Multi-dimensional rapid detection and comprehensive analysis system for skin care product formula

Abstract

The invention relates to the technical field of skin care product detection, and discloses a multi-dimensional rapid detection and comprehensive analysis system for a skin care product formula. The method comprises the following steps: acquiring infrared spectrum data of a skin care product through a component spectrum acquisition device, and extracting characteristic peak information to generate an original spectrum characteristic vector; the formula component analysis engine matches the standard spectral characteristics, identifies active components and concentration proportions and outputs a component identification list; the component interaction relation modeling module constructs a component collaboration or antagonism relation network and generates a graph; the stability prediction unit analyzes component stability change trends in different environments and outputs a prediction report; the safety evaluation module calculates a safety threshold value and a risk level and generates an evaluation matrix; the formula optimization suggestion generator screens replaceable components or adjusts the concentration, and outputs optimization suggestions; and the detection report integration unit carries out structured integration on the result to form a final report. According to the system, multi-link integrated detection is realized, and the detection requirements of skin care products are met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of skin care product detection, in particular to a multi-dimensional rapid detection and comprehensive analysis system for skin care product formula. BACKGROUND

[0002] With the vigorous development of the cosmetics industry, the market size continues to expand, and the iteration speed of new ingredients and new formulas is accelerating, ingredient safety has become a core element of brand competition, and consumers' concern about product safety is also increasing. However, behind the rapid development of the industry, there are many safety challenges. Data shows that more than 60% of annual cosmetic adverse reaction reports are caused by ingredient problems, common allergenic ingredients include fragrances, preservatives, heavy metals, etc. Some products also have problems such as cross-contamination of raw materials, illegal addition, etc., which cause heavy metals to exceed the standard, threatening the health of consumers.

[0003] The current skin care product ingredient risk presents the characteristics of high incidence, concealment and complexity, and the risk sources cover multiple links such as raw material production, formula design, production and processing, storage and transportation. In raw material production, inferior raw materials and illegal addition occur from time to time; in formula design, enterprises blindly compound multiple active ingredients for the pursuit of efficacy, ignoring the interaction between ingredients, such as the skin irritation that may be enhanced after the compounding of multiple antioxidants; in the storage and transportation stage, changes in temperature and humidity will affect the stability of ingredients, leading to problems such as ineffective preservatives and microbial over-standard. More notably, some ingredients may decompose to produce toxic substances or react with other ingredients to generate irritating compounds under certain conditions, and such hidden risks are difficult to detect.

[0004] The existing detection scheme cannot meet the current risk prevention and control needs, and there are lagging, single and limited aspects in the technical aspect. The lagging is that the update speed of the detection method cannot keep up with the emergence of new ingredients, such as the lack of suitable detection methods for synthetic biology ingredients, leading to distorted evaluation results; the single is that it is mainly for the detection of known single ingredients, ignoring the combined risk brought by ingredient interaction and metabolites, such as the toxic substances that may be produced after the metabolism of sunscreen agents and emulsifiers after compounding; the limitation is that the detection sensitivity is insufficient, which cannot effectively identify trace impurities. In the standard system aspect, the detection methods in different regions are not unified, there is no detection standard for emerging ingredients, and there is a gap with international standards. In the implementation aspect, the detection cost is high, which is difficult for small and medium-sized enterprises to bear, the self-detection ability of enterprises is weak, the coverage of random inspection is limited, and the data sharing mechanism is not perfect, which further aggravates the market risk. SUMMARY

[0005] The purpose of the present application is to provide a multi-dimensional rapid detection and comprehensive analysis system for skin care product formula to solve the problems raised in the background.

[0006] In order to achieve the above object, the present application provides a multi-dimensional rapid detection and comprehensive analysis system for skin care product formula, which comprises:

[0007] The component spectrum acquisition device acquires infrared spectrum data of the skin care product, extracts spectrum characteristic peak position and intensity distribution, and generates an original spectrum characteristic vector;

[0008] The formula component analysis engine receives the original spectrum characteristic vector, matches standard spectrum characteristics in the preset component database, identifies active components and concentration ratios in the skin care product, and outputs a component identification list;

[0009] The component interaction relationship modeling module retrieves a chemical interaction knowledge base based on the component identification list, constructs a synergistic or antagonistic relationship network between components, and generates a component interaction relationship graph;

[0010] The stability prediction unit analyzes the component stability change trend under different environmental conditions according to the component interaction relationship graph, and outputs a stability prediction report;

[0011] The safety evaluation module combines the stability prediction report and a toxicology database to calculate the safety threshold and potential risk level of each component, and generates a safety evaluation matrix;

[0012] The formula optimization suggestion generator filters replaceable components or adjusts the concentration ratio based on the safety evaluation matrix and the component interaction relationship graph, and outputs an optimized formula suggestion table;

[0013] The detection report integration unit structurally integrates the component identification list, the stability prediction report, the safety evaluation matrix, and the optimized formula suggestion table to generate a final detection report document.

[0014] Preferably, the component spectrum acquisition device comprises:

[0015] The spectrum preprocessing submodule performs baseline correction and noise filtering processing on the original infrared spectrum data, and extracts the effective spectrum interval;

[0016] The characteristic peak extraction submodule identifies the characteristic peak position and half-peak width parameter in the effective spectrum interval, and marks the characteristic peak intensity distribution;

[0017] The vector encoding submodule converts the characteristic peak position, half-peak width parameter, and intensity distribution into a multi-dimensional numerical vector to generate an original spectrum characteristic vector.

[0018] Preferably, the formula component analysis engine comprises:

[0019] The standard characteristic matching submodule performs similarity calculation on the original spectrum characteristic vector and the standard spectrum characteristics in the preset component database to identify candidate components with a matching degree exceeding a threshold value;

[0020] The concentration inversion submodule calculates the actual concentration ratio of the candidate ingredients in the skin care product according to the characteristic peak intensity and the concentration calibration curve;

[0021] The list generation submodule integrates the candidate ingredients and their concentration ratios, sorts them by priority, and outputs an ingredient identification list.

[0022] Preferably, the ingredient interaction relationship modeling module includes:

[0023] The relationship retrieval submodule extracts the known interaction records of each ingredient in the ingredient identification list from the chemical interaction knowledge base;

[0024] The network construction submodule converts the interaction records into a directed graph structure, with nodes representing ingredients and edges representing synergistic or antagonistic directions and strengths;

[0025] The graph generation submodule performs topological sorting and weight normalization on the directed graph structure, and outputs an ingredient interaction relationship graph.

[0026] Preferably, the stability prediction unit includes:

[0027] The environmental parameter input interface receives temperature, humidity, and light intensity simulation conditions;

[0028] The change trend analysis submodule calculates the degradation rate and derivative generation probability of each ingredient under different environmental conditions based on the ingredient interaction relationship graph;

[0029] The report generation submodule integrates the degradation rate and derivative generation probability by environmental condition, and outputs a stability prediction report.

[0030] Preferably, the safety evaluation module includes:

[0031] The threshold query submodule obtains the skin irritation and sensitization safety thresholds of each ingredient in the ingredient identification list from the toxicology database;

[0032] The risk calculation submodule combines the derivative generation probability in the stability prediction report to evaluate the potential composite risk level;

[0033] The matrix generation submodule maps the safety thresholds and risk levels of each ingredient into a two-dimensional evaluation matrix to generate a safety evaluation matrix.

[0034] Preferably, the formula optimization suggestion generator includes:

[0035] The alternative ingredient screening submodule retrieves alternative ingredients with similar functions and higher safety thresholds from the preset ingredient database according to the high-risk ingredient identification in the safety evaluation matrix;

[0036] The proportion adjustment sub-module re-computes the recommended concentration proportion of the substitute ingredient based on the synergistic strength in the ingredient interaction relationship graph;

[0037] The suggestion table generation sub-module sorts the substitute ingredient and its recommended concentration proportion according to the optimization priority, and outputs an optimized formula suggestion table.

[0038] Preferably, the detection report integration unit comprises:

[0039] The data verification sub-module performs logical consistency checking on the ingredient identification list, the stability prediction report, the safety evaluation matrix and the optimized formula suggestion table;

[0040] The structured template sub-module maps and fills the verified data according to the field requirements of the preset report template;

[0041] The document generation sub-module converts the filled template into an editable document format, and generates a final detection report document.

[0042] Preferably, the system further comprises:

[0043] The user feedback interface receives modification opinions or supplementary data for the final detection report document;

[0044] The dynamic update engine adjusts the ingredient interaction relationship graph and the safety evaluation matrix according to the modification opinions or supplementary data, and re-generates the optimized formula suggestion table and the final detection report document.

[0045] Preferably, the dynamic update engine comprises:

[0046] The version control sub-module records the report version after each modification and the corresponding data change record;

[0047] The incremental learning sub-module inputs the user feedback data as a new sample into the formula ingredient analysis engine, and updates the standard spectrum feature matching rule.

[0048] Compared with the prior art, the present application has the following advantages:

[0049] The component spectrum acquisition device directly obtains the infrared spectrum data of the skin care product and extracts the characteristic peak information, and the pre-set ingredient database is combined to realize rapid identification of active ingredients and concentration proportion. Compared with the traditional detection method relying on complex pretreatment, the operation process is greatly simplified, the ingredient detection period is shortened, and relying on the fingerprint spectrum characteristics of infrared spectrum, complex ingredients such as natural plant extracts can be accurately identified, which is suitable for the current situation of increasing types of ingredients in skin care products.

[0050] The ingredient identification list output by the formula ingredient analysis engine provides a clear basis for subsequent analysis. The ingredient interaction relationship modeling module retrieves the chemical interaction knowledge base based on this, and constructs a relationship network and an interaction map that can intuitively present the synergistic or antagonistic effects between ingredients. This visualization of ingredient interactions breaks the limitations of traditional detection, which only focuses on single ingredients, and can discover hidden problems such as the reaction of plant extracts with preservatives to generate irritating compounds, making it impossible for the potential effects of ingredient compounding to escape.

[0051] The stability prediction unit analyzes the stability change trend under different environmental conditions based on the ingredient interaction relationship map, and can predict in advance the problems such as ingredient degradation and preservative failure caused by temperature and humidity changes during storage and transportation, providing reference for product storage condition formulation and shelf life determination, and reducing product unqualified conditions caused by supply chain problems.

[0052] The safety evaluation module combines the stability prediction report with the toxicology database to generate a safety evaluation matrix that clearly shows the safety threshold and potential risk level of each ingredient, helping enterprises and regulatory authorities quickly grasp the product safety status without relying on complex multiple rounds of detection to achieve risk identification, adapting to the needs of regulatory deepening and industry safety leading.

[0053] The formula optimization suggestion generator filters replaceable ingredients or adjusts the concentration ratio based on the evaluation matrix and interaction map, and outputs an optimization suggestion table that provides specific directions for enterprise formula improvement, avoiding the risks of blind compounding, while reducing the trial-and-error cost of enterprise formula adjustment, especially for small and medium-sized enterprises that lack professional testing capabilities.

[0054] The detection report integration unit structures and integrates various results to generate a final report covering multi-dimensional information such as ingredients, stability, safety, and optimization suggestions, achieving systematic presentation of detection results, meeting the needs of enterprise production control, regulatory department sampling, and consumer right-to-know protection in multiple scenarios, and promoting the transformation of skin care product detection from single ingredient analysis to whole-chain risk control. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 The working principle diagram of the multi-dimensional rapid detection and comprehensive analysis system for skin care product formulas described in the present application;

[0056] Figure 2 The flowchart for spectrum acquisition and feature extraction;

[0057] Figure 3 The flowchart for ingredient interaction relationship modeling. DETAILED DESCRIPTION

[0058] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the scope of protection of the present application.

[0059] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the scope of protection of the present application. Figure 1 The present application provides a multi-dimensional rapid detection and comprehensive analysis system for skin care product formula, which comprises: rapidly obtaining the component information of the skin care product by spectral technology, and performing in-depth analysis based on the interaction between components, and finally generating a comprehensive detection report containing formula analysis, stability prediction, safety evaluation and optimization suggestions. When the system is running, the component spectrum acquisition device first performs infrared spectrum scanning on the skin care product sample to obtain the original infrared spectrum data, and then accurately extracts the position and intensity distribution information of the spectral characteristic peaks from the data, and converts the information into a structured original spectrum feature vector. The formula component analysis engine receives the feature vector, matches and calculates it with a large number of standard spectral features stored in the preset component database, identifies the specific active components contained in the skin care product and their respective concentration ratios, and outputs a structured component identification list. The component interaction relationship modeling module retrieves the chemical interaction knowledge base based on the list, finds the known synergistic or antagonistic relationship between the identified components, and constructs a relationship network describing the mutual influence between the components, i.e. a component interaction graph. The stability prediction unit uses the graph to simulate and analyze the stability change trend of each component under different environmental conditions such as temperature, humidity and light intensity, especially the degradation rate and possible derivatives, thereby outputting a stability prediction report. The safety evaluation module integrates the results of the stability prediction report and information from the toxicology database to calculate the safety threshold and potential risk level of each component and its possible derivatives, and generates a quantitative safety evaluation matrix. The formula optimization suggestion generator takes the matrix and the component interaction graph as the basis to propose suggestions for alternative components that are functionally similar but safer and more stable, or adjust the concentration ratio, for components with high risk or poor stability, to form an optimized formula suggestion table. Finally, the detection report integration unit logically checks and structures all the intermediate results, including the component identification list, the stability prediction report, the safety evaluation matrix and the optimized formula suggestion table, to generate a final detection report document with a standard format and complete content.

[0060] Embodiment 1: see Figure 2, the ingredient spectrum acquisition device as the front-end sensing unit of the system, its core task is to convert the physical entity of skin care product sample into digital features for computational analysis, the device contains a series of closely coordinated sub-modules to complete this conversion. The spectrum preprocessing submodule first processes the original interferogram signal collected by the Fourier transform infrared spectrometer, the original data inevitably contains system noise from the instrument itself and the environment and baseline tilt due to sample scattering and other factors, the preprocessing process uses iterative least squares method to fit the spectral baseline, through multiple iterations to find an optimal polynomial curve to represent the baseline, then subtract the baseline corresponding to the original spectral data points to achieve baseline correction, for the residual random noise, the sliding window based smoothing filter algorithm is applied, which effectively suppresses the interference of high frequency noise by local polynomial fitting of data points within the preset width window, while maximizing the retention of the true shape characteristics of the spectrum. After completing these purification steps, the system automatically intercepts the spectral band range containing the vibration information of key functional groups such as carbon-hydrogen bond, hydroxyl group and carboxyl group as the effective spectral interval according to the general feature interval division rules of the international standard infrared spectrum database, and discards those marginal bands with weak signal or serious interference.

[0061] The feature peak extraction submodule then performs feature recognition on the purified spectral curve, this submodule uses a peak detection algorithm based on derivative spectrum, by calculating the first and second derivatives of the spectrum to accurately determine the position and boundary of the spectral peak, because the derivative zero crossing corresponds to the extreme point of the original spectrum, i.e. the peak top position, and the derivative extreme point corresponds to the inflection point of the original spectrum which can be used to determine the start and end points of the peak. For each potential feature peak identified, the submodule calculates its full width at half maximum, i.e. the width at half the peak height, this parameter is closely related to the chemical environment of the molecule, and records the absorbance value corresponding to the peak top as the peak intensity, all identified feature peaks are sorted by their wave number from high to low and assigned a unique temporary identifier, forming a feature peak list containing three attributes of position, width and intensity.

[0062] The vector encoding submodule is responsible for structuring the feature information in the list form into a one-dimensional numerical vector. The submodule predefines a fixed-length vector template, with the total dimension determined by the maximum number of feature peaks expected to be processed by the system. Each feature peak occupies three consecutive dimensions to store its wave value, half-peak width value, and intensity value, respectively. To eliminate the numerical differences caused by different physical dimensions, the half-peak width and intensity data are normalized before being filled into the vector, mapping them linearly to the interval of zero to one. The wave value is encoded using relative displacement, i.e., subtracting a reference value from the actual wave number to reduce the numerical size. The data of all feature peaks are sequentially filled into the corresponding positions of the one-dimensional array in descending order of wave number. For cases where the actual number of peaks is less than the maximum number, the remaining dimensions are filled with specific identifiers. The final generated original spectrum feature vector is a digital object that can be efficiently processed by a computer and fully carries the chemical fingerprint information of the sample.

[0063] The work of the formula component analysis engine begins with a deep interpretation of the feature vector. The standard feature matching submodule internally maintains a database containing thousands of standard spectrum features of common cosmetic raw materials. Each standard feature is stored in the form of a vector with the same dimension and encoding rules as the original spectrum feature vector. The matching process uses the similarity measurement method under the vector space model to calculate the cosine of the angle between the input vector and each standard vector in the database. The closer the value is to one, the more consistent the directions of the two vectors, i.e., the more similar the spectrum shapes. The system administrator can set a similarity threshold, and only those database components with a cosine similarity exceeding the threshold will be screened as candidate components. To improve matching efficiency, the submodule uses the inverted index technology to pre-organize the database, allowing the matching calculation to quickly focus on the subset of feature peaks common to the input vector, thereby significantly reducing unnecessary global comparisons.

[0064] The concentration inversion submodule gives quantitative meaning to the qualitative analysis results. This submodule accesses an independent concentration calibration curve library, which stores the change model of the feature peak intensity of each possible component at different concentrations. These models are usually pre-established through a series of standard samples with known concentrations. For a given candidate component, the submodule extracts the intensity data of its key feature peaks from the input vector and then substitutes them into the calibration curve equation corresponding to that component for inversion calculation. If the component has multiple feature peaks available for quantification, the submodule will use a weighted average algorithm to combine the calculation results of each peak to obtain a more robust concentration estimate. During the calculation process, factors such as the dilution factor that may exist during sample preparation are also considered for necessary correction. Finally, the estimated mass percentage or concentration range of the component in the skin care product under test is output.

[0065] The list generation submodule is responsible for integrating the scattered results generated by the aforementioned processes into a structured report. This submodule first ranks all candidate components according to their matching similarity scores and calculated concentration values, typically giving higher weights to components with higher similarity. The ranked component list is then assembled into a structured data object, such as a JavaScript Object Notation (JSON) object. This object is an array, with each element representing an identified component. Each element is a sub-object containing multiple fields, including at least the component's standard name, unique identifier in the database, calculated concentration value and unit, matching similarity score, and characteristic peak number for quantification. The generated component identification list is not only used by subsequent modules within the system but can also be directly displayed to users as the initial component analysis result.

[0066] The implementation process of the entire embodiment 1 reflects the progressive conversion from analog signals to digital features and then to chemical information. Spectral preprocessing ensures the quality of input data, feature peak extraction captures essential information from data, vector encoding makes it suitable for computational analysis, and the formula component analysis engine maps digital features back to specific chemical components and their contents through large-scale comparison and quantitative models, providing the most basic and crucial data input for the subsequent advanced analysis functions of the entire system. The design of each submodule fully considers the complex situations that may be encountered in practical applications, such as data noise, baseline drift, and spectral overlap, to enhance the robustness and accuracy of the system through algorithm selection and process design.

[0067] Embodiment 2: Refer to Figure 3 The operation of the component interaction relationship modeling module begins with the activation of the relationship retrieval submodule. Upon receiving the structured component identification list from the formula component analysis engine, this submodule immediately initiates a batch query process of the background chemical interaction knowledge base. The chemical interaction knowledge base is a structured database meticulously compiled by humans, similar in structure to a professional knowledge graph. It stores component pair interaction data extracted from a large number of scientific literature, patents, and experimental reports. The query process takes each component in the list as the subject and retrieves all recorded object components that interact with the subject component from the knowledge base. The returned results typically include the type of interaction, the strength level of the interaction, the possible conditions for the interaction, and the literature source and evidence level of the record. The relationship retrieval submodule needs to preliminarily integrate the returned redundant or potentially conflicting information. For example, when multiple descriptions of the interaction between the same pair of components are retrieved from different sources, the submodule will generate a comprehensive judgment based on pre-set rules.

[0068] The network construction submodule integrates the retrieved binary relationship data, which is still in a discrete state, into a global relationship network. This submodule uses a directed graph model in graph theory as its core data structure, abstracts each independent component in the component identification list as a node in the graph, and assigns attributes to each node, such as component name, concentration, and functional classification. A directed edge between any two nodes represents the interaction relationship obtained from the relationship retrieval results, with the direction of the edge pointing from the influencing component to the affected component. For example, if component A can enhance the stability of component B, the direction of the edge is from A to B. The weight attribute of the edge is used to quantify the strength and nature of the interaction, with positive weights for synergistic interactions and negative weights for antagonistic interactions. The absolute value of the weight can be mapped and assigned according to the strength levels recorded in the knowledge base. For complex multi-component systems, there may be indirect interactions, and the network construction submodule has the ability to infer indirect relationships. For example, by querying records of A affecting B and B affecting C, it can be inferred that A may indirectly affect C through B. Although the strength of this indirect influence path may be attenuated, it is crucial for understanding the overall behavior of the formula. The initial directed graph constructed may contain multiple connected components or isolated nodes, reflecting the coverage of the current knowledge base and the known extent of the relationships between components in the specific formula.

[0069] The atlas generation submodule then performs a series of optimization and standardization processes on the initially constructed directed graph, making it more suitable for subsequent analysis tasks such as stability prediction. Topological sorting is a key process in which the algorithm attempts to linearly arrange the nodes in the graph, such that for any directed edge in the graph, the starting node is arranged before the terminal node. This sorting helps to more clearly demonstrate the influence flow and dependency relationships between components. Although not all graphs containing interaction relationships can be completely topologically sorted (as there may be circular influences), the algorithm will try to find a maximum linear node sequence. Weight normalization is performed to make the strengths of different interactions comparable. This process maps the weight values of all edges in the graph to a unified interval, such as zero to one or negative one to positive one. Common methods include min-max normalization or Z-score standardization. The normalized weights make the relative strengths of synergistic or antagonistic interactions clear at a glance. After completing these processes, the submodule encapsulates the final graph structure data, including the node set, edge set, and all their attributes, into a data object. This object is the component interaction relationship atlas. The atlas can be output in various forms, such as adjacency matrices or edge lists for subsequent machine calculations, or rendered into intuitive network graphs using a visualization engine. The size of the nodes in the graph can map the component concentrations, and the thickness and color of the edges can map the strength and type of interaction, providing a powerful tool for formula engineers to intuitively understand the complex relationships within the formula.

[0070] The whole process of ingredient interaction modeling embodies the sublimation from data retrieval to knowledge extraction, the relationship retrieval sub-module ensures that the modeling fact is derived from verified scientific knowledge, the network construction sub-module weaves scattered facts into a related network, and the atlas generation sub-module makes this network clear, regular and easy to use through algorithm optimization. The output of this module: ingredient interaction relationship atlas, is no longer a simple list of ingredients, but a model that reveals the potential of internal dynamic interaction of the formula, which presents the systematic characteristics of the formula and lays a solid foundation for predicting its behavior under various conditions. The quality of this atlas is highly dependent on the breadth, depth and accuracy of the background chemical interaction knowledge base, therefore, the continuous maintenance and update of the knowledge base, such as automatically extracting relationship data from newly published literature through natural language processing technology, is crucial to keeping the whole system up-to-date. At the same time, the construction logic of the atlas also takes into account the complexity of actual cosmetic formulations, which can handle systems containing dozens of ingredients, and ensures that the calculation process is completed within an acceptable time through efficient graph algorithms.

[0071] The environmental parameter input interface of the stability prediction unit plays the role of a bridge between the system and the external environmental conditions. This interface is designed in a highly flexible and configurable mode, allowing users to set simulation conditions in multiple ways. Users can input single, deterministic temperature, humidity, and light intensity values to simulate specific storage scenarios, or input a parameter variation sequence to simulate accelerated aging tests or day-night, seasonal replacement effects. The original parameters received by the interface are format-verified and unit-uniformly converted, then packaged into a standard environmental condition parameter object, which serves as the initial driving condition for subsequent simulation calculations. The variation trend analysis sub-module is the core of the stability prediction operation, which contains an understanding of chemical kinetics and the complexity of the formulation system. This sub-module has pre-set basic thermodynamic and kinetic parameters for each known cosmetic ingredient, such as activation energy, pre-exponential factor, etc. These parameters form the basis of the prediction. When receiving specific environmental parameter combinations, the sub-module will solve the degradation kinetics equation for each ingredient in the ingredient identification list under the given environmental conditions. A model for estimating the residual rate of an ingredient after a specific time is as follows:

[0072]

[0073] Where: R i,t represents the residual rate fraction of ingredient i after time t, its value is between 0 and 1, k i is a reaction rate constant closely related to the properties of ingredient i and the environmental temperature, the influence of temperature on k i is usually described by the Arrhenius formula, t represents the duration of the simulation, and αi is a parameter related to the degradation mechanism of component i, which is about 1 for a simple first-order reaction, and can deviate from 1 for more complex diffusion-controlled or other mechanisms. The complexity of this submodule lies in that it does not calculate the degradation of each component in isolation, but rather modifies the underlying kinetic equations based on the information provided by the interaction map. For example, when the map indicates that component A has a catalytic degradation effect on component B, an enhancement factor proportional to the concentration of component A is introduced when calculating the degradation rate constant k B of component B, thereby simulating this catalytic effect; conversely, if there is a stabilizing synergistic effect, the effective rate constant can be correspondingly reduced. In addition to calculating the concentration decay of the parent components, this submodule also integrates a degradation pathway knowledge base to predict the main degradation products and their generation kinetics over time, calculating the estimated concentrations of various potential degradation products at a given time point.

[0074] The report generation submodule organizes the vast numerical results output by the change trend analysis submodule into a stability prediction report with high readability, which usually adopts a hierarchical structure, first summarizing the overall stability performance of the formula under the given simulation conditions, such as indicating the active ingredient most sensitive to specific environmental factors. The main body of the report then elaborates in detail by environmental condition classification, under the chapter of "high temperature and high humidity", it may list the predicted half-life of the main active ingredients, the percentage of activity retention at the end of the simulation period, and highlight the names of potential degradation products and their maximum estimated concentrations, while citing the interaction map to explain the reasons for certain abnormal degradation behaviors, such as "due to the antagonistic effect of components X and Y, the degradation rate of Y is increased by about 30% compared to when it exists alone". The report not only contains data, but also generates risk prompts according to pre-set rules, such as when a certain degradation product concentration is predicted to exceed its safety threshold, it will be prominently marked. The threshold query submodule of the safety assessment module is the benchmark setting link of risk assessment, which is connected to a toxicology database integrating multiple national regulations and scientific literature, each record in the database is associated with specific experimental data sources and applicable conditions. The query process takes a list of component identifiers and newly identified potential degradation products in the stability prediction report as input, batch retrieves the key toxicology parameters of each plant, including but not limited to skin irritation concentration limit, eye irritation classification, sensitization probability, phototoxicity data, and the legal upper limit of the concentration of some components. The database design takes into account the variability and uncertainty of the data, for the case of multiple research results for the same component, the submodule will give a recommended reference threshold range based on data quality, novelty and authority.

[0075] The risk calculation sub-module performs the core risk quantification task, which is a dynamic and comprehensive process. It calculates a base risk index for each initial ingredient in the formula, which is a combination of the ingredient's use concentration and its closeness to the threshold values of skin irritation, sensitization, etc. More importantly, it must integrate the results of stability prediction and assess the potential risks brought by ingredient transformation. For example, even if the initial ingredient A is safe, it can generate a more sensitizing degradation product A1 after storage. The final risk level of ingredient A will be determined by its own risk and the predicted risk of its degradation product A1, and the contribution weight of degradation product A1 is related to the product of its predicted generation amount at the end of the simulation period and its unit risk. The calculation will consider multiple risk endpoints (irritation, sensitization, etc.), and generate independent risk scores for each endpoint, and finally summarize into a comprehensive risk level. The matrix generation sub-module integrates all the above evaluation results into a compact but informative two-dimensional evaluation matrix. This matrix usually takes ingredients as rows and different safety indicators (such as skin irritation risk, eye irritation risk, sensitization risk, light safety risk, etc.) as columns. Each cell of the matrix not only contains a general risk level symbol (such as high, medium, low or color coding), but also contains concise data support, such as "use concentration: 0.5%, reference safe concentration threshold: 1%". For ingredients with increased risk due to degradation products, a note will be made in the remarks column, "Note: Long-term storage may generate degradation product XX, whose predicted risk is high". This safety evaluation matrix finally provides a clear and direct action guide for the formula optimization suggestion generator, clearly indicating the risk points that need to be prioritized.

[0076] Example 4: The core task of the formulation optimization suggestion generator is to propose practical improvement measures for the risk points revealed in the safety assessment matrix, which starts with the activation of the alternative ingredient screening submodule. This submodule first parses the safety assessment matrix to identify all ingredients labeled as "high risk" or "medium risk", as well as those that show significant instability in the stability prediction report, which are listed as potential replacement targets. For example, assume that in a moisturizing cream formula, the preservative "pentylene glycol" is labeled as having potential sensitization risk in the safety assessment matrix, and the stability prediction report indicates that it may degrade at high temperatures, so it is locked as a target ingredient that needs to be optimized. Then, the submodule accesses the preset ingredient database, which not only contains the spectral characteristics of the ingredients, but also a knowledge base containing ingredient function classification, chemical properties, safety profile and compatibility information. The system takes the function of the target ingredient as the main retrieval condition, and in the above example, it takes "preservative" as the core function for retrieval, while setting a filter that requires the safety score of the candidate replacement ingredient in the toxicology database to be better than the target ingredient, and its chemical stability index to meet the requirements. The search may return multiple results, such as "octanol", "ethyl hexyl glycerol" and other ingredients with preservative functions and better safety profiles. The submodule will further call the ingredient interaction graph to check whether these candidate replacement ingredients and other main active ingredients in the formula (such as moisturizers, antioxidants) have known antagonistic effects, to ensure that the introduction of new ingredients will not destroy the overall synergistic effect of the formula. After this screening, a list of alternative ingredient candidates for each target ingredient is generated, sorted by comprehensive matching degree.

[0077] The proportion adjustment submodule needs to determine a scientifically reasonable recommended addition concentration for the candidate replacement ingredient after obtaining it. This process is not simply an equal replacement. The submodule will first query the effective concentration range and safe use limit of the candidate replacement ingredient itself, which comes from the regulatory database and technical materials provided by the raw material supplier. More importantly, it needs to deeply analyze the ingredient interaction relationship map and evaluate the chain effect of the replacement behavior on the formulation system. Continuing with the example of replacing "pentylene glycol", assuming that the concentration of "pentylene glycol" in the original formula is 0.5%, and the interaction relationship map shows that it has a slight synergistic effect with another moisturizing ingredient "panthenol". When "octylene glycol" is selected as the replacement, the proportion adjustment submodule will first confirm the commonly used effective concentration range of "octylene glycol" (such as 0.2%-0.8%), and then analyze whether there is a known interaction between "octylene glycol" and "panthenol". If the database shows that there is no interaction between the two, the system may recommend an "octylene glycol" addition amount close to the original concentration, such as 0.5%; but if the interaction map shows that "octylene glycol" has a strong synergistic effect with another ingredient in the formula "nicotinamide", allowing the same effect to be achieved at a lower concentration, the submodule may recommend a lower, more economical and safe margin concentration, such as 0.3%. The adjustment calculation is a multi-objective optimization process that needs to consider efficacy, safety and cost.

[0078] The suggestion table generation submodule formats the aforementioned results into clear operation guidelines, and the optimized formula suggestion table it produces not only lists the replacement options, but also clarifies the logic of the replacement and the expected benefits. Referring to Table 1, a simplified suggestion table is shown.

[0079] Table 1: Optimized Formula Suggestion Table

[0080]

[0081] The report integration unit is the collection point of the information flow, and its data verification submodule performs an important quality control step before integration begins. The submodule checks the internal consistency of all input data according to pre-set business rules. For example, it checks whether the proposed replacement "pentylene glycol" in the optimization suggestion table actually exists in the ingredient identification list and whether the concentration value matches the record in the list; it checks whether the degradation products mentioned in the stability prediction report have been assessed in the safety assessment matrix. If a logical conflict is found, such as replacing an ingredient that is not on the list at all, the system will pause the integration process and issue a warning, prompting the relevant modules to review the data.

[0082] The structured template sub-module defines the skeleton and style of the entire final report, which is a prefabricated digital document framework containing all fixed elements such as title, chapters, table, chart position, etc. The work of the sub-module is to fill the data in the checked ingredient identification list, stability prediction report, safety evaluation matrix and optimization formula suggestion table into the corresponding positions of the template according to the one-to-one mapping relationship. For example, the content of the ingredient identification list will be filled into the table of chapter 2 "ingredient identification results" of the report; the summary information of the safety evaluation matrix will be converted into the risk level list in chapter 3 "safety evaluation", supplemented by a brief description; and the detailed optimization formula suggestion table will be inserted into chapter 4 "formula optimization suggestion" as a whole. This process ensures that each generated report has a unified, professional appearance and structure.

[0083] The document generation sub-module completes the final conversion from data to finished product, which receives the complete template object that has been filled with data, calls the underlying document processing engine to render it into the editable or read-only format specified by the user. Common output formats include portable document format (PDF) for non-changeable formal report distribution, and editable document format (such as DOCX) to allow formula engineers to make further annotations and modifications based on it. The generated final test report document is a self-contained file that logically presents the entire reasoning process and final conclusion from the spectral analysis of the original sample to the specific optimization scheme, providing users with a complete technical basis from problem diagnosis to solution.

[0084] Example 5: The user feedback interface is designed as a structured data input platform, allowing authorized formula engineers or quality control personnel to provide review comments on the final test report document generated by the system. The interface is usually integrated into the side panel of the report preview interface or as part of the online review tool. Users can add comments next to specific paragraphs, data tables, or conclusions in the report. The types of comments can be corrections to facts, questions about reasoning, or the addition of new information. For example, the system may analyze the spectral characteristic peak of a certain essence as "nicotinamide", but the user confirms that the ingredient is "nicotinic acid ethyl ester" based on the actual raw material records. The user can add a correction comment to the corresponding ingredient identification result in the report through the feedback interface, and upload the supplier-provided raw material test certificate as supporting materials. Another common feedback is to supplement the storage conditions of specific products that the system does not have, such as the user noting that the product will actually be stored and sold in a constant-temperature refrigerator, which is different from the system's default normal temperature prediction conditions. The user can select "update environmental parameters" through the interface and input the new temperature range. All feedback data submitted through the interface is time-stamped, user-identified, and associated with the original report version number, and is stored in a dedicated feedback database for processing.

[0085] The dynamic update engine is automatically activated after monitoring new user feedback data. Its core responsibility is to evaluate the effectiveness of the feedback and trigger the update process of the system's internal data and models accordingly. First, the engine will analyze the feedback content and identify its type and associated data objects. For example, the engine will confirm whether the "nicotinic acid ethyl ester" proposed by the user exists in the system's standard ingredient database. If it exists, the engine will call the formula ingredient parsing engine, using the user-corrected ingredient name as an important reference, to re-match the original spectral feature vector. This process may involve adjusting the weight parameters of the matching algorithm to make the spectral features more inclined to match the user-confirmed ingredient. Once the ingredient list is corrected, the engine will start a chain update process, sequentially re-executing ingredient interaction relationship modeling, stability prediction, safety evaluation, and formula optimization suggestion generation. Finally, an updated final test report document containing user-corrected information will be generated. For feedback that supplements storage conditions, the engine will directly call the stability prediction unit to perform simulation calculations again using the new environmental parameters, and update the relevant parts of the report regarding stability and risk accordingly.

[0086] The version control submodule inside the dynamic update engine is responsible for maintaining the traceability of the entire modification history. Whenever a new version of the report is generated due to user feedback, this submodule automatically creates a version record. This record contains the unique number of the new version, the time of generation, a summary of the user feedback that triggered this update, and a detailed data change log. The change log precisely records which data has been changed, for example, “ingredient ‘Unknown X’ corrected to ‘Ethyl Nicotinate’”, “stability prediction environmental condition updated from ‘25°C’ to ‘4°C’”, “risk level of ingredient Y adjusted from ‘medium’ to ‘low’ due to improved stability”. This detailed version management allows any user to clearly trace the complete evolution of the report from the initial draft to the current version, understand the reason and specific content of each modification, which is crucial for quality system audit and review of the formulation development process.

[0087] The incremental learning submodule is an advanced function that enables the system intelligence to continuously improve. It converts the high-value feedback data confirmed by users into the system’s own knowledge. The submodule sets a confirmation mechanism, for example, when a report that has been corrected by the user is finally marked as “audit passed”, the correction data related to it is recognized as a high-quality and high-confidence labeled sample. Taking the correction of ingredient identification as an example, “Ethyl Nicotinate” and its corresponding original spectral feature vector constitute a valuable correction sample pair. The incremental learning submodule regularly adds these new correction samples to the training sample set of the standard feature matching submodule of the formula ingredient analysis engine. In subsequent model retraining, the machine learning algorithm learns the subtle differences between the spectral features of “Ethyl Nicotinate” and the previously misjudged “Nicotinamide”, so that in the future, it can more accurately distinguish similar spectra. Similarly, user-confirmed stability data that is consistent with the actual situation can also be used to fine-tune the parameters in the degradation kinetics model. Through this continuous, real feedback-based learning, the accuracy of the system’s ingredient identification and the credibility of the prediction model can be continuously optimized over time, gradually reducing the dependence on manual correction.

[0088] It should be noted that the relational terms, such as first and second, and the like, are used solely to distinguish one from another entity or action, without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms “comprises,” “comprising,” or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0089] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A multi-dimensional rapid detection and comprehensive analysis system for skincare product formulations, characterized in that, include: The component spectral acquisition device acquires the infrared spectral data of skin care products, extracts the position and intensity distribution of spectral characteristic peaks, and generates the original spectral feature vector; The formula ingredient analysis engine receives the original spectral feature vector, matches it with the standard spectral features in the preset ingredient database, identifies the active ingredients and concentration ratios in the skin care product, and outputs a list of ingredient identifiers. The component interaction modeling module, based on the component identifier list, retrieves the chemical interaction knowledge base, constructs a network of synergistic or antagonistic relationships between components, and generates a component interaction graph. The stability prediction unit analyzes the trend of component stability changes under different environmental conditions based on the component interaction relationship map and outputs a stability prediction report. The safety assessment module combines the stability prediction report and the toxicology database to calculate the safety threshold and potential risk level of each component and generate a safety assessment matrix. Based on the safety assessment matrix and component interaction graph, the formulation optimization suggestion generator screens replaceable components or adjusts their concentration ratios, and outputs an optimized formulation suggestion table. The test report integration unit integrates the ingredient identification list, stability prediction report, safety assessment matrix, and optimized formulation suggestion table in a structured manner to generate the final test report document.

2. The multi-dimensional rapid detection and comprehensive analysis system for skincare product formulations according to claim 1, characterized in that, The component spectrum acquisition device includes: The spectral preprocessing submodule performs baseline correction and noise filtering on the raw infrared spectral data and extracts the effective spectral range. The feature peak extraction submodule identifies the position and half-width parameter of the feature peaks within the effective spectral range and marks the intensity distribution of the feature peaks. The vector encoding submodule converts the characteristic peak position, half-peak width parameter and intensity distribution into a multi-dimensional numerical vector to generate the original spectral feature vector.

3. The multi-dimensional rapid detection and comprehensive analysis system for skincare product formulations according to claim 2, characterized in that, The formula component analysis engine includes: The standard feature matching submodule calculates the similarity between the original spectral feature vector and the standard spectral features in the preset component database, and identifies candidate components whose matching degree exceeds a threshold. The concentration inversion submodule calculates the actual concentration ratio of the candidate ingredient in the skin care product based on the characteristic peak intensity and concentration calibration curve; The list generation submodule integrates the candidate components and their concentration ratios, sorts them by priority, and outputs a list of component identifiers.

4. The multi-dimensional rapid detection and comprehensive analysis system for skincare product formulations according to claim 3, characterized in that, The component interaction modeling module includes: The relationship retrieval submodule extracts known interaction relationship records for each component in the component identifier list from the chemical interaction knowledge base; The network construction submodule converts the interaction relationship records into a directed graph structure, where nodes represent components and edges represent the direction and intensity of cooperative or antagonistic effects; The graph generation submodule performs topological sorting and weight normalization on the directed graph structure, and outputs a component interaction relationship graph.

5. The multi-dimensional rapid detection and comprehensive analysis system for skincare product formulations according to claim 4, characterized in that, The stability prediction unit includes: The environmental parameter input interface receives simulated conditions such as temperature, humidity, and light intensity. The trend analysis submodule calculates the degradation rate and derivative generation probability of each component under different environmental conditions based on the component interaction relationship map. The report generation submodule categorizes and integrates the degradation rate and derivative generation probability according to environmental conditions, and outputs a stability prediction report.

6. The multi-dimensional rapid detection and comprehensive analysis system for skincare product formulations according to claim 5, characterized in that, The security assessment module includes: The threshold query submodule retrieves the skin irritation and sensitization safety thresholds for each ingredient in the ingredient identifier list from the toxicology database; The risk calculation submodule combines the derivative generation probability in the stability prediction report to assess the potential composite risk level. The matrix generation submodule maps the safety thresholds and risk levels of each component into a two-dimensional evaluation matrix, generating a safety evaluation matrix.

7. The multi-dimensional rapid detection and comprehensive analysis system for skincare product formulations according to claim 6, characterized in that, The recipe optimization suggestion generator includes: The alternative ingredient screening submodule retrieves alternative ingredients with similar functions and higher safety thresholds from a preset ingredient database based on the high-risk ingredient identifiers in the safety assessment matrix. The proportion adjustment submodule recalculates the recommended concentration ratio of the substitute component based on the synergistic effect strength in the component interaction relationship map; The suggestion table generation submodule sorts the alternative ingredients and their recommended concentration ratios according to optimization priority and outputs an optimized formulation suggestion table.

8. The multi-dimensional rapid detection and comprehensive analysis system for skincare product formulations according to claim 7, characterized in that, The test report integration unit includes: The data verification submodule performs a logical consistency check on the ingredient identification list, stability prediction report, safety assessment matrix, and optimized formulation suggestion table. The structured template submodule maps and populates the validated data according to the field requirements of the preset report template; The document generation submodule converts the filled template into an editable document format and generates the final test report document.

9. The multi-dimensional rapid detection and comprehensive analysis system for skincare product formulations according to claim 8, characterized in that, Also includes: The user feedback interface receives suggestions for modification or supplementary data for the final test report document; The dynamic update engine adjusts the component interaction graph and safety assessment matrix based on the modification suggestions or supplementary data, and regenerates the optimized formulation suggestion table and the final test report document.

10. The multi-dimensional rapid detection and comprehensive analysis system for skincare product formulations according to claim 9, characterized in that, The dynamic update engine includes: The version control submodule records the report version and corresponding data change records after each modification. The incremental learning submodule uses user feedback data as new samples to input into the formula component analysis engine and updates the standard spectral feature matching rules.

Images