Search

CN-121725925-B - Automatic beverage formula generation method based on artificial intelligence

CN121725925BCN 121725925 BCN121725925 BCN 121725925BCN-121725925-B

Abstract

The invention relates to the technical field of beverage formulas AI and discloses an automatic beverage formula generation method based on artificial intelligence. The method comprises the steps of collecting flavor characteristic data of five raw pulps of raspberry, mulberry, medlar, pomegranate and purple grape and health component data containing synergistic effects of basic nutrients, plant compounds and derivative compounds of the basic nutrients, and establishing a structured raw material database. And extracting features of database data through a machine learning model, generating flavor vectors and health component vectors of all raw materials, and realizing quantitative characterization of the data. And (3) performing multi-objective optimization on the two types of vectors by adopting an optimization algorithm, and screening out candidate formula sets which simultaneously meet a preset flavor threshold and a health threshold. And further screening the candidate formulas according to the user preference data, finally determining the five tiger magical pulp formulas, and outputting the five tiger magical pulp formulas to a formula execution system. The method accurately meets the personalized demands of users, and provides a reliable technical path for standardized production of five tiger magical pulp.

Inventors

  • LU FENG

Assignees

  • 福建榕基软件股份有限公司
  • 河南滋身润心生物科技有限公司

Dates

Publication Date
20260505
Application Date
20260212

Claims (8)

  1. 1. An automatic beverage formula generation method based on artificial intelligence is characterized by comprising the following steps: (1) Establishing a multi-dimensional raw material database, namely collecting flavor characteristic data and health component data of five raw materials of raspberry, mulberry, medlar, pomegranate and purple grape, and establishing a structured raw material database; (2) Feature extraction and vector generation, namely performing feature extraction on the database data in the step (1) through a machine learning model to generate flavor vectors and health component vectors of all raw materials; (3) Multi-objective optimization screening, namely performing multi-objective optimization on the flavor vector and the health component vector in the step (2) by adopting an optimization algorithm, and screening a candidate formula set which simultaneously meets a preset flavor threshold value and a health threshold value; (4) The personalized recipe determination, namely carrying out secondary screening on the candidate recipe set in the step (3) in combination with user preference data to generate a final five tiger magical pulp recipe, and outputting the final recipe to a recipe execution system; the multi-objective optimization of the flavor vector and the health component vector in the step (2) by adopting an optimization algorithm in the step (3) comprises the following steps: constructing a multi-objective optimization function with the aim of maximizing flavor coordination and health benefits; carrying out iterative solution on the multi-objective optimization function by adopting a genetic algorithm to generate a candidate formula set meeting a preset flavor threshold and a health threshold; The proportioning weight of each raw material is dynamically adjusted in the iterative process, so that the diversity of candidate formula sets is ensured; the first objective function is a flavor coordination objective function, and is used for quantifying the flavor coordination degree of the mixed formula, and the specific form is as follows: ; Wherein, the Represents the first The mass ratio weight of the seed raw materials in the formula is as follows , Corresponding to raspberry raw pulp, Corresponding to mulberry raw pulp, Corresponding to the original pulp of medlar, Corresponding to the pomegranate primary pulp, Corresponding to the raw grape pulp; Represents the first The flavor vector of the seed raw material is generated by deep neural network for reducing dimension and is A dimension real number vector; Representing a preset ideal flavor vector, and constructing by researching flavor characteristics of main stream composite original pulp drinks in the market and combining scoring data of a professional evaluation team; representing the flavor harmony score with a value range of The closer the value is to 1, the more coordinated the formula flavor and the ideal flavor; The second objective function is a health benefit objective function, which is used for evaluating the comprehensive health value of the formula, and the specific form is as follows: ; Wherein, the Represents the first Seed material of the first kind The content of the key health components is detected by high performance liquid chromatography and normalized; Represents the first The weight coefficient of the seed health component is set according to the nutritional value of the component; Representing the total number of species of the key health component; Representing health benefit score, the value range is Higher values indicate higher health value of the formulation; The secondary screening of the candidate formula set in the step (3) by combining the user preference data in the step (4) comprises the following steps: Acquiring historical selection data of a user, and analyzing preference trends of the user on sweetness, acidity and health components; calculating the matching degree of the candidate formula set and the user preference data by adopting a collaborative filtering algorithm; and selecting the candidate formula with the highest matching degree as the final five tiger magical pulp formula.
  2. 2. The automatic beverage formulation generating method based on artificial intelligence according to claim 1, wherein the collecting of the flavor profile data and the health component data of the five raw pulps of raspberry, mulberry, medlar, pomegranate, and purple grape in step (1) comprises: extracting volatile flavor compound data including sweetness, acidity and aroma intensity of Rubi fructus primary pulp, mori fructus primary pulp, lycii Frutus primary pulp, punica granatum primary pulp and purple grape primary pulp by gas chromatography-mass spectrometry; extracting polyphenol content, vitamin level and antioxidant activity of Rubi fructus juice, mori fructus juice, lycii Frutus juice, punica granatum juice and fructus Vitis Viniferae juice by high performance liquid chromatography; And storing the data of the volatile flavor compounds, the polyphenol content, the vitamin level and the antioxidant activity into a raw material database after normalization treatment, so as to ensure the accuracy and comparability of the data.
  3. 3. The automated artificial intelligence based beverage formulation generation method according to claim 2, wherein the feature extraction of the database data of step (1) by the machine learning model of step (2) comprises: performing dimension reduction treatment on the volatile flavor compound data by adopting a deep neural network to generate flavor vectors of all raw materials; Adopting a random forest algorithm to perform characteristic selection on the polyphenol content, the vitamin level and the antioxidant activity to generate a healthy component vector of each raw material; The flavor vector and the health component vector are associated to corresponding raw material entries in a raw material database.
  4. 4. The automated artificial intelligence based beverage formulation generation method of claim 3, further comprising a dynamic data update procedure: monitoring the change of the flavor characteristic data and the health component data of each raw material in the raw material database in real time; Triggering a machine learning model to regenerate a flavor vector and a health component vector when the change of the data is detected and the change quantity exceeds the preset fluctuation threshold flavor data plus or minus 5% and the health data plus or minus 3%; And (3) re-executing the multi-objective optimization screening in the step (3) to the personalized recipe determination in the step (4) based on the updated flavor vector and the health component vector.
  5. 5. The automated artificial intelligence based beverage formulation generation method according to claim 4, wherein the real-time monitoring of changes in flavor profile data and health component data for each ingredient in the ingredient database comprises: Collecting freshness indexes and component stability indexes of all raw materials through an Internet of things sensor; If the freshness index or the ingredient stability index exceeds the preset range, marking the corresponding raw material data as a state to be updated; feature extraction and optimization calculations are re-performed only on raw material data marked as a state to be updated.
  6. 6. The method of claim 5, further comprising a feedback optimization mechanism: after a final five tiger magical pulp formula is generated, recording flavor characteristic data and health component data of the final five tiger magical pulp formula; And feeding the recorded flavor characteristic data and health component data back to a machine learning model for optimizing a subsequent characteristic extraction process.
  7. 7. The method of claim 6, further comprising contribution analysis: Creating a formula generation log, and storing input data, an intermediate vector, compensation parameters and an output formula when generating the five tiger magma formula each time; Generating a log based on the formula to analyze the flavor contribution and health contribution of each raw material; And dynamically adjusting the initial weight of each raw material in the raw material database according to the flavor contribution degree and the health contribution degree.
  8. 8. The method of claim 7, further comprising biasing the control flow: After the formula execution system executes the five tiger magical pulp formula, collecting sensory evaluation data and component detection data of the actual finished product; Comparing the sensory evaluation data and the component detection data with theoretical data, and calculating a formula execution deviation; If the formula execution deviation exceeds the allowable range, jumping to a dynamic data updating flow if the raw material fluctuation causes, calibrating an execution system if the equipment is in error, jumping to a feedback optimization mechanism if the algorithm is in defect, and triggering a flow for regenerating the five tiger magma formula.

Description

Automatic beverage formula generation method based on artificial intelligence Technical Field The invention belongs to the technical field of artificial intelligence of beverage formulas, and particularly relates to an automatic beverage formula generation method based on artificial intelligence. Background Today, where the beverage industry is vigorously developing, consumer demands for beverages are no longer merely resting on the basis of thirst quenching, but increasingly focus on the uniqueness of the flavor and the health attributes of the ingredients. Under the trend, the composite raw pulp beverage with excellent taste and high nutritive value is gradually in the way of exposing the head and the corner, and becomes the focus of market attention. The composite raw pulp drink represented by five tiger magma is popular in this wave by virtue of the unique formula of selecting various natural raw pulps as raw materials. However, despite the increasing market demand, the formulation development of current "wuhushenjiang" and similar composite raw pulp beverages still relies heavily on traditional artificial experience patterns. Although the mode has certain practical accumulation advantages, the limitation is obvious, on one hand, the manual experience is difficult to fully cover the complex nutrition matching and flavor coordination requirements, and on the other hand, the traditional mode is worry about innovation efficiency and precision facing the increasingly diversified health requirements of consumers. This clearly presents a new challenge for further development of the industry. In the traditional formula development process, research and development personnel mainly rely on subjective cognition of different primary pulp flavors and ingredients, and the proportion of each raw material is adjusted through repeated experiments so as to realize balance of flavors and health. However, this process is not only time consuming and costly, but also has a lengthy trial-and-error period and overall lower efficiency. In addition, because of individual difference of human perception of flavor and limited quantitative analysis capability of healthy components, the flavor synergistic effect and reasonable collocation of the healthy components generated after mixing different raw pulps are difficult to accurately control. This limitation results in the end developed formulation being deficient in terms of stability, which makes it difficult to meet the consumer's dual needs for flavor consistency and health standardization. With the increasing subdivision of the consumer population, the demands of different users for beverage flavor (e.g., sweetness in acid, intensity) and health attributes (e.g., content of particular nutritional ingredients) exhibit significant differences. However, the conventional beverage development mode is very hard to realize in response to such personalized demands. While some formulation developments in the prior art have attempted to introduce data acquisition and analysis approaches, these approaches have often been limited to single-dimensional optimizations-e.g., focused on flavor tuning alone or focused on the tuning of healthy ingredients alone-failing to achieve multi-objective collaborative optimization of flavor and health. In addition, the lack of effective combination of the technical means and the user preference data leads to the failure of the effective combination to accurately match the personalized requirements of users, and finally, the developed formula has weak market adaptability, and the diversified market requirements are difficult to meet. At present, raw material data are mostly managed in a scattered state, and a unified database is not established to integrate and store flavor characteristic data and health component data of various raw pulps. The current situation causes low data calling efficiency in the research and development process, and is difficult to realize efficient multiplexing and deep analysis of data, so that the intelligent and accurate process of formula development is restricted to a certain extent. Therefore, how to break through the limitation of traditional artificial experience, a set of automatic formula generation method which can integrate raw material data, realize multi-objective optimization of flavor and health and combine user preference is constructed, and the method has become a key requirement for promoting the development and upgrading of the formula of five tiger magma and similar composite magma beverage. Disclosure of Invention The invention aims to provide an artificial intelligence-based automatic generation method for a beverage formula, which aims to solve the problems in the background technology. In order to achieve the above object, the present invention provides an artificial intelligence based automatic beverage formulation generating method, which comprises: Collecting flavor characteristic data of five primary