, characterized in that, include: The set range of bone protein mass percentage is obtained. Based on the amino acid vectors of bone and soybean protein and the reference pattern, the relative abundance scores of essential amino acids under each percentage in the range of bone protein mass percentage are calculated, and the lower envelope valence curve reflecting the continuous change of the complementary shortcomings of the two proteins with the ratio is constructed. The replacement of limiting amino acid markers is analyzed based on the lower envelope valence curve to determine the transition boundary region. The limiting amino acids before and after the replacement, as well as the non-limiting amino acids that meet the approximation conditions, are recorded as the active limiting amino acid set. If no transition occurs, the local high value region is used as the candidate region. Within the candidate area, the theoretical ratio is discretized into actual physical feeding nodes by combining the purity of raw material protein, equipment metering step distance and total amount of single batch feeding, and the lower envelope effect value corresponding to each actual physical feeding node is calculated in reverse. Valid nodes that meet the constraints are screened out through closure verification. The raw material analysis uncertainty of the amino acid detection error is obtained, and the lower envelope effect value of each effective node is converted into a confidence interval. By comparing the upper and lower limits of the confidence intervals of each effective node, cross-domination judgment is performed to screen out dominated nodes whose lower confidence limit is lower than the upper confidence limit of other nodes. When there are multiple valid nodes and the optimal node cannot be uniquely determined by cross-dominance determination, supplementary sampling and testing are performed on the detection items corresponding to the active restriction amino acid set to update the confidence interval of the valid node and perform cross-dominance determination again; if a unique decision still cannot be made, the conservative strategy is adopted for output.

2. The method for optimizing the complementary amino acid ratio of bone marrow and soy milk dual proteins according to claim 1, characterized in that, The construction of the lower envelope potency curve that continuously varies with the proportion of bone protein includes: Based on the amino acid vectors of bone and soybean proteins and the reference pattern, the relative abundance score of each essential amino acid under each bone protein mass percentage is calculated. The relative abundance score is the ratio of the mixed amino acid content to the reference abundance benchmark value. The minimum relative abundance score for each of the bone protein mass percentages is taken as the lower envelope potency value to form a lower envelope potency curve that changes continuously with the bone protein mass percentage.

3. The method for optimizing the amino acid complementarity ratio of bone marrow and soy milk double protein according to claim 2, characterized in that, The active limiting amino acid set includes: During the continuous change of the bone protein mass percentage, the limiting amino acid markers that form the lower envelope effect value are recorded for each percentage. When the restricted amino acid identifier corresponding to adjacent percentage points changes, the interval where the change occurs is located as the restricted amino acid transition boundary region, and the restricted amino acid identifiers before and after the change are included in the active restricted amino acid set; For any bone protein mass percentage, if the difference between the relative abundance score of an essential amino acid that is not identified as a restricted amino acid and the lower envelope effect value meets a preset relative proximity tolerance boundary, then the amino acid is identified as a candidate restricted amino acid and incorporated into the active restricted amino acid set.

4. The method for optimizing the complementary amino acid ratio of bone marrow and soy milk dual proteins according to claim 3, characterized in that, The preset relative proximity tolerance boundary determination criteria include: If the difference between the relative abundance score of an essential amino acid and the lower envelope value does not exceed the product of the relative proximity tolerance coefficient and the larger of the lower envelope value and the preset positive lower limit, then the amino acid is identified as a potential limiting amino acid that satisfies the relative proximity tolerance boundary.

5. The method for optimizing the complementary amino acid ratio of bone marrow and soy milk dual proteins according to claim 3, characterized in that, The criteria for confirming the transition boundary region also include: If the difference between the relative abundance score of a potential limiting amino acid and the lower envelope value consistently satisfies the relative proximity tolerance boundary, and the relative abundance score difference between potential limiting amino acids shows a sign change or slope convergence, then the neighborhood is confirmed to belong to the transition boundary region, and the relevant amino acid is incorporated into the active limiting amino acid set.

6. The method for optimizing the complementary amino acid ratio of bone marrow and soy milk dual proteins according to claim 3, characterized in that, The methods for determining the candidate region include: Identify the local high-value region on the lower envelope titer curve. The local high-value region is the region where the lower envelope titer reaches a local peak and the limiting amino acid identifier of the lower envelope titer is stable and traceable within the region. The neighborhood around the ratio region is expanded according to the step size to form the candidate region, and the batch is marked as a batch without clearly defined amino acid replacement. The active restricted amino acid set is the set of restricted amino acid identifiers that form the lower envelope value within the candidate region.

7. The method for optimizing the amino acid complementarity ratio of bone marrow and soy milk double protein according to claim 1, characterized in that, The reverse calculation of the lower envelope effect value corresponding to each actual physical feeding node includes: Obtain the protein content coefficients of bone phase raw materials and soybean phase raw materials, as well as the equivalent mass step distance of bone phase equipment, the equivalent mass step distance of soybean phase equipment, and the total feed amount constraint threshold for a single batch. Based on the theoretical bone protein content ratio in the candidate area and the total mass of the target protein in the current batch, the theoretical target bone protein content and the theoretical target soybean protein content are calculated, and the theoretical target protein content is converted into the theoretical physical feed mass using the bone phase raw material protein content coefficient and the soybean phase raw material protein content coefficient. Based on the equivalent mass step distance of the bone phase equipment and the equivalent mass step distance of the bean phase equipment, the theoretical physical feeding mass is rounded to obtain the number of steps of the bone phase equipment and the number of steps of the bean phase equipment, and the corresponding actual physical feeding mass is determined. The actual bone protein content and actual soybean protein content are calculated in reverse based on the actual physical feed mass and the protein content coefficient, and the proportion of actual bone protein content is calculated; the corresponding lower envelope potency is recalculated based on the proportion of actual bone protein content.

8. The method for optimizing the complementary amino acid ratio of bone marrow and soy milk dual proteins according to claim 7, characterized in that, The valid nodes that satisfy the constraints after closure verification include: The difference between the sum of the actual physical feed mass and the single batch feed total quantity constraint threshold is calculated to obtain the total loading deviation. The difference between the actual bone protein mass percentage and the theoretical bone protein mass percentage is calculated to obtain the ratio mapping deviation; A node is defined as a valid node if the total loading deviation does not exceed the preset total closure tolerance, the ratio mapping deviation does not exceed the preset ratio mapping tolerance, and the limiting amino acid corresponding to the lower envelope effect value recalculated based on the actual bone protein content belongs to the set of active limiting amino acids.

9. The method for optimizing the amino acid complementarity ratio of bone marrow and soy milk double protein according to claim 1, characterized in that, The execution of the cross-dominance determination includes: For each of the aforementioned valid nodes, the deviation is obtained by multiplying the lower envelope effect value by the raw material analysis uncertainty. The lower envelope effect value minus the deviation is used as the confidence lowerfimum, and the deviation is added to it as the confidence upperfimum. The valid nodes are compared pairwise. When the confidence lower bound of one node is greater than the confidence upper bound of another node, it is determined that the former node dominates the latter node, and the dominated node is eliminated.

10. A system for optimizing the complementary amino acid ratio of bone marrow and soy milk dual proteins, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement the method for optimizing the amino acid complementarity ratio of bone marrow and soy milk dual proteins according to any one of claims 1-9.

Images