Digital control method and system based on cosmetic production characteristics

Abstract

The invention provides a digital control method and system based on cosmetic production characteristics, and belongs to the technical field of data processing.The method comprises the steps that firstly, a digital twinborn model of cosmetic product characteristics is built, and the model is related to formula data, technological parameters, sensory characteristics and physicochemical indexes of a product; and output sensory characteristics and physicochemical indexes can be predicted according to input formula data and process parameters. Then, order demands at least containing one or more sensory characteristic preferences of product texture, skin feeling and fragrance are obtained; then, one or more recommended production parameter sets are determined based on the recommended production parameter set order requirements and the recommended production parameter set digital twin model, the recommended production parameter sets defining corresponding recipe data, process parameters and sensory characteristics. And finally, controlling production equipment to execute production according to the recommended production parameter set. According to the invention, accurate control from personalized sensory requirements to production parameters is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, and in particular to a digital control method and system based on cosmetic production characteristics. BACKGROUND

[0002] Traditional cosmetic production control methods rely heavily on fixed formulations and the experience of operators. Process parameters in the production process are usually preset based on historical experience, while the core quality of the final product, especially the sensory characteristics such as texture and skin feel, is mainly determined by manual sensory evaluation after production is completed.

[0003] With the increasing demand for personalized cosmetics from consumers, the traditional production system based on fixed formulations cannot quickly and economically realize small-batch and customized production, and it is difficult to directly convert abstract sensory needs such as "refreshing feeling" and "moisturizing degree" into precise production instructions. Secondly, the production quality stability is insufficient and excessively dependent on personal experience. The production effect is significantly affected by the subjective judgment of the operator, and there may be differences in sensory experience between different batches of products, resulting in quality fluctuations. SUMMARY

[0004] The embodiments of the present application provide a digital control method and system based on cosmetic production characteristics to improve the above problems.

[0005] To achieve the above purpose, the technical scheme adopted by the present application is as follows: In a first aspect, the present application provides a digital control method based on cosmetic production characteristics, which is suitable for a digital control system based on cosmetic production characteristics, the system comprising a controller and a production device, the method being executed by the controller and comprising: constructing a digital twin model of cosmetic product characteristics, wherein the digital twin model is associated with the formulation data, process parameters, sensory characteristics and physicochemical indexes of the product, and the digital twin model is used to output predicted sensory characteristics and physicochemical indexes according to input formulation data and process parameters; obtaining order requirements, wherein the order requirements at least include one or more sensory characteristic preferences for product texture, skin feel and fragrance type; determining one or more recommended production parameter sets based on the order requirements and the digital twin model, wherein the recommended production parameter sets are used to determine the corresponding formulation data, process parameters and sensory characteristics in the production process; controlling the production device to execute production based on the recommended production parameter sets.

[0006] In combination with the first aspect, optionally, a digital twin model of the cosmetic product features is constructed, wherein the digital twin model is associated with the formula data, process parameters, sensory properties, and physicochemical indexes of the product, and the digital twin model is used to output predicted sensory properties and physicochemical indexes according to input formula data and process parameters, including: a product feature knowledge graph is constructed, wherein the nodes of the product feature knowledge graph at least include raw materials, process units, sensory attributes, and physicochemical indexes; historical production data is obtained, and the association relationship between the nodes of the product feature knowledge graph is determined based on the historical production data; a multi-level prediction model is constructed based on the product feature knowledge graph, wherein a first layer model of the multi-level prediction model is used to predict the state of an intermediate product according to formula data and process parameters, and a second layer model of the multi-level prediction model is used to predict sensory attributes and physicochemical indexes according to the state of the intermediate product; the product feature knowledge graph and the multi-level prediction model are data-coupled to form the digital twin model.

[0007] In combination with the first aspect, optionally, one or more recommended production parameter sets are determined based on order demand and the digital twin model, wherein the recommended production parameter sets are used to determine corresponding formula data, process parameters, and sensory properties in the production process, including: the sensory properties are converted into a target sensory property vector; the target sensory property vector and the preset physicochemical indexes are taken as optimization objectives, and an iterative search is performed in the formula data and the process parameters based on a multi-objective optimization algorithm; one or more formula data and recommended production parameter sets matched therewith that simultaneously satisfy the optimization objectives are determined based on iterative forward prediction.

[0008] In combination with the first aspect, optionally, historical production data is obtained, and the association relationship between the nodes of the product feature knowledge graph is determined based on the historical production data, including: the historical production data is preprocessed, feature data related to each node is extracted, and a structured data set is formed; an association rule mining algorithm and / or a causal inference model are applied to obtain analysis results corresponding to statistical dependency relationships and potential causal directions between different node features in the structured data set; according to the analysis results, association edges between the nodes in the product feature knowledge graph are established, and each association edge is given a relationship type and / or a relationship strength weight.

[0009] In combination with the first aspect, optionally, an association rule mining algorithm and / or a causal inference model are applied to obtain analysis results corresponding to statistical dependency relationships and potential causal directions between different node features in the structured data set, including: traverse and combine the node features in the structured dataset, determine the frequent co-occurrence or co-variable rule of the feature combination, and determine the support and confidence of the feature combination; determine the set of association rules with dependency relationship from the feature combination based on the preset confidence threshold and support threshold; based on the determined association rules, determine the order and direction of the influence between the node features to determine the causal direction; output the set of association rules and the causal direction as the analysis result.

[0010] In combination with the first aspect, optionally, a product feature knowledge graph is constructed, wherein the nodes of the product feature knowledge graph at least include raw materials, process units, sensory attributes, and physicochemical indicators, including: define the node entities of the knowledge graph, wherein: the raw material node contains the chemical composition, physical properties, and functional category attributes of the raw material; the process unit node contains the unit operation type, equipment parameter range, and input and output material state attributes; the sensory attribute node contains the texture, skin feel, and appearance corresponding score vector; the physicochemical indicator node contains the viscosity, pH value, particle size distribution, and stability test results of the product.

[0011] In combination with the first aspect, optionally, the product feature knowledge graph is data-coupled with a multi-level prediction model to form a digital twin model, including: based on the node entities and association relationships in the product feature knowledge graph, determine the feature vector; input the feature vector as the prior constraint condition and feature of the multi-level prediction model, and fuse it with the corresponding numerical features in the historical production data, and input it into the multi-level prediction model for training; map the output result of the multi-level prediction model back to the product feature knowledge graph as the predicted value of the corresponding node.

[0012] In the second aspect, the present application proposes a digital control system based on the production characteristics of cosmetics, which includes a controller and production equipment, and is configured to: construct a digital twin model of cosmetic product features, wherein the digital twin model is associated with the formula data, process parameters, sensory characteristics, and physicochemical indicators of the product, and is used to output the predicted sensory characteristics and physicochemical indicators according to the input formula data and process parameters; obtain order requirements, wherein the order requirements at least include one or more sensory characteristic preferences for product texture, skin feel, and fragrance type; determine one or more recommended production parameter sets based on the order demand and the digital twin model, wherein the recommended production parameter sets are used to determine corresponding formula data, process parameters, and sensory properties in the production process; control the production equipment to perform production based on the recommended production parameter sets.

[0013] In combination with the second aspect, the system is optionally configured to: build a digital twin model of the cosmetic product features, wherein the digital twin model associates formula data, process parameters, sensory properties, and physicochemical indexes of the product, and the digital twin model is used to output predicted sensory properties and physicochemical indexes according to input formula data and process parameters, including: build a product feature knowledge graph, wherein nodes of the product feature knowledge graph at least include raw materials, process units, sensory properties, and physicochemical indexes; obtain historical production data and determine the association between the nodes of the product feature knowledge graph based on the historical production data; build a multi-level prediction model based on the product feature knowledge graph, wherein a first layer model of the multi-level prediction model is used to predict intermediate product states according to formula data and process parameters, and a second layer model of the multi-level prediction model is used to predict sensory properties and physicochemical indexes according to the intermediate product states; couple the product feature knowledge graph and the multi-level prediction model to form the digital twin model.

[0014] In combination with the second aspect, the system is optionally configured to: determine one or more recommended production parameter sets based on the order demand and the digital twin model, wherein the recommended production parameter sets are used to determine corresponding formula data, process parameters, and sensory properties in the production process, including: convert the sensory properties into a target sensory property vector; take the target sensory property vector and the preset physicochemical indexes as optimization objectives, and perform iterative search in the formula data and the process parameters based on a multi-objective optimization algorithm; determine one or more formula data and matching recommended production parameter sets that simultaneously satisfy the optimization objectives based on iterative forward prediction.

[0015] In combination with the second aspect, the system is optionally configured to: obtain historical production data and determine the association between the nodes of the product feature knowledge graph based on the historical production data, including: preprocess the historical production data, extract feature data related to each node, and form a structured data set; The association rule mining algorithm and / or the causal inference model are applied to obtain analysis results corresponding to statistical dependency relationships and potential causal directions between different node features in the structured data set. According to the analysis results, association edges between nodes are established in the product feature knowledge graph, and each association edge is assigned a relationship type and / or a relationship strength weight.

[0016] In combination with the second aspect, the system is optionally configured to: The association rule mining algorithm and / or the causal inference model are applied to obtain analysis results corresponding to statistical dependency relationships and potential causal directions between different node features in the structured data set, including: The node features in the structured data set are traversed and combined to determine feature combinations that frequently co-occur or have a co-variance rule, and to determine the support and confidence of the feature combinations; Based on the preset confidence threshold and support threshold, a set of association rules having dependency relationships is determined from the feature combinations; Based on the determined association rules, the order and direction of the influence between the node features are determined to determine the causal direction; The set of association rules and the causal direction are output as analysis results.

[0017] In combination with the second aspect, the system is optionally configured to: A product feature knowledge graph is constructed, wherein the nodes of the product feature knowledge graph at least include raw materials, process units, sensory attributes, and physicochemical indicators, including: The node entities of the knowledge graph are defined, wherein: The raw material node includes chemical composition, physical properties, and functional category attributes of the raw material; The process unit node includes unit operation type, equipment parameter range, and input and output material state attributes; The sensory attribute node includes a score vector corresponding to texture, skin feel, and appearance; The physicochemical indicator node includes viscosity, pH value, particle size distribution, and stability test results of the product.

[0018] In combination with the second aspect, the system is optionally configured to: The product feature knowledge graph is coupled with a multi-level prediction model to form a digital twin model, including: Based on the node entities and association relationships in the product feature knowledge graph, a feature vector is determined; The feature vector is taken as a prior constraint condition and feature input of the multi-level prediction model, and is fused with corresponding numerical features in historical production data, and is input into the multi-level prediction model for training; The output result of the multi-level prediction model is mapped back to the product feature knowledge graph as a predicted value of a corresponding node.

[0019] The third aspect of the embodiment of the present application provides an electronic device, and the electronic device comprises: at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to the first aspect of the embodiment of the present application.

[0020] The fourth aspect of the embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the method according to the first aspect of the embodiment of the present application.

[0021] In summary, the above method and system have the following technical effects: The present application discloses a kind of digital control method and system based on cosmetic production characteristics, which first constructs the digital twin model of cosmetic product features, the model is associated with the formula data, process parameters, sensory characteristics and physicochemical indexes of product, can be according to the formula data and process parameters of input predicted output sensory characteristics and physicochemical indexes.It is then obtained at least containing one or more sensory characteristics preferences of order demand such as product texture, skin feeling, fragrance type.Then, based on order demand and digital twin model, determine one or more recommended production parameter set, parameter set defines corresponding formula data, process parameters and sensory characteristics.Finally, according to recommended production parameter set control production equipment executes production.The present application realizes from individualized sensory demand to the accurate control of production parameter. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 A flowchart of a kind of digital control method based on cosmetic production characteristics is provided for the embodiment of the present application.

[0023] Figure 2 A flowchart of the specific composition of the digital twin model of cosmetic product features constructed in the embodiment of the present application is provided. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments of the present application.Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0025] The embodiment of the present application provides a digital control method based on cosmetic production characteristics, which is suitable for a digital control system based on cosmetic production characteristics. The system comprises a controller and production equipment, and the method is executed by the controller. Please refer to Figure 1 The method comprises the following steps: S101: Constructing a digital twin model of cosmetic product characteristics, wherein the digital twin model is associated with formula data, process parameters, sensory characteristics and physicochemical indexes of a product, and the digital twin model is used for outputting predicted sensory characteristics and physicochemical indexes according to input formula data and process parameters.

[0026] It can be understood that the historical and real-time data in cosmetic production. These data mainly include two types: one is input data, that is, accurate raw material formula proportioning and specific production process parameters; the other is output data, that is, detection results of various physical and chemical indexes corresponding to the final product, and quantitatively processed product sensory evaluation data. Based on the collected data, a calculable mathematical correlation can be established between the input variables and the output variables.

[0027] This process is usually realized by constructing and training a specific prediction model. The model can learn and internalize the mapping relationship from specific formula and process conditions to the physicochemical indexes and sensory characteristics of the final product.

[0028] Exemplarily, Specifically, as an implementation manner, step S101 can comprise: S1011: Constructing a product characteristic knowledge graph, wherein nodes of the product characteristic knowledge graph at least include raw materials, process units, sensory attributes and physicochemical indexes.

[0029] Specifically, the product characteristic knowledge graph refers to defining and organizing core concepts and their mutual relationships in the field of cosmetic production in a structured network form.

[0030] Exemplarily, the raw material node can contain chemical composition, physical properties and functional category attributes of the raw material, the process unit node can contain unit operation type, equipment parameter range and input and output material state attributes, the sensory attribute node can contain score vectors corresponding to texture, skin feeling and appearance, and the physicochemical index node can contain viscosity, pH value, particle size distribution and stability test results of the product.

[0031] In the present embodiment, the basic constituting units in the knowledge graph, i.e. the nodes, need to be defined first. These nodes are set into four basic categories: raw materials representing all the ingredients used, process units representing specific operations in the production process, sensory attributes describing the subjective experience characteristics of the product, and physicochemical indicators indicating the objective measurement values of the product. Each node category corresponds to a key dimension in the production knowledge system. For example, a raw material node can contain the chemical composition, physical properties and functional category attributes of the raw material, a process unit node can contain the unit operation type, equipment parameter range and input and output material state attributes, a sensory attribute node can contain the score vector corresponding to texture, skin feel and appearance, and a physicochemical indicator node can contain the viscosity, pH value, particle size distribution and stability test results of the product.

[0032] For example, a specific raw material node records its chemical name, molecular formula, physical state, supplier information and common functional role in the formula. A process unit node may contain the equipment type it belongs to, the adjustable parameter range of the equipment, and the expected input and output material state of the unit operation. The sensory attribute node needs to be associated with the descriptive scores quantified by standard methods, such as numerical scales from "refreshing" to "sticky". The physicochemical indicator node is explicitly directed to specific detection items and standard units, such as viscosity value, pH value or concentration of a specific ingredient. Therefore, the final knowledge graph is a semantic network composed of these nodes carrying specific attributes and connection edges representing the logical and causal relationships between nodes established in subsequent steps.

[0033] S1012: Obtain historical production data, and determine the association relationship between the nodes of the product feature knowledge graph based on the historical production data.

[0034] For example, the historical production data can be pre-processed to extract feature data related to each node to form a structured data set. It can be understood that historical data can be collected from production records, laboratory test reports and sensory evaluation archives. These data usually contain complete information chain corresponding to each production batch, such as specific raw materials used and their accurate proportioning, process parameters such as temperature, time, stirring speed recorded in each production process, and physicochemical indicators such as viscosity, pH value of the final product of the batch, and sensory characteristic description or score evaluated by professionals.

[0035] Subsequently, these multi-source heterogeneous historical data are cleaned, aligned and structured to enable mapping with the node categories defined in the knowledge graph. For example, the raw material code in the database is mapped to the raw material node in the graph, and the steps in the operation order are mapped to the process unit node.

[0036] By analyzing the co-occurrence patterns, statistical correlations and change trends between data entries, the potential relationships between nodes of different categories are identified and inferred. For example, by analyzing a large amount of historical batch data, it may be found that when the amount of raw material A increases, the score of product on sensory attribute B shows a regular upward trend; or when the control parameter of process unit C reaches a certain range, the measured value of physicochemical index D will be stable in a certain interval. These data-verified association relationships are added as connection lines to the initially constructed knowledge graph framework, thereby connecting independent nodes into a semantic network containing rich production experience and causal logic. For example, association rule mining algorithms and / or causal inference models can be applied to obtain the analysis results corresponding to the statistical dependence relationships and potential causal directions between different node features in the structured data set.

[0037] In particular, as an implementation, please refer to Figure 2 may include the following steps: S201: Traverse and combine the node features in the structured data set, determine the feature combinations that frequently co-occur or have co-variation rules, and determine the support and confidence of the feature combinations.

[0038] As can be understood, by combining and comparing all defined node features in the data set two by two or in multiple elements, feature pairs or feature sets that repeatedly appear together in history can be found. For example, it may find that raw material A and raw material B are present together in more than 80% of the essence liquid formulations; Or, it is observed that whenever the emulsification temperature feature value is in a certain specific interval, the measured value of product viscosity also stably falls within another corresponding interval.

[0039] These repeatedly appearing and statistically linked patterns are called frequent item sets or co-variation combinations.

[0040] For each discovered feature combination, the first is the support, which measures the universality of the pattern, and the calculation method is the proportion of the number of times the feature combination appears in all historical data samples to the total number of samples. The second is the confidence, which measures the reliability of the relationship between the features in the pattern, and the calculation method is the conditional probability that the remaining features in the combination also appear when one or part of the features in the combination appear.

[0041] S202: Based on the preset confidence threshold and support threshold, determine the set of association rules with dependent relationships from the feature combinations.

[0042] Specifically, the support value of each feature combination is compared with a preset lower limit of support, and combinations with too low frequency are filtered out because they lack universality and are not sufficient to constitute meaningful empirical rules. Then, the confidence of the combinations filtered by support is calculated, and the value is compared with a preset lower limit of confidence, further filtering out combinations that are frequently occurring but have unstable internal relationships, i.e., the conclusion feature does not always accompany the current feature when the current feature appears.

[0043] The feature combinations remaining after the two rounds of threshold filtering are considered to be statistically significant strong association patterns. For example, a confirmed rule can be expressed as: "If the formula contains raw materials A and B at the same time, and the emulsification temperature is set in the range of T1 to T2, then the particle size of the final product has a high probability of being less than D value". All such determined rules constitute an association rule set representing the dependence relationship between various elements in the production process.

[0044] S203: Based on the determined association rules, the order and direction of the influence between node features are determined to determine the causal direction.

[0045] It can be understood that, for example, if historical data shows that the addition of raw material X is strongly related to the increase of final viscosity, the system will check the production log: raw material X is always added in the early stage of the production process, and viscosity is always measured at the end of the process. This irreversibility in time provides evidence that the addition of raw material X is the cause of the increase of viscosity.

[0046] For complex cases that lack clear timing or have interactive effects, the system will use a causal discovery algorithm for analysis. Specifically, if there is a direct causal relationship between two variables A and B, the statistical dependence between them still exists even under the condition of controlling other variables; On the contrary, if the association is caused by other common causes, the dependence may disappear after controlling a specific variable. The system builds a network diagram representing the causal probability relationship by testing the conditional independence between different variable sets, thereby inferring the most likely causal direction between the features.

[0047] S204: The set of association rules and the causal direction are output as the analysis result.

[0048] It can be understood that the association rule set is a list composed of multiple "if-then" form rules, and each rule is attached with the specific support and confidence values calculated. These rules quantitatively describe the stable coexistence or covariation relationship between different production features. The causal direction determination is a further clarification of the interaction logic between the features in the above association rules, indicating which feature is the cause and which feature is the effect in the form of directed connection, or marking out complex cases such as bidirectional influence.

[0049] S205: According to the analysis result, the association edge between the nodes in the product feature knowledge graph is established, and the relationship type and / or relationship strength weight is assigned to each association edge.

[0050] Specifically, a directed or undirected connection line can be created between the corresponding nodes of the knowledge graph. For example, according to a rule, a connection edge is established between the "raw material A" node and the "sensory attribute B" node. At the same time, according to the causal direction analysis, the edge is marked as pointing from "raw material A" to "sensory attribute B".

[0051] Then, the quantitative indicators calculated by each association rule, such as support and confidence, are mapped to the weight attributes of the association edge. The higher the support, the more extensive the historical cases covered by the relationship; the higher the confidence, the stronger the possibility of the relationship. The weight value makes the relationship in the graph have a comparable and calculable strength dimension.

[0052] S1013: Based on the product feature knowledge graph, a multi-level prediction model is constructed, wherein a first layer model of the multi-level prediction model is used to predict the intermediate product state according to the formula data and process parameters, and a second layer model of the multi-level prediction model is used to predict the sensory attributes and physicochemical indexes according to the intermediate product state.

[0053] It can be understood that the first layer model receives specific formula composition data and detailed process parameter settings as input. Its design goal is to simulate and predict the state of the intermediate product at a certain key stage or node in the production process. These states are the observable or inferable intermediate attributes of the product before it is finally formed, such as the particle size distribution and rheological properties of the emulsion at the end of the emulsification stage, or the real-time temperature and viscosity changes of the material during the heating reaction process. The output of this layer model is a quantitative prediction of these intermediate results.

[0054] The second layer model takes the intermediate product state predicted by the first layer model as its core input. Its design goal is to further deduce and predict various characteristics of the final product based on these intermediate states. These characteristics include two aspects: one is the physical and chemical indexes that can be measured by instruments, such as the stability, pH value, and active ingredient content of the finished product; the other is the experience attributes that need to be evaluated by the senses, such as texture, skin feel, and aroma characteristics. The output of this layer model is a comprehensive prediction value of the quality of the final product.

[0055] Specifically, in the present embodiment, firstly, a feature vector can be determined based on the node entities and the associated relationships in the product feature knowledge graph. Illustratively, the graph embedding technology can be used to mathematically process each node entity and each associated relationship in the knowledge graph. The goal of this process is to generate a fixed-length, dense numerical array, i.e., a feature vector, for each node and relationship.

[0056] Then, the feature vector is input as a prior constraint condition of the multi-level prediction model and is fused with the corresponding numerical features in the historical production data, and is input into the multi-level prediction model for training. Specifically, training data can be prepared from two aspects: one is the node and relationship features extracted from the knowledge graph and vectorized; and the other is the traditional numerical features corresponding to the graph nodes extracted from the historical production records, such as the specific amount of raw materials, the accurate numerical value of the temperature.

[0057] Then, the knowledge graph feature vector and the corresponding numerical feature vector can be connected into a longer comprehensive feature vector. Of course, in some other embodiments, different weights can be dynamically assigned to the two types of features before being fused, so that the model can autonomously determine the importance of different information sources. In the model training stage, these fused feature vectors serve as input data for the model. At the same time, the associated logic defined in the knowledge graph is converted into a constraint condition for model training. Illustratively, a special regularization term can be added to the loss function of the model to achieve this. The regularization term calculates the degree of consistency between the model's prediction results and the known logical relationships in the knowledge graph, and penalizes predictions that obviously violate the graph logic.

[0058] For example, if the knowledge graph clearly states that "raw material A will increase the viscosity", but the model learns the opposite prediction trend, then the regularization term will generate a larger penalty value, prompting the model to adjust the parameters to conform to this prior knowledge. Finally, the output results of the multi-level prediction model can be mapped back to the product feature knowledge graph as the predicted values of the corresponding nodes. When the multi-level prediction model completes a prediction, the system will analyze the output result data. These result data usually correspond to the predicted values of the sensory attributes or physicochemical indicators of a specific product formula. According to the preset mapping rules, these numerical prediction results are associated and matched with the corresponding nodes in the product feature knowledge graph. For example, the model's prediction of "viscosity value of 3500 mPa·s" will be matched to the "physicochemical indicator" node representing the specific product in the knowledge graph, and stored as an attribute named "model prediction value" of the node. Similarly, the predicted "refreshing degree score of 0.85" will be written into the corresponding attribute field of the corresponding "sensory attribute" node.

[0059] S1014: Coupling the product feature knowledge graph with the multi-level prediction model to form a digital twin model.

[0060] It can be understood that the feature vector is fused with the numerical features in the historical data as the input of the multi-level prediction model, so that the learning process thereof depends not only on the statistical law of the data, thereby improving the rationality and interpretability of the model prediction. After the multi-level prediction model performs simulation calculation on the new product scheme, the prediction result output by the model is automatically mapped and written back to the corresponding node in the knowledge graph, and is stored as the prediction attribute of the node.

[0061] S102: Obtaining an order demand, wherein the order demand at least includes a preference for one or more sensory characteristics of product texture, skin feel, and fragrance type.

[0062] Specifically, the system can obtain structured order data. In addition to the information such as product type and quantity, the order data must include the requirements for the sensory experience of the final product. These requirements usually focus on one or more key sensory dimensions, such as texture, skin feel, and fragrance type. These descriptions can be in the form of natural language, rating scores, or selection labels, which are not limited in the present application.

[0063] After obtaining the original requirement description, the system can convert these sensory preferences with subjective color into a data format that can be recognized and processed by the subsequent digital twin model. For example, converting “light texture” into a quantitative indicator pointing to a high score on the pre-defined “texture” dimension, or mapping a series of fragrance keywords to a specific fragrance feature vector.

[0064] S103: Determining one or more recommended production parameter sets based on the order demand and the digital twin model, wherein the recommended production parameter sets are used to determine the corresponding formula data, process parameters, and sensory characteristics in the production process.

[0065] It can be understood that the explicit sensory characteristic preferences in the order are first converted into quantifiable target vectors that can be recognized by the digital twin model. For example, “extremely refreshing skin feel” is set as the highest score to be achieved on a certain sensory dimension.

[0066] Subsequently, an automated iterative search process is started in the virtual production environment constructed by the digital twin model, with the target vector and the core physicochemical indicators that the product must meet as the optimization objectives. This process explores all possible combinations of formula raw material ratios and process parameter settings. In each iteration, the system calls the digital twin model to perform a virtual production simulation of the current formula and process combination, quickly predicting the sensory characteristics and physicochemical indicators of the product under this combination. The system compares these predicted results with the set optimization objectives, evaluates their degree of conformity, and automatically adjusts the search direction to find better solutions.

[0067] It can be understood that after multiple rounds of iterative simulation and optimization calculation, the system will finally output one or more optimal or near-optimal solutions. Each solution is a recommended production parameter set, which contains the exact formula composition, detailed process parameters at each step, and expected sensory characteristic results that should be used to achieve the order requirements. These parameter sets provide clear and executable digital work orders for actual production.

[0068] S104: Control the production equipment to perform production based on the recommended production parameter set.

[0069] The present application discloses a digital control method based on the production characteristics of cosmetics. The method first constructs a digital twin model of cosmetic product characteristics. The model is associated with product formula data, process parameters, sensory characteristics, and physicochemical indicators, and can predict the output of sensory characteristics and physicochemical indicators based on input formula data and process parameters. Next, the order requirements including one or more sensory characteristics preferences of product texture, skin feel, and fragrance are obtained. Then, based on the order requirements and the digital twin model, one or more recommended production parameter sets are determined, which define the corresponding formula data, process parameters, and sensory characteristics. Finally, the production equipment is controlled to perform production according to the recommended production parameter set. The present application realizes precise control from personalized sensory requirements to production parameters.

[0070] Based on the same inventive concept, the embodiments of the present application also propose a digital control system based on the production characteristics of cosmetics. The system includes a controller and production equipment, and is configured to: construct a digital twin model of cosmetic product characteristics, wherein the digital twin model is associated with product formula data, process parameters, sensory characteristics, and physicochemical indicators, and is used to output predicted sensory characteristics and physicochemical indicators based on input formula data and process parameters; obtain order requirements, wherein the order requirements include one or more sensory characteristics preferences of product texture, skin feel, and fragrance; determine one or more recommended production parameter sets based on the order demand and the digital twin model, wherein the recommended production parameter sets are used to determine corresponding formula data, process parameters, and sensory properties in the production process; control the production equipment to perform production based on the recommended production parameter sets.

[0071] Optionally, the system is configured to: build a digital twin model of the cosmetic product features, wherein the digital twin model associates formula data, process parameters, sensory properties, and physicochemical indicators of the product, and the digital twin model is used to output predicted sensory properties and physicochemical indicators based on input formula data and process parameters, including: build a product feature knowledge graph, wherein the nodes of the product feature knowledge graph include at least raw materials, process units, sensory attributes, and physicochemical indicators; obtain historical production data and determine the association between the nodes of the product feature knowledge graph based on the historical production data; based on the product feature knowledge graph, build a multi-level prediction model, wherein the first layer model of the multi-level prediction model is used to predict the state of intermediate products based on formula data and process parameters, and the second layer model of the multi-level prediction model is used to predict sensory attributes and physicochemical indicators based on the state of intermediate products; couple the product feature knowledge graph and the multi-level prediction model to form a digital twin model.

[0072] Optionally, the system is configured to: determine one or more recommended production parameter sets based on the order demand and the digital twin model, wherein the recommended production parameter sets are used to determine corresponding formula data, process parameters, and sensory properties in the production process, including: convert the sensory properties into a target sensory property vector; take the target sensory property vector and the preset physicochemical indicators as optimization objectives, and perform iterative search in the formula data and process parameters based on a multi-objective optimization algorithm; based on the iterative forward prediction, determine one or more formula data that simultaneously satisfy the optimization objectives and the recommended production parameter sets matched therewith.

[0073] Optionally, the system is configured to: obtain historical production data and determine the association between the nodes of the product feature knowledge graph based on the historical production data, including: preprocess the historical production data, extract feature data related to each node, and form a structured data set; apply an association rule mining algorithm and / or a causal inference model to obtain analysis results corresponding to the statistical dependency relationships and potential causal directions between different node features in the structured data set; According to the analysis result, the association edges between the nodes in the product feature knowledge graph are established, and the relationship type and / or relationship strength weight are assigned to each association edge.

[0074] Optionally, the system is configured to: apply the association rule mining algorithm and / or the causal inference model to obtain the analysis result corresponding to the statistical dependency relationship and the potential causal direction between different node features in the structured data set, including: traverse and combine the node features in the structured data set, determine the feature combination that frequently appears together or has a co-variation rule, and determine the support and confidence of the feature combination; determine the set of association rules with dependency relationships from the feature combination based on the preset confidence threshold and support threshold; determine the sequence and direction of the influence between the node features based on the determined association rules, to determine the causal direction; output the set of association rules and the causal direction as the analysis result.

[0075] Optionally, the system is configured to: construct a product feature knowledge graph, wherein the nodes of the product feature knowledge graph at least include raw materials, process units, sensory attributes, and physicochemical indicators, including: define the node entities of the knowledge graph, wherein: the raw material node contains chemical composition, physical properties, and functional category attributes of the raw material; the process unit node contains unit operation type, equipment parameter range, and input and output material state attributes; the sensory attribute node contains a score vector corresponding to texture, skin feel, and appearance; the physicochemical indicator node contains viscosity, pH value, particle size distribution, and stability test results of the product.

[0076] Optionally, the system is configured to: couple the product feature knowledge graph and the multi-level prediction model for data to form a digital twin model, including: determine the feature vector based on the node entities and association relationships in the product feature knowledge graph; input the feature vector as the prior constraint condition and feature of the multi-level prediction model, and fuse it with the corresponding numerical features in the historical production data, and input the multi-level prediction model for training; map the output result of the multi-level prediction model back to the product feature knowledge graph as the predicted value of the corresponding node.

[0077] The application discloses a digital control system based on cosmetic production characteristics, which comprises the following steps: firstly, a digital twin model of cosmetic product characteristics is constructed, the model is associated with product formula data, process parameters, sensory characteristics and physicochemical indexes, and the output sensory characteristics and physicochemical indexes can be predicted according to the input formula data and process parameters; secondly, order requirements of at least one or more sensory characteristics, such as product texture, skin feeling and fragrance type, are obtained; thirdly, one or more recommended production parameter sets are determined based on the order requirements and the digital twin model, the parameter sets define corresponding formula data, process parameters and sensory characteristics; and finally, the production equipment is controlled to perform production according to the recommended production parameter set. The application realizes accurate control from personalized sensory requirements to production parameters.

[0078] Based on the same inventive concept, the embodiments of the present application also provide an electronic device, which comprises: at least one processor; and a memory connected with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the automatic overheat protection method based on the universal testing machine.

[0079] In addition, to achieve the above object, the embodiments of the present application also provide a computer readable storage medium storing a computer program, and the computer program is executed by a processor to implement the automatic overheat protection method based on the universal testing machine.

[0080] The various constituent components of the electronic device will be described in detail as follows: The processor is the control center of the electronic device, which can be one processor or a plurality of processing elements. For example, the processor is one or more central processing units (CPU), application specific integrated circuits (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application, such as one or more microprocessors (digital signal processor, DSP), or one or more field programmable gate arrays (FPGA).

[0081] Optionally, the processor can execute various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0082] The memory is configured to store a software program for implementing the scheme of the application, and the processor is configured to control the execution. The specific implementation can refer to the method embodiments described above, and will not be described here.

[0083] Alternatively, the memory can be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disk storage, an optical disk storage (including a compact disk, a laser disk, an optical disk, a digital versatile disk, a Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but not limited to. The memory can be integrated with the processor or exist independently and be coupled to the processor through the interface circuit of the electronic device, and the embodiments of the application do not make specific limitations in this regard.

[0084] The transceiver is configured to communicate with the network device or the terminal device.

[0085] Alternatively, the transceiver can include a receiver and a transmitter. The receiver is configured to implement the receiving function, and the transmitter is configured to implement the transmitting function.

[0086] Alternatively, the transceiver can be integrated with the processor or exist independently and be coupled to the processor through the interface circuit of the router, and the embodiments of the application do not make specific limitations in this regard.

[0087] In addition, the technical effects of the electronic device can refer to the technical effects of the data transmission method of the method embodiments described above, and will not be described here.

[0088] It should be appreciated that a processor in the embodiments of the present application can be a central processing unit (CPU). The processor can also be other general purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic, discrete hardware components, etc. The general purpose processor can be a microprocessor or the processor can be any conventional processor.

[0089] It should also be appreciated that the memory in the embodiments of the present application can be a volatile memory or a nonvolatile memory, or can include both volatile and nonvolatile memory. The nonvolatile memory can be a read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), electrically EPROM (EEPROM), or flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example, and not limitation, many forms of random access memory (RAM) are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0090] The above-described embodiments can be implemented in whole or in part by software, hardware (e.g., circuitry), firmware, or any combination thereof. When implemented in software, the above-described embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions according to the embodiments of the present application are entirely or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center through wired (e.g., infrared, wireless, microwave, etc.) or wireless means. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0091] It should be understood that the term "and / or" herein merely describes an association relationship of associated objects, which means that there can be three relationships, for example, A and / or B can represent three cases of A alone, A and B together, and B alone, where A and B can be singular or plural. In addition, the character " / " herein generally represents an "or" relationship between the associated objects before and after it, but it can also represent an "and / or" relationship, which can be understood in the context before and after it.

[0092] In the present application, "at least one" means one or more, and "multiple" means two or more. "At least one of the following" or the like means any combination of the items, including any combination of single item or multiple items. For example, at least one of a, b, or c can represent a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, and c can be single or multiple.

[0093] It should be understood that in various embodiments of the present application, the size of the sequence number of the above-described processes does not mean the order of execution, and the execution order of the processes should be determined by their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0094] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on specific applications and design constraints. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

Claims

1. A digital control method based on the characteristics of cosmetic production, characterized in that, A digital control system applicable to cosmetic production characteristics, the system comprising a controller and production equipment, the method being executed by the controller, comprising: A digital twin model of cosmetic product characteristics is constructed, wherein the digital twin model is associated with the product's formula data, process parameters, sensory characteristics, and physicochemical indicators, and the digital twin model is used to output predicted sensory characteristics and physicochemical indicators based on the input formula data and process parameters. Obtain order requirements, wherein the order requirements include at least a preference for one or more sensory characteristics of the product, such as texture, feel, and fragrance; Based on the order requirements and the digital twin model, one or more recommended production parameter sets are determined, wherein the recommended production parameter sets are used to determine the corresponding formula data, process parameters, and sensory characteristics during the production process; The production equipment is controlled to perform production based on the recommended set of production parameters.

2. The digital control method based on the characteristics of cosmetic production according to claim 1, characterized in that, A digital twin model of cosmetic product characteristics is constructed, wherein the digital twin model is associated with the product's formula data, process parameters, sensory characteristics, and physicochemical indicators. The digital twin model is used to output predicted sensory characteristics and physicochemical indicators based on the input formula data and process parameters, including: Construct a product feature knowledge graph, wherein the nodes of the product feature knowledge graph include at least raw materials, process units, sensory attributes, and physicochemical indicators; Acquire historical production data, and determine the relationships between the nodes of the product feature knowledge graph based on the historical production data; Based on the product feature knowledge graph, a multi-level prediction model is constructed. The first layer of the multi-level prediction model is used to predict the state of intermediate products based on the formula data and the process parameters. The second layer of the multi-level prediction model is used to predict the sensory attributes and physicochemical indicators based on the state of intermediate products. The product feature knowledge graph is coupled with the multi-level prediction model to form the digital twin model.

3. The digital control method based on the characteristics of cosmetic production according to claim 2, characterized in that, Based on the order demand and the digital twin model, one or more recommended production parameter sets are determined, wherein the recommended production parameter sets are used to determine the corresponding formula data, process parameters, and sensory characteristics during the production process, including: The sensory characteristics are transformed into a target sensory characteristic vector; Using the target sensory characteristic vector and preset physicochemical indicators as optimization objectives, an iterative search is performed on the formula data and the process parameters based on a multi-objective optimization algorithm; Based on iterative forward prediction, one or more of the formula data that simultaneously satisfy the optimization objective and the matching set of recommended production parameters are determined.

4. The digital control method based on the characteristics of cosmetic production according to claim 2, characterized in that, Acquiring historical production data and determining the relationships between the nodes of the product feature knowledge graph based on the historical production data includes: The historical production data is preprocessed to extract feature data related to each node, forming a structured dataset; By applying association rule mining algorithms and / or causal inference models, the statistical dependencies and potential causal directions between different node features in the structured dataset are analyzed. Based on the analysis results, connections between nodes are established in the product feature knowledge graph, and each connection is assigned a relation type and / or relation strength weight.

5. The digital control method based on the characteristics of cosmetic production according to claim 4, characterized in that, By applying association rule mining algorithms and / or causal inference models, the analytical results corresponding to the statistical dependencies and potential causal directions among different node features in the structured dataset are obtained, including: The node features in the structured dataset are traversed and combined to determine feature combinations that frequently co-occur or have co-variation patterns, and the support and confidence of the feature combinations are determined. Based on preset confidence and support thresholds, a set of association rules with dependencies is determined from the feature combinations. Based on the determined association rules, the order and direction of influence between node features are determined to determine the causal direction; The set of association rules and the causal direction are output as the analysis results.

6. The digital control method based on the characteristics of cosmetic production according to claim 2, characterized in that, Construct a product feature knowledge graph, wherein the nodes of the product feature knowledge graph include at least raw materials, process units, sensory attributes, and physicochemical indicators, including: Define the node entities of the knowledge graph, where: The raw material node includes the chemical composition, physical properties, and functional category attributes of the raw material; The process unit node includes the unit operation type, equipment parameter range, and input / output material status attributes; The sensory attribute nodes include scoring vectors corresponding to texture, skin feel, and appearance; The physicochemical index nodes include the product's viscosity, pH value, particle size distribution, and stability test results.

7. The digital control method based on the characteristics of cosmetic production according to claim 2, characterized in that, The digital twin model is formed by coupling the product feature knowledge graph with the multi-level prediction model, including: Based on the node entities and their associations in the product feature knowledge graph, feature vectors are determined; The feature vector is used as the prior constraint and feature input of the multi-level prediction model, and is fused with the corresponding numerical features in the historical production data, and then input into the multi-level prediction model for training. The output of the multi-level prediction model is mapped back to the product feature knowledge graph and used as the predicted value of the corresponding node.

8. A digital control system based on the characteristics of cosmetic production, characterized in that, The system includes a controller and production equipment, and the system is configured as follows: A digital twin model of cosmetic product characteristics is constructed, wherein the digital twin model is associated with the product's formula data, process parameters, sensory characteristics, and physicochemical indicators, and the digital twin model is used to output predicted sensory characteristics and physicochemical indicators based on the input formula data and process parameters. Obtain order requirements, wherein the order requirements include at least a preference for one or more sensory characteristics of the product, such as texture, feel, and fragrance; Based on the order requirements and the digital twin model, one or more recommended production parameter sets are determined, wherein the recommended production parameter sets are used to determine the corresponding formula data, process parameters, and sensory characteristics during the production process; The production equipment is controlled to perform production based on the recommended set of production parameters.

9. An electronic device, characterized in that, include: At least one processor; And, a memory communicatively connected to at least one of the processors; The memory stores instructions that can be executed by at least one of the processors, which are executed by at least one of the processors to enable at least one of the processors to perform a digital control method based on cosmetic production characteristics as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements a digital control method based on the characteristics of cosmetic production as described in any one of claims 1-7.

Images