A deep learning-based intelligent optimization system and method for raw material proportioning of meal replacement powder

Abstract

The application relates to the technical field of meal replacement powder raw material proportioning optimization, and discloses a meal replacement powder raw material proportioning intelligent optimization system and method based on deep learning, which comprises the following modules: a raw material feature acquisition module, a proportioning demand construction module, a raw material correlation analysis module, a proportioning parameter optimization module and a proportioning deviation correction module, wherein the proportioning deviation correction module is configured to perform raw material proportion increase, decrease, replacement or maintenance adjustment on a candidate raw material proportioning scheme according to an index deviation state, and determine a target raw material proportioning scheme; and the optimization feedback updating module is configured to acquire actual evaluation data corresponding to the target raw material proportioning scheme, update the correlation between the raw material features according to the actual evaluation data, and optimize subsequent meal replacement powder raw material proportioning based on the updated correlation. The application can accurately determine the influence law of different raw material features on each target index.

Description

Technical Field

[0001] This invention relates to the field of meal replacement powder ingredient ratio optimization technology, and more specifically, to a deep learning-based intelligent optimization system and method for meal replacement powder ingredient ratio. Background Technology

[0002] The current methods for determining and adjusting the ingredient ratios of meal replacement powders mainly rely on the long-term industry experience accumulated by R&D personnel to make manual adjustments. This makes it difficult to fully combine the preset target nutritional needs, the physicochemical properties of different meal replacement ingredients, and the actual consumption application environment to achieve synergistic optimization of multiple core indicators. Ultimately, this results in meal replacement powders that generally have unbalanced nutritional structures, insufficient powder storage stability, decreased mixing performance when actually prepared and consumed, and poor suitability for users of different ages and with different needs.

[0003] Therefore, it is necessary to design a deep learning-based intelligent optimization system and method for the ingredient ratio of meal replacement powder to solve the problems existing in the current technology. Summary of the Invention

[0004] In view of this, the present invention proposes a deep learning-based intelligent optimization system and method for the proportion of ingredients in meal replacement powder, aiming to solve the technical problems that existing manual optimization of meal replacement powder proportions is difficult to achieve synergistic optimization of multiple core indicators, resulting in unbalanced nutritional structure, poor stability and mixing performance, and insufficient suitability for different populations.

[0005] In one aspect, this invention proposes a deep learning-based intelligent optimization system for the ingredient ratio of meal replacement powder, comprising: The raw material feature acquisition module is configured to collect raw material attribute data, historical formula data, and formula evaluation data corresponding to the raw materials of the meal replacement powder, and to extract features from the raw material attribute data to obtain raw material features; The ratio requirement construction module is configured to establish a ratio optimization unit based on the application requirement data corresponding to the target meal replacement powder, determine the target indicator set corresponding to the ratio optimization unit, and determine the current ratio state and the ratio state to be adjusted based on the target indicator set. The raw material correlation analysis module is configured to analyze the correlation between different raw material characteristics based on the historical formula data, formula evaluation data and raw material characteristics, and determine the influence of each raw material characteristic on each target indicator in the target indicator set. The proportioning parameter optimization module is configured to determine the proportioning adjustment parameters corresponding to each raw material based on the influence relationship and the proportioning state to be adjusted, and to generate a candidate raw material proportioning scheme based on the proportioning adjustment parameters. The ratio deviation correction module is configured to obtain the evaluation information corresponding to the candidate raw material ratio scheme, compare the evaluation information with the target index set, determine the index deviation status corresponding to the candidate raw material ratio scheme, and perform raw material ratio increase, decrease, replacement or maintenance adjustment on the candidate raw material ratio scheme according to the index deviation status to determine the target raw material ratio scheme. The optimization feedback update module is configured to obtain the actual evaluation data corresponding to the target raw material ratio scheme, update the correlation between raw material characteristics based on the actual evaluation data, and optimize the subsequent meal replacement powder raw material ratio based on the updated correlation.

[0006] Furthermore, when establishing a formulation optimization unit based on the application demand data corresponding to the target meal replacement powder, and determining the target index set corresponding to the formulation optimization unit, the following are included: Obtain application demand data corresponding to the target meal replacement powder, extract demand features from the application demand data, and obtain target population characteristics, nutritional demand characteristics, consumption environment characteristics, and product performance demand characteristics. The nutritional requirement constraints corresponding to the target meal replacement powder are determined based on the characteristics and nutritional needs of the target population. Based on the characteristics of the edible environment and the characteristics of the product performance requirements, determine the sensory performance requirements and processing adaptation requirements corresponding to the target meal replacement powder. A formulation optimization unit is established based on the nutritional requirement constraints, sensory performance requirements, and processing adaptation requirements, and a set of target indicators corresponding to the target meal replacement powder is determined based on the formulation optimization unit.

[0007] Furthermore, when determining the current proportioning state and the proportioning state to be adjusted based on the set of target indicators, the process includes: Obtain the ratio parameter information corresponding to the current meal replacement powder ingredient ratio scheme, and determine the current ratio characteristics of each ingredient based on the ratio parameter information; Based on the current proportioning characteristics, the target indicators in the target indicator set are matched to determine the indicator satisfaction status corresponding to the current raw material proportioning scheme; Based on the difference between the indicator satisfaction state and the target indicator set, determine the proportion deviation characteristics corresponding to the current proportion scheme; The current proportioning state is determined based on the proportioning deviation characteristics, and the proportioning state to be adjusted is determined based on the raw material adjustment requirements corresponding to the target indicators with deviations in the current proportioning state.

[0008] Furthermore, when analyzing the correlation between different raw material characteristics using deep learning to determine the influence of each raw material characteristic on each target indicator in the target indicator set, the following steps are taken: The historical formula data and formula evaluation data are obtained, and the raw material combination characteristics in the historical formula data are correlated and matched with the indicator evaluation characteristics in the formula evaluation data to obtain raw material ratio correlation samples. Based on the raw material ratio associated samples, the raw material characteristics corresponding to each raw material are fused to obtain a raw material combination feature set; wherein, the raw material characteristics include nutritional characteristics, physicochemical characteristics and sensory characteristics; The raw material combination feature set is input into the deep learning analysis process to perform feature mapping on the correlation between different raw material features and determine the degree of correlation between each raw material feature and each target indicator in the target indicator set. The influence of each raw material characteristic on the corresponding target indicator in the target indicator set is determined based on the degree of correlation.

[0009] Furthermore, when determining the proportion adjustment parameters for each raw material based on the aforementioned influence relationship and the proportion to be adjusted, the following steps are included: Obtain the index deviation information corresponding to the state of the ratio to be adjusted, and determine the target index to be optimized and the offset direction corresponding to the target index to be optimized based on the index deviation information; Based on the aforementioned influence relationship, the response relationship of raw material characteristics to the target index to be optimized is analyzed to determine the response direction of different raw material characteristic changes to the target index to be optimized. Based on the response direction and the state of the ratio to be adjusted, target raw materials that meet the adjustment requirements of the target indicators to be optimized are selected, and the ratio adjustment trend of each target raw material is determined. Based on the aforementioned ratio adjustment trend and the current ratio parameters corresponding to each target raw material, calculate the ratio adjustment parameters corresponding to each target raw material.

[0010] Further, when comparing the evaluation information with the target indicator set to determine the indicator deviation status corresponding to the candidate raw material ratio scheme, the process includes: The evaluation information corresponding to the candidate raw material ratio scheme is obtained, and the indicator features of the evaluation information are extracted to obtain the actual indicator features corresponding to the candidate raw material ratio scheme. Based on the set of target indicators, determine the indicator constraint range corresponding to each target indicator, and match the actual indicator characteristics with the indicator constraint range to determine the indicator compliance status corresponding to the candidate raw material ratio scheme; Based on the compliance status of the indicators, determine the deviation type corresponding to each target indicator in the candidate raw material ratio scheme, and determine the deviation direction of the corresponding indicators based on the deviation type; Based on the direction of the index deviation and the set of target indicators, the index deviation status corresponding to the candidate raw material ratio scheme is determined.

[0011] Furthermore, when determining the target raw material ratio scheme by adjusting the proportions of raw materials by increasing, decreasing, replacing, or maintaining them according to the deviation status of the indicators, the process includes: Based on the deviation status of the indicators, determine the indicator adjustment requirements corresponding to the candidate raw material ratio scheme, and determine the target raw material to be adjusted based on the indicator adjustment requirements; Based on the raw material characteristics and influence relationships corresponding to the target raw material to be adjusted, determine the response direction of the target raw material to the target indicator; When the actual indicator characteristics corresponding to the target indicator are lower than the indicator range corresponding to the target indicator set, and the target raw material to be adjusted has a positive impact on the target indicator, the proportioning parameter corresponding to the target raw material to be adjusted is increased; When the actual indicator characteristics corresponding to the target indicator are higher than the indicator range corresponding to the target indicator set, and the target raw material to be adjusted has a positive impact on the target indicator, the proportioning parameter corresponding to the target raw material to be adjusted is reduced. When the adjustment result of the target raw material to be adjusted causes deviations in other target indicators, the target raw material to be adjusted shall be replaced according to the adjustment requirements corresponding to the target indicators. When the actual indicator characteristics corresponding to the target indicator meet the target indicator set, and there is no need to adjust the target raw material to be adjusted, the proportioning parameters corresponding to the target raw material to be adjusted shall be maintained. The target raw material ratio scheme is determined based on the adjusted raw material ratio parameters.

[0012] Furthermore, when updating the correlation between raw material characteristics based on the actual evaluation data, and optimizing the subsequent meal replacement powder ingredient ratio based on the updated correlation, the process includes: Obtain the actual evaluation data corresponding to the target raw material ratio scheme, and determine the actual indicator characteristics corresponding to the target raw material ratio scheme based on the actual evaluation data; Based on the correlation analysis between the actual indicator characteristics and the raw material characteristics corresponding to the target raw material ratio scheme, the actual response relationship of the changes in actual raw material characteristics to each target indicator in the target indicator set is determined. By comparing the actual response relationship with the influence relationship, the changes in the influence relationship between the raw material characteristics and each target indicator in the target indicator set are determined; The influence relationship of the raw material characteristics on each target indicator in the target indicator set is updated according to the changes in the influence relationship, and the subsequent ratio of meal replacement powder raw materials is optimized based on the updated influence relationship.

[0013] Furthermore, when performing correlation analysis between the actual indicator characteristics and the raw material characteristics corresponding to the target raw material ratio scheme to determine the actual response relationship of changes in actual raw material characteristics to each target indicator in the target indicator set, the following steps are included: Obtain the raw material ratio parameter change information corresponding to the target raw material ratio scheme, and determine the change characteristics corresponding to each raw material characteristic based on the raw material ratio parameter change information; Based on the time correspondence and feature matching between the change characteristics and the actual indicator characteristics, the response mapping relationship between the change in raw material characteristics and the change in target indicators is determined. Based on the response mapping relationship, analyze the actual impact direction of the changes in the characteristics of each raw material on the corresponding target indicators in the target indicator set, and determine the actual response direction corresponding to the target indicator. Based on the actual direction of influence and the actual direction of response, determine the actual response relationship between the actual changes in raw material characteristics and each target indicator in the target indicator set.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention utilizes deep learning to pre-learn the correlation between historical formula data, raw material characteristics, and formula evaluation data. This allows for the precise determination of the influence patterns of different raw material characteristics on various target indicators. It eliminates the need for researchers to rely solely on experience for ratio adjustments, enabling synergistic optimization of multiple preset target indicators. This effectively ensures a balanced nutritional structure in the developed meal replacement powder, while also improving the powder's storage stability and reconstitution performance. Furthermore, it allows for adaptability optimization for different application needs and population groups, significantly improving the accuracy and efficiency of meal replacement powder formula development. Simultaneously, by continuously updating the influence relationship between raw material characteristics and target indicators using actual evaluation data, it continuously optimizes the subsequent ratio development process, further enhancing the accuracy of ratio optimization.

[0015] In another aspect, this invention also proposes a method for intelligent optimization of the ingredient ratio of meal replacement powder based on deep learning, comprising the following steps: Collect raw material attribute data, historical formula data, and formula evaluation data corresponding to the raw materials of the meal replacement powder, and extract features from the raw material attribute data to obtain raw material characteristics; Establish a ratio optimization unit based on the application demand data corresponding to the target meal replacement powder, determine the target index set corresponding to the ratio optimization unit, and determine the current ratio state and the ratio state to be adjusted based on the target index set; Based on the historical formula data, formula evaluation data, and raw material characteristics, the correlation between different raw material characteristics is analyzed through deep learning to determine the influence of each raw material characteristic on each target indicator in the target indicator set. Based on the aforementioned influence relationships and the state of the proportions to be adjusted, determine the proportion adjustment parameters corresponding to each raw material, and generate candidate raw material proportion schemes based on the proportion adjustment parameters. The evaluation information corresponding to the candidate raw material ratio scheme is obtained, and the evaluation information is compared with the target index set to determine the index deviation status corresponding to the candidate raw material ratio scheme. Based on the index deviation status, the raw material ratio scheme is adjusted by increasing, decreasing, replacing or maintaining the raw material ratio to determine the target raw material ratio scheme. Obtain the actual evaluation data corresponding to the target raw material ratio scheme, update the correlation between raw material characteristics based on the actual evaluation data, and optimize the subsequent meal replacement powder raw material ratio based on the updated correlation.

[0016] It is understandable that the aforementioned intelligent optimization system and method for the proportion of meal replacement powder ingredients based on deep learning have the same beneficial effects, and will not be elaborated further here. Attached Figure Description

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A structural block diagram of the intelligent optimization system for the proportion of raw materials in meal replacement powder based on deep learning, provided in an embodiment of the present invention; Figure 2 A flowchart of a deep learning-based intelligent optimization method for the ingredient ratio of meal replacement powder provided in an embodiment of the present invention. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] See Figure 1 As shown in some embodiments of this application, this embodiment provides a deep learning-based intelligent optimization system for the ingredient ratio of meal replacement powder, including: The raw material feature acquisition module is configured to collect raw material attribute data, historical formula data, and formula evaluation data corresponding to the raw materials of the meal replacement powder, and to extract features from the raw material attribute data to obtain raw material features; The ratio requirement construction module is configured to establish a ratio optimization unit based on the application requirement data corresponding to the target meal replacement powder, determine the target indicator set corresponding to the ratio optimization unit, and determine the current ratio state and the ratio state to be adjusted based on the target indicator set. The raw material correlation analysis module is configured to analyze the correlation between different raw material characteristics based on the historical formula data, formula evaluation data and raw material characteristics, and determine the influence of each raw material characteristic on each target indicator in the target indicator set. The proportioning parameter optimization module is configured to determine the proportioning adjustment parameters corresponding to each raw material based on the influence relationship and the proportioning state to be adjusted, and to generate a candidate raw material proportioning scheme based on the proportioning adjustment parameters. The ratio deviation correction module is configured to obtain the evaluation information corresponding to the candidate raw material ratio scheme, compare the evaluation information with the target index set, determine the index deviation status corresponding to the candidate raw material ratio scheme, and perform raw material ratio increase, decrease, replacement or maintenance adjustment on the candidate raw material ratio scheme according to the index deviation status to determine the target raw material ratio scheme. The optimization feedback update module is configured to obtain the actual evaluation data corresponding to the target raw material ratio scheme, update the correlation between raw material characteristics based on the actual evaluation data, and optimize the subsequent meal replacement powder raw material ratio based on the updated correlation.

[0021] It is understood that the deep learning-based intelligent optimization system for meal replacement powder ingredient ratios provided in this embodiment achieves a closed-loop process from raw material data collection, ratio requirement analysis, correlation modeling, dynamic parameter optimization to deviation correction and model iterative updates through the collaborative work of multiple modules. Specifically, the raw material feature collection module provides basic data support for the system; the ratio requirement construction module transforms user-end application scenarios into quantifiable technical indicators; the raw material correlation analysis module utilizes deep neural networks to mine high-dimensional nonlinear relationships, overcoming the limitations of traditional empirical rules; the ratio parameter optimization module performs precise control based on influence relationships, generating scientifically reasonable candidate solutions; the ratio deviation correction module introduces a feedback mechanism to ensure that the solution meets actual performance requirements under multi-objective constraints; and the optimization feedback update module continuously absorbs the actual evaluation results of new formulas, enabling online learning and adaptive evolution of the model, thereby continuously improving the accuracy and generalization ability of subsequent ratio recommendations.

[0022] Specifically, when establishing a formulation optimization unit based on the application demand data corresponding to the target meal replacement powder, and determining the set of target indicators corresponding to the formulation optimization unit, the following are included: Obtain application demand data corresponding to the target meal replacement powder, extract demand features from the application demand data, and obtain target population characteristics, nutritional demand characteristics, consumption environment characteristics, and product performance demand characteristics. The nutritional requirement constraints corresponding to the target meal replacement powder are determined based on the characteristics and nutritional needs of the target population. Based on the characteristics of the edible environment and the characteristics of the product performance requirements, determine the sensory performance requirements and processing adaptation requirements corresponding to the target meal replacement powder. A formulation optimization unit is established based on the nutritional requirement constraints, sensory performance requirements, and processing adaptation requirements, and a set of target indicators corresponding to the target meal replacement powder is determined based on the formulation optimization unit.

[0023] It is understandable that determining the nutritional requirements constraints for the target meal replacement powder based on the characteristics and nutritional needs of the target population means that, for people of different age groups (e.g., children, adults, elderly), specific physiological states (e.g., pregnancy, post-operative recovery), or health goals (e.g., fat loss, blood sugar control, muscle gain), the system can automatically analyze their corresponding nutrient intake standards, such as protein content not less than 25g / 100g, dietary fiber content in the range of 10-20g / 100g, and glycemic index (GI) ≤ 55, thus forming binding nutritional indicator boundaries. Simultaneously, considering the characteristics of the consumption environment (e.g., room temperature preparation, cold drink scenarios, or high temperature and humidity storage conditions) and product performance requirements (e.g., solubility ≥ 90%, clumping rate ≤ 3%, water activity Aw ≤ 0.60 within the shelf life), the technical requirements for sensory and processing dimensions are further refined. These multi-dimensional constraints together constitute the target indicator set, providing a clear optimization direction and evaluation benchmark for subsequent formulation optimization.

[0024] It is understandable that determining the sensory performance requirements and processing suitability requirements for the target meal replacement powder based on the characteristics of the consumption environment and the product performance requirements means that the system can identify the specific requirements of users for the physical properties and sensory experience of the meal replacement powder in different usage scenarios. For example, when stored in a high-temperature and high-humidity environment, it is necessary to ensure that the powder has good anti-caking properties and low hygroscopicity; in the cold beverage preparation scenario, the raw materials are required to have rapid dispersibility and low-temperature solubility stability; while in portable ready-to-eat applications, more attention is paid to the flowability, particle uniformity, and smoothness of the powder. These requirements are transformed into quantifiable index constraints, such as preparation time ≤ 60 seconds, sedimentation rate ≤ 5%, particle size D90 ≤ 250 μm, and color L value in the range of 75–90, as sensory performance requirements. Meanwhile, the processing adaptation requirements cover technical parameters that match the production process, such as a raw material particle size distribution range of 50–300 μm suitable for dry mixing processes, an adhesion characteristic threshold of 10–30 N suitable for wet granulation, or an upper temperature resistance limit of ≤80℃ to meet the heat sensitivity requirements of spray drying.

[0025] It is understood that establishing a proportioning optimization unit based on the aforementioned nutritional requirements, sensory performance requirements, and processing suitability requirements means that the system structurally integrates these multi-dimensional constraints to construct a unified optimization decision space. In this space, each type of constraint is mapped to a computable objective function or boundary condition, thus forming a comprehensive set of target indicators covering the three dimensions of nutrition, sensory performance, and processing. The proportioning optimization unit uses this set of indicators as the optimization benchmark to drive subsequent raw material selection and proportion adjustment processes. For example, when the target population is patients in the postoperative recovery period, the system not only sets nutritional constraints such as high protein and easy digestibility (e.g., protein content ≥30g / 100g, fat content ≤8g / 100g, dietary fiber content ≥12g / 100g), but also simultaneously introduces requirements for low viscosity and rapid dissolution, such as preparation time ≤45 seconds and viscosity ≤300mPa·s, as well as powder flowability standards suitable for aseptic filling (e.g., angle of repose ≤40°, moisture content ≤5%). This ensures that the final formula meets core nutritional needs while also considering practical user experience and production feasibility. Through this multi-objective collaborative modeling mechanism, the formulation optimization unit can effectively avoid performance imbalances caused by single-dimensional optimization, achieving overall optimization of meal replacement powder products in terms of function, quality, and process.

[0026] Specifically, when determining the current proportioning state and the proportioning state to be adjusted based on the set of target indicators, the following are included: Obtain the ratio parameter information corresponding to the current meal replacement powder ingredient ratio scheme, and determine the current ratio characteristics of each ingredient based on the ratio parameter information; Based on the current proportioning characteristics, the target indicators in the target indicator set are matched to determine the indicator satisfaction status corresponding to the current raw material proportioning scheme; Based on the difference between the indicator satisfaction state and the target indicator set, determine the proportion deviation characteristics corresponding to the current proportion scheme; The current proportioning state is determined based on the proportioning deviation characteristics, and the proportioning state to be adjusted is determined based on the raw material adjustment requirements corresponding to the target indicators with deviations in the current proportioning state.

[0027] Understandably, the formulation parameter information includes the type, proportion, mass fraction, and combination relationship of each ingredient in the current meal replacement powder formulation. The system analyzes the formulation parameter information to determine the actual proportion of each ingredient, establishes a corresponding relationship with the collected ingredient characteristics, and obtains the current formulation characteristics of each ingredient, including nutritional contribution characteristics, physicochemical performance contribution characteristics, and sensory performance contribution characteristics, which respectively characterize the effect status of the corresponding indicators of each ingredient.

[0028] Understandably, the target indicator set is a set of nutritional, sensory, and processing performance indicators constructed by combining the target population, nutritional needs, consumption environment, and product performance requirements. The system calculates the actual indicator characteristics of the current formulation based on the current proportion characteristics, matches them with the constraints of the target indicator set, determines the satisfaction status of each target indicator, and judges whether the formula meets the design requirements.

[0029] Understandably, the difference in indicators represents the degree of deviation between the actual indicator characteristics and the target indicator set. The system calculates the deviation for indicators that do not meet the standards, determines the main raw material factors causing the deviation by combining the influence relationship between raw material characteristics and target indicators, and then classifies the proportioning deviation characteristics into nutritional, sensory, and processing performance deviation characteristics according to the type of deviation, providing a basis for raw material adjustment.

[0030] Understandably, the system determines the current state of the meal replacement powder ingredient ratio based on the type, direction, and degree of deviation of the ratio deviation characteristics. Then, combined with the influence of each ingredient's characteristics on the target indicators, it determines the target ingredient to be adjusted, the direction of adjustment, and the adjustment range, forming the ratio state to be adjusted, which provides a basis for subsequent ratio optimization.

[0031] Specifically, when analyzing the correlation between different raw material characteristics using deep learning to determine the influence of each raw material characteristic on each target indicator in the target indicator set, the following steps are included: The historical formula data and formula evaluation data are obtained, and the raw material combination characteristics in the historical formula data are correlated and matched with the indicator evaluation characteristics in the formula evaluation data to obtain raw material ratio correlation samples. Based on the raw material ratio associated samples, the raw material characteristics corresponding to each raw material are fused to obtain a raw material combination feature set; wherein, the raw material characteristics include nutritional characteristics, physicochemical characteristics and sensory characteristics; The raw material combination feature set is input into the deep learning analysis process to perform feature mapping on the correlation between different raw material features and determine the degree of correlation between each raw material feature and each target indicator in the target indicator set. The influence of each raw material characteristic on the corresponding target indicator in the target indicator set is determined based on the degree of correlation.

[0032] In this embodiment, the deep learning analysis process includes feature encoding, feature interaction, and indicator response analysis.

[0033] Understandably, historical formula data includes the raw material composition, addition ratio, combination method, and formula adjustment records of historical meal replacement powders; formula evaluation data includes nutritional, physicochemical properties, sensory evaluation, and user feedback evaluation data for the corresponding historical formulas. The system performs structured analysis on the historical formula data, extracting the raw material combination characteristics of different raw material combinations. At the same time, it analyzes the formula evaluation data to extract indicator evaluation characteristics, and then establishes a raw material ratio correlation sample containing "raw material combination characteristics - indicator evaluation characteristics" according to the correspondence between the two, providing a data foundation for subsequent analysis.

[0034] Understandably, raw material characteristics are a set of performance characteristics extracted based on the properties of raw materials, including nutritional characteristics, physicochemical characteristics, and sensory characteristics, which respectively characterize the impact of raw materials on the nutritional composition, processing and storage performance, and consumption experience of meal replacement powder. The system, based on the raw material combination relationships of samples related to raw material ratios, weights and fuses the characteristics of multiple raw materials in the same formula according to the addition ratio, to obtain a raw material combination characteristic set reflecting the overall formula characteristics. This avoids evaluation bias caused by single-raw material analysis. For example, for a historical formula containing 30% soy protein, 40% oat flour, 20% inulin, and 10% sweetener, the raw material combination characteristic set of that formula is obtained by fusing the characteristics of each raw material according to the ratio.

[0035] Understandably, the deep learning analysis process is used to analyze the nonlinear relationship between raw material combination characteristics and formula evaluation results in a large amount of historical data: the system takes the raw material combination feature set as input and the corresponding indicator evaluation features as output reference, extracts the potential relationship between the changes of the two through multi-layer feature mapping, identifies the change law of different raw material parameters and target indicators, and finally obtains the degree of correlation between each raw material feature and the target indicator, and establishes the correspondence between the two.

[0036] Understandably, the system analyzes the influence direction and effect of raw material characteristics on target indicators based on the degree of correlation, determining the influence relationship of each raw material characteristic on the corresponding target indicator. This influence relationship includes both the direction and type of influence. The direction of influence is categorized as either increasing or decreasing, and the type of influence is categorized as direct or synergistic. For example, the protein characteristics of plant protein have a positive influence relationship with protein content; increasing the proportion of plant protein can increase protein content. The physicochemical characteristics of highly hygroscopic raw materials have a negative influence relationship with powder stability; increasing the proportion of this raw material will reduce storage stability. Dietary fiber characteristics have a synergistic influence on satiety indicators. This relationship indicates that adjustments to the combination of dietary fiber and protein raw materials can jointly affect the target indicator. Through this method, the system establishes the influence relationship between raw material characteristics and the set of target indicators, providing a basis for subsequently determining the direction of raw material adjustment and the parameters for adjusting the proportions based on the state of the mixture to be adjusted.

[0037] Specifically, when determining the proportion adjustment parameters for each raw material based on the aforementioned influence relationships and the state of the proportion to be adjusted, the following steps are included: Obtain the index deviation information corresponding to the state of the ratio to be adjusted, and determine the target index to be optimized and the offset direction corresponding to the target index to be optimized based on the index deviation information; Based on the aforementioned influence relationship, the response relationship of raw material characteristics to the target index to be optimized is analyzed to determine the response direction of different raw material characteristic changes to the target index to be optimized. Based on the response direction and the state of the ratio to be adjusted, target raw materials that meet the adjustment requirements of the target indicators to be optimized are selected, and the ratio adjustment trend of each target raw material is determined. Based on the aforementioned ratio adjustment trend and the current ratio parameters corresponding to each target raw material, calculate the ratio adjustment parameters corresponding to each target raw material.

[0038] Understandably, the indicator deviation information includes the current actual value of the target indicator to be optimized, the target constraint range, and the magnitude of the deviation. Based on this, the system can directly identify the target indicator that needs adjustment, and whether the indicator is currently above the target upper limit or below the target lower limit, thus determining the direction of the shift. For example, if the actual protein content is 20g / 100g, and the target requirement is no less than 25g / 100g, then the target indicator to be optimized is determined to be protein content, and the shift direction is that it is too low and needs to be increased. The system calls the influence relationship between raw material characteristics and target indicators obtained in advance through deep learning, matching the influence relationship of raw materials corresponding to the target indicator to be optimized one by one, clarifying the direction of change of the target indicator to be optimized when different raw material characteristics change, providing a basis for selecting and adjusting raw materials. Continuing with the example of low protein content, according to the influence relationship, it can be seen that the protein characteristics of soy protein isolate and whey protein have a direct positive impact on the protein content indicator; increasing the amount added can directly increase the protein content. However, the protein characteristics of maltodextrin have almost no positive contribution to this indicator, thus determining the response direction corresponding to different raw material characteristics.

[0039] Understandably, the system combines the response direction and the offset direction to select raw materials that can adjust the target indicator to the target range as target raw materials, and determines the corresponding raw material ratio adjustment trend. In the above case, the protein content needs to be increased. Soy protein isolate and whey protein, which positively affect this indicator, meet the adjustment requirements and are therefore identified as target raw materials, with the corresponding adjustment trend being to increase the addition ratio. Maltodextrin has no positive contribution; if there is already a total addition constraint, the adjustment trend is to decrease the addition ratio to make room for the target raw material. Finally, the system combines the current ratio parameters of each target raw material with the magnitude of the deviation and the degree of influence of the raw material on the indicator to calculate the specific adjustment range for each target raw material, obtaining the corresponding ratio adjustment parameters, completing the quantitative adjustment of the raw material ratio, and providing accurate parameter basis for generating optimized candidate ratio schemes.

[0040] Specifically, when comparing the evaluation information with the target indicator set to determine the indicator deviation status corresponding to the candidate raw material ratio scheme, the following steps are included: The evaluation information corresponding to the candidate raw material ratio scheme is obtained, and the indicator features of the evaluation information are extracted to obtain the actual indicator features corresponding to the candidate raw material ratio scheme. Based on the set of target indicators, determine the indicator constraint range corresponding to each target indicator, and match the actual indicator characteristics with the indicator constraint range to determine the indicator compliance status corresponding to the candidate raw material ratio scheme; Based on the compliance status of the indicators, determine the deviation type corresponding to each target indicator in the candidate raw material ratio scheme, and determine the deviation direction of the corresponding indicators based on the deviation type; Based on the direction of the index deviation and the set of target indicators, the index deviation status corresponding to the candidate raw material ratio scheme is determined.

[0041] Understandably, the evaluation information for candidate raw material formulation schemes includes estimated values ​​of various nutritional indicators calculated through raw material ratio adjustment parameters, as well as estimated values ​​of sensory performance and processing adaptability performance indicators derived through influence relationships. The system extracts these estimated values ​​as actual indicator features and compares them one by one with the preset constraint range in the target indicator set to determine whether each indicator meets the requirements: if the actual indicator feature is within the constraint range, it is marked as compliant; if it exceeds the constraint range, it is marked as non-compliant, and the corresponding deviation direction is recorded, such as higher than the upper limit or lower than the lower limit. Finally, the comparison results of all indicators are integrated to obtain the complete indicator deviation status of the candidate formulation scheme, providing a clear basis for subsequent judgment on whether the optimization requirements are met.

[0042] Specifically, when determining the target raw material ratio scheme by adjusting the proportions of raw materials by increasing, decreasing, replacing, or maintaining them according to the deviation status of the indicators, the process includes: Based on the deviation status of the indicators, determine the indicator adjustment requirements corresponding to the candidate raw material ratio scheme, and determine the target raw material to be adjusted based on the indicator adjustment requirements; Based on the raw material characteristics and influence relationships corresponding to the target raw material to be adjusted, determine the response direction of the target raw material to the target indicator; When the actual indicator characteristics corresponding to the target indicator are lower than the indicator range corresponding to the target indicator set, and the target raw material to be adjusted has a positive impact on the target indicator, the proportioning parameter corresponding to the target raw material to be adjusted is increased; When the actual indicator characteristics corresponding to the target indicator are higher than the indicator range corresponding to the target indicator set, and the target raw material to be adjusted has a positive impact on the target indicator, the proportioning parameter corresponding to the target raw material to be adjusted is reduced. When the adjustment result of the target raw material to be adjusted causes deviations in other target indicators, the target raw material to be adjusted shall be replaced according to the adjustment requirements corresponding to the target indicators. When the actual indicator characteristics corresponding to the target indicator meet the target indicator set, and there is no need to adjust the target raw material to be adjusted, the proportioning parameters corresponding to the target raw material to be adjusted shall be maintained. The target raw material ratio scheme is determined based on the adjusted raw material ratio parameters.

[0043] It should be noted that the index deviation status is used to characterize the deviation between the actual index characteristics of the candidate raw material formulation and the target index set, specifically including the type of deviation index, the direction of deviation, and the degree of deviation. The system will first analyze the index deviation status to identify the target index that does not meet the requirements, and then clarify the specific adjustment needs based on the deviation type corresponding to the target index. For example, when the actual protein content is lower than the range set by the target index set, it can be determined that there is a need for protein supplementation; when the actual sweetness evaluation value exceeds the range set by the target index set, it can be determined that there is a need for sweetness reduction; when the actual reconstitution performance index does not meet the target requirements, it can be determined that there is a need for solubility improvement.

[0044] It should be noted that the system will match the relevant raw materials that cause the deviation in the target indicators based on the influence relationship between the characteristics of the raw materials and the target indicators, and determine the target raw materials that can be adjusted to regulate the corresponding target indicators. For example, when the protein content is insufficient, the system will identify protein raw materials such as soy protein and whey protein as target raw materials to be adjusted based on the influence relationship; when the sugar content is too high, sweeteners or high-sugar raw materials will be identified as target raw materials to be adjusted, thereby establishing a correspondence between the indicator adjustment needs and the target raw materials.

[0045] It should be noted that the characteristics of the target raw material to be adjusted include its nutritional, physicochemical, and sensory characteristics. The influence relationship here refers to the interaction between the raw material characteristics and the target indicator obtained through deep learning analysis mentioned earlier. The system combines the raw material characteristics of the target raw material to be adjusted, as well as the direction of its influence on the target indicator, to determine the trend of the target indicator after adjusting the raw material ratio. For example, when the system determines that the nutritional characteristics of soybean protein have a positive influence on the protein content indicator, it means that increasing the proportion of soybean protein can increase the protein content; when the system determines that the carbohydrate characteristics of maltodextrin have a positive influence on the caloric indicator, it means that increasing the proportion of maltodextrin will increase the caloric indicator; when the system determines that the physicochemical characteristics of a highly hygroscopic raw material have a negative influence on the powder stability indicator, it means that increasing the proportion of this raw material may reduce the powder stability. Through the above analysis, the response direction corresponding to each target raw material to be adjusted can be determined, providing a basis for subsequent adjustments such as increasing, decreasing, or replacing raw materials.

[0046] This adjustment method is applicable to scenarios where the target indicator does not meet the minimum requirement, but the corresponding raw material can improve the target indicator. The system first determines whether the actual indicator characteristics are lower than the lower limit of the target indicator set. After confirming that the target indicator is not met, it further determines whether the target raw material to be adjusted can have a positive effect on the target indicator based on the influence relationship. If the above conditions are met, the addition ratio of the target raw material to be adjusted in the candidate raw material ratio scheme is increased. For example, if the target indicator set requires the protein content of meal replacement powder to be 25g / 100g to 35g / 100g, while the protein content calculated by the candidate raw material ratio scheme is 22g / 100g, and the system determines that the corresponding characteristics of whey protein have a positive impact on the protein content indicator, then the addition ratio of whey protein is increased, for example, from 15% to 20%, to improve the overall protein indicator and make it meet the target indicator requirements.

[0047] This adjustment method applies to scenarios where target indicators exceed the allowable range. When the actual characteristic of a target indicator exceeds the maximum constraint value of the target indicator set, the system first identifies the relevant raw materials causing the excessively high indicator, and then determines whether the raw material will further increase the target indicator based on the influence relationship. If the raw material has a positive impact on the target indicator, the proportion of the raw material is reduced to bring the target indicator back to a reasonable range. For example, if the target indicator set requires a sugar content ≤10g / 100g, while the sugar content calculated by the candidate raw material ratio scheme is 14g / 100g, and the system determines that there is a positive influence relationship between the sensory characteristics of a certain sweetener and the sugar content indicator, the proportion of that sweetener is reduced, for example from 5% to 2%, thereby reducing the sugar content indicator and meeting the requirements of the target indicator set.

[0048] This adjustment method is suitable for resolving conflicts in indicators arising from adjustments to a single raw material. After adjusting the proportion of a raw material by increasing or decreasing it, the system recalculates the actual indicator characteristics corresponding to the adjusted raw material combination to determine whether the adjustment will cause other target indicators to deviate from the target indicator set. For example, increasing the proportion of a certain protein raw material may increase protein content, but it may lead to decreased powder solubility or a rougher texture. In this case, it can be determined that the raw material adjustment may introduce new indicator deviations. The system can then replace the target raw material to be adjusted by considering the adjustment requirements of the target indicators and the interrelationships between raw material characteristics, selecting alternative raw materials with the same target efficacy but fewer side effects. For example, replacing plant protein, which has a significant negative impact on solubility, with whey protein, which has better solubility, can reduce its adverse effects on processing performance indicators while ensuring the improvement in protein content.

[0049] This adjustment method avoids ineffective adjustments to ingredient ratios that already meet optimization requirements. The system checks each current target indicator to ensure it falls within the constraints of the target indicator set. Then, considering the influence relationships of the target ingredient to be adjusted, it determines whether the ingredient requires optimization. If the target indicator meets the requirements and the current ingredient ratio will not cause deviations in other indicators, the ratio parameters corresponding to that ingredient are kept unchanged to maintain the stability of the current formula. For example, if the current oat flour ratio is 40%, and its dietary fiber content, texture, and mixing performance all meet the requirements of the target indicator set, and there are no deviations in other indicators, the oat flour ratio is directly retained to avoid unnecessary adjustments that could lead to a decline in overall formula performance.

[0050] The system recombines the raw material ratio parameters after additions, reductions, replacements, or maintenance to form new raw material combinations. It then recalculates the nutritional, sensory, and processing performance indicators corresponding to the adjusted scheme. If the adjusted scheme meets the constraints of the target indicator set, it is determined as the target raw material ratio scheme. If there are still indicator deviations, iterative adjustments are performed based on the deviation status until a target raw material ratio scheme that meets the target indicator set requirements is obtained. Through this process, the raw material ratio scheme for meal replacement powder can be automatically adjusted from candidate schemes to optimized schemes, improving the accuracy and adaptability of the formula optimization process.

[0051] Specifically, when updating the correlation between raw material characteristics based on the actual evaluation data, and optimizing the subsequent meal replacement powder ingredient ratio based on the updated correlation, the process includes: Obtain the actual evaluation data corresponding to the target raw material ratio scheme, and determine the actual indicator characteristics corresponding to the target raw material ratio scheme based on the actual evaluation data; Based on the correlation analysis between the actual indicator characteristics and the raw material characteristics corresponding to the target raw material ratio scheme, the actual response relationship of the changes in actual raw material characteristics to each target indicator in the target indicator set is determined. By comparing the actual response relationship with the influence relationship, the changes in the influence relationship between the raw material characteristics and each target indicator in the target indicator set are determined; The influence relationship of the raw material characteristics on each target indicator in the target indicator set is updated according to the changes in the influence relationship, and the subsequent ratio of meal replacement powder raw materials is optimized based on the updated influence relationship.

[0052] Understandably, the actual evaluation data refers to the real nutritional indicators, sensory evaluation scores, and processing performance parameters obtained through testing after the target raw material ratio scheme has been actually processed. This data directly reflects the actual performance of the ratio scheme. The system compares the actual indicator characteristics obtained from the actual testing with the theoretical indicator characteristics derived from optimization. Then, combining this with the actual characteristic parameters of each raw material in the current ratio, it analyzes the true correspondence between changes in raw material characteristics and changes in actual indicator characteristics, obtaining the actual response relationship. This actual response relationship is then compared with the initial influence relationship obtained from deep learning pre-training to determine the differences, thus revealing the changes in the influence relationship between raw material characteristics and each target indicator. Based on these changes, the original influence relationship is corrected and updated, continuously accumulating real data from actual production. This makes the judgment of the impact of raw material adjustments on indicators more accurate during subsequent ratio optimization, thereby continuously improving the actual adaptability of the system's optimization results and forming a data-driven optimization closed loop.

[0053] Specifically, when determining the actual response relationship of changes in actual raw material characteristics to each target indicator in the target indicator set by performing correlation analysis between the actual indicator characteristics and the raw material characteristics corresponding to the target raw material ratio scheme, the following steps are included: Obtain the raw material ratio parameter change information corresponding to the target raw material ratio scheme, and determine the change characteristics corresponding to each raw material characteristic based on the raw material ratio parameter change information; Based on the time correspondence and feature matching between the change characteristics and the actual indicator characteristics, the response mapping relationship between the change in raw material characteristics and the change in target indicators is determined. Based on the response mapping relationship, analyze the actual impact direction of the changes in the characteristics of each raw material on the corresponding target indicators in the target indicator set, and determine the actual response direction corresponding to the target indicator. Based on the actual direction of influence and the actual direction of response, determine the actual response relationship between the actual changes in raw material characteristics and each target indicator in the target indicator set.

[0054] Understandably, each optimization adjustment of the raw material ratio will create a trajectory of change in the raw material ratio parameters, resulting in corresponding changes in raw material characteristics. The system records the increase or decrease of all raw material characteristics from the initial candidate scheme to the final target scheme. This information is then synchronized with the actual detected changes in the target indicators in terms of time and features, clarifying which type of raw material characteristic change corresponds to which target indicator change, thus establishing a mapping relationship between raw material characteristic changes and indicator changes. Based on this, further analysis is conducted to determine whether the target indicator increases or decreases after the increase or decrease of raw material characteristics. This yields the actual impact direction and response degree of the raw materials on the target indicator in the real-world scenario. Finally, a complete actual response relationship is integrated, providing real data support for correcting the impact relationship of the original deep learning model, allowing the system's optimization judgment capability to continuously improve with practical applications.

[0055] See Figure 2 As shown in some embodiments of this application, this embodiment provides a method for intelligent optimization of the ingredient ratio of meal replacement powder based on deep learning, including the following steps: S100: Collect raw material attribute data, historical formula data and formula evaluation data corresponding to the raw materials of the meal replacement powder, and extract features from the raw material attribute data to obtain raw material characteristics; S200: Establish a ratio optimization unit based on the application demand data corresponding to the target meal replacement powder, determine the target index set corresponding to the ratio optimization unit, and determine the current ratio state and the ratio state to be adjusted based on the target index set; S300: Based on the historical formula data, formula evaluation data, and raw material characteristics, the correlation between different raw material characteristics is analyzed through deep learning to determine the influence of each raw material characteristic on each target indicator in the target indicator set; S400: Based on the influence relationship and the state of the ratio to be adjusted, determine the ratio adjustment parameters corresponding to each raw material, and generate a candidate raw material ratio scheme based on the ratio adjustment parameters; S500: Obtain the evaluation information corresponding to the candidate raw material ratio scheme, compare the evaluation information with the target index set, determine the index deviation status corresponding to the candidate raw material ratio scheme, and perform raw material ratio increase, decrease, replacement or maintenance adjustment on the candidate raw material ratio scheme according to the index deviation status to determine the target raw material ratio scheme; S600: Obtain the actual evaluation data corresponding to the target raw material ratio scheme, update the correlation between raw material characteristics based on the actual evaluation data, and optimize the subsequent meal replacement powder raw material ratio based on the updated correlation.

[0056] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. 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 goods 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.

[0057] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods 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.

[0058] 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.

[0059] 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.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A deep learning-based intelligent optimization system for the ingredient ratio of meal replacement powder, characterized in that, include: The raw material feature acquisition module is configured to collect raw material attribute data, historical formula data, and formula evaluation data corresponding to the raw materials of the meal replacement powder, and to extract features from the raw material attribute data to obtain raw material features; The ratio requirement construction module is configured to establish a ratio optimization unit based on the application requirement data corresponding to the target meal replacement powder, determine the target indicator set corresponding to the ratio optimization unit, and determine the current ratio state and the ratio state to be adjusted based on the target indicator set. The raw material correlation analysis module is configured to analyze the correlation between different raw material characteristics based on the historical formula data, formula evaluation data and raw material characteristics, and determine the influence of each raw material characteristic on each target indicator in the target indicator set. The proportioning parameter optimization module is configured to determine the proportioning adjustment parameters corresponding to each raw material based on the influence relationship and the proportioning state to be adjusted, and to generate a candidate raw material proportioning scheme based on the proportioning adjustment parameters. The ratio deviation correction module is configured to obtain the evaluation information corresponding to the candidate raw material ratio scheme, compare the evaluation information with the target index set, determine the index deviation status corresponding to the candidate raw material ratio scheme, and perform raw material ratio increase, decrease, replacement or maintenance adjustment on the candidate raw material ratio scheme according to the index deviation status to determine the target raw material ratio scheme. The optimization feedback update module is configured to obtain the actual evaluation data corresponding to the target raw material ratio scheme, update the correlation between raw material characteristics based on the actual evaluation data, and optimize the subsequent meal replacement powder raw material ratio based on the updated correlation.

2. The intelligent optimization system for the proportion of raw materials in meal replacement powder based on deep learning according to claim 1, characterized in that, When establishing a formulation optimization unit based on the application demand data corresponding to the target meal replacement powder, and determining the set of target indicators corresponding to the formulation optimization unit, the following are included: Obtain application demand data corresponding to the target meal replacement powder, extract demand features from the application demand data, and obtain target population characteristics, nutritional demand characteristics, consumption environment characteristics, and product performance demand characteristics. The nutritional requirement constraints corresponding to the target meal replacement powder are determined based on the characteristics and nutritional needs of the target population. Based on the characteristics of the edible environment and the characteristics of the product performance requirements, determine the sensory performance requirements and processing adaptation requirements corresponding to the target meal replacement powder. A formulation optimization unit is established based on the nutritional requirement constraints, sensory performance requirements, and processing adaptation requirements, and a set of target indicators corresponding to the target meal replacement powder is determined based on the formulation optimization unit.

3. The intelligent optimization system for the proportion of raw materials in meal replacement powder based on deep learning according to claim 2, characterized in that, When determining the current proportioning status and the proportioning status to be adjusted based on the set of target indicators, the following are included: Obtain the ratio parameter information corresponding to the current meal replacement powder ingredient ratio scheme, and determine the current ratio characteristics of each ingredient based on the ratio parameter information; Based on the current proportioning characteristics, the target indicators in the target indicator set are matched to determine the indicator satisfaction status corresponding to the current raw material proportioning scheme; Based on the difference between the indicator satisfaction state and the target indicator set, determine the proportion deviation characteristics corresponding to the current proportion scheme; The current proportioning state is determined based on the proportioning deviation characteristics, and the proportioning state to be adjusted is determined based on the raw material adjustment requirements corresponding to the target indicators with deviations in the current proportioning state.

4. The intelligent optimization system for the proportion of raw materials in meal replacement powder based on deep learning according to claim 3, characterized in that, When analyzing the correlation between different raw material characteristics using deep learning to determine the influence of each raw material characteristic on each target indicator in the target indicator set, the following methods are used: The historical formula data and formula evaluation data are obtained, and the raw material combination characteristics in the historical formula data are correlated and matched with the indicator evaluation characteristics in the formula evaluation data to obtain raw material ratio correlation samples. Based on the raw material ratio associated samples, the raw material characteristics corresponding to each raw material are fused to obtain a raw material combination feature set; wherein, the raw material characteristics include nutritional characteristics, physicochemical characteristics and sensory characteristics; The raw material combination feature set is input into the deep learning analysis process to perform feature mapping on the correlation between different raw material features and determine the degree of correlation between each raw material feature and each target indicator in the target indicator set. The influence of each raw material characteristic on the corresponding target indicator in the target indicator set is determined based on the degree of correlation.

5. The intelligent optimization system for the proportion of raw materials in meal replacement powder based on deep learning according to claim 4, characterized in that, When determining the proportion adjustment parameters for each raw material based on the aforementioned influence relationships and the state of the proportions to be adjusted, the following are included: Obtain the index deviation information corresponding to the state of the ratio to be adjusted, and determine the target index to be optimized and the offset direction corresponding to the target index to be optimized based on the index deviation information; Based on the aforementioned influence relationship, the response relationship of raw material characteristics to the target index to be optimized is analyzed to determine the response direction of different raw material characteristic changes to the target index to be optimized. Based on the response direction and the state of the ratio to be adjusted, target raw materials that meet the adjustment requirements of the target indicators to be optimized are selected, and the ratio adjustment trend of each target raw material is determined. Based on the aforementioned ratio adjustment trend and the current ratio parameters corresponding to each target raw material, calculate the ratio adjustment parameters corresponding to each target raw material.

6. The intelligent optimization system for the proportion of raw materials in meal replacement powder based on deep learning according to claim 5, characterized in that, When comparing the evaluation information with the target indicator set to determine the indicator deviation status corresponding to the candidate raw material ratio scheme, the following steps are included: The evaluation information corresponding to the candidate raw material ratio scheme is obtained, and the indicator features of the evaluation information are extracted to obtain the actual indicator features corresponding to the candidate raw material ratio scheme. Based on the set of target indicators, determine the indicator constraint range corresponding to each target indicator, and match the actual indicator characteristics with the indicator constraint range to determine the indicator compliance status corresponding to the candidate raw material ratio scheme; Based on the compliance status of the indicators, determine the deviation type corresponding to each target indicator in the candidate raw material ratio scheme, and determine the deviation direction of the corresponding indicators based on the deviation type; Based on the direction of the index deviation and the set of target indicators, the index deviation status corresponding to the candidate raw material ratio scheme is determined.

7. The intelligent optimization system for the proportion of raw materials in meal replacement powder based on deep learning according to claim 6, characterized in that, When determining the target raw material ratio scheme by adjusting the proportions of raw materials by increasing, decreasing, replacing, or maintaining them according to the deviation status of the indicators, the following steps are included: Based on the deviation status of the indicators, determine the indicator adjustment requirements corresponding to the candidate raw material ratio scheme, and determine the target raw material to be adjusted based on the indicator adjustment requirements; Based on the raw material characteristics and influence relationships corresponding to the target raw material to be adjusted, determine the response direction of the target raw material to the target indicator; When the actual indicator characteristics corresponding to the target indicator are lower than the indicator range corresponding to the target indicator set, and the target raw material to be adjusted has a positive impact on the target indicator, the proportioning parameter corresponding to the target raw material to be adjusted is increased; When the actual indicator characteristics corresponding to the target indicator are higher than the indicator range corresponding to the target indicator set, and the target raw material to be adjusted has a positive impact on the target indicator, the proportioning parameter corresponding to the target raw material to be adjusted is reduced. When the adjustment result of the target raw material to be adjusted causes deviations in other target indicators, the target raw material to be adjusted shall be replaced according to the adjustment requirements corresponding to the target indicators. When the actual indicator characteristics corresponding to the target indicator meet the target indicator set, and there is no need to adjust the target raw material to be adjusted, the proportioning parameters corresponding to the target raw material to be adjusted shall be maintained. The target raw material ratio scheme is determined based on the adjusted raw material ratio parameters.

8. The intelligent optimization system for the proportion of raw materials in meal replacement powder based on deep learning according to claim 7, characterized in that, When updating the correlation between raw material characteristics based on the actual evaluation data, and optimizing the subsequent meal replacement powder ingredient ratio based on the updated correlation, the following steps are included: Obtain the actual evaluation data corresponding to the target raw material ratio scheme, and determine the actual indicator characteristics corresponding to the target raw material ratio scheme based on the actual evaluation data; Based on the correlation analysis between the actual indicator characteristics and the raw material characteristics corresponding to the target raw material ratio scheme, the actual response relationship of the changes in actual raw material characteristics to each target indicator in the target indicator set is determined. By comparing the actual response relationship with the influence relationship, the changes in the influence relationship between the raw material characteristics and each target indicator in the target indicator set are determined; The influence relationship of the raw material characteristics on each target indicator in the target indicator set is updated according to the changes in the influence relationship, and the subsequent ratio of meal replacement powder raw materials is optimized based on the updated influence relationship.

9. The intelligent optimization system for the proportion of raw materials in meal replacement powder based on deep learning according to claim 8, characterized in that, When performing correlation analysis between the actual indicator characteristics and the raw material characteristics corresponding to the target raw material ratio scheme, and determining the actual response relationship of changes in actual raw material characteristics to each target indicator in the target indicator set, the following is included: Obtain the raw material ratio parameter change information corresponding to the target raw material ratio scheme, and determine the change characteristics corresponding to each raw material characteristic based on the raw material ratio parameter change information; Based on the time correspondence and feature matching between the change characteristics and the actual indicator characteristics, the response mapping relationship between the change in raw material characteristics and the change in target indicators is determined. Based on the response mapping relationship, analyze the actual impact direction of the changes in the characteristics of each raw material on the corresponding target indicators in the target indicator set, and determine the actual response direction corresponding to the target indicator. Based on the actual direction of influence and the actual direction of response, determine the actual response relationship between the actual changes in raw material characteristics and each target indicator in the target indicator set.

10. A deep learning-based intelligent optimization method for the ingredient ratio of meal replacement powder, applied to the deep learning-based intelligent optimization system for the ingredient ratio of meal replacement powder as described in any one of claims 1-9, characterized in that, include: Collect raw material attribute data, historical formula data, and formula evaluation data corresponding to the raw materials of the meal replacement powder, and extract features from the raw material attribute data to obtain raw material characteristics; Establish a ratio optimization unit based on the application demand data corresponding to the target meal replacement powder, determine the target index set corresponding to the ratio optimization unit, and determine the current ratio state and the ratio state to be adjusted based on the target index set; Based on the historical formula data, formula evaluation data, and raw material characteristics, the correlation between different raw material characteristics is analyzed through deep learning to determine the influence of each raw material characteristic on each target indicator in the target indicator set. Based on the aforementioned influence relationships and the state of the proportions to be adjusted, determine the proportion adjustment parameters corresponding to each raw material, and generate candidate raw material proportion schemes based on the proportion adjustment parameters. The evaluation information corresponding to the candidate raw material ratio scheme is obtained, and the evaluation information is compared with the target index set to determine the index deviation status corresponding to the candidate raw material ratio scheme. Based on the index deviation status, the raw material ratio scheme is adjusted by increasing, decreasing, replacing or maintaining the raw material ratio to determine the target raw material ratio scheme. Obtain the actual evaluation data corresponding to the target raw material ratio scheme, update the correlation between raw material characteristics based on the actual evaluation data, and optimize the subsequent meal replacement powder raw material ratio based on the updated correlation.

Images