An AI-based method for precise optimization of rice and flour product formulations

Abstract

The application discloses a kind of based on AI's rice, flour product formula precision optimization method, belong to food formula optimization field.This method is realized by neural network modeling combined with genetic algorithm global optimization, the intelligent design and precision optimization of rice, flour product raw material powder formula.First, determine the flower with pleasant fragrance, the plant with natural pigment and the plant rich in polyphenol, freeze-drying and crushing are carried out to the above-mentioned three types of plant raw materials, then mixed with rice flour or flour, obtain rice, flour product composite raw material powder, and carry out the production of rice, flour product;Formulation test and sensory evaluation results are used to construct neural network model, and the model is globally optimized using genetic algorithm to determine the optimal formula of raw material powder.The optimized rice, flour product has specific floral fragrance, color and taste, and can meet the needs of individual customization.

Description

Technical Field

[0001] This invention belongs to the field of food formulation optimization, and specifically discloses an AI-based method for precise optimization of rice and flour product formulations. Background Technology

[0002] With increasing consumer awareness of health and growing demand for diversified food products, traditional rice and flour product formulas often struggle to simultaneously achieve multiple quality characteristics such as taste, color, and flavor. Existing technologies largely rely on experience and single experimental methods for formula design, making it difficult to systematically integrate the characteristics of various raw materials, and often requiring significant time and cost.

[0003] Liu Meiyu (2017) disclosed a method for making potato bread using unfermented frozen dough (publication number: CN106720055A), which includes: (1) accurately weighing 170g of high-gluten wheat flour, 30g of potato flour, 118g of water, 40g of white sugar, 8g of high-protein defatted high-calcium milk powder, 0.8g of bread improver, 20g of butter, 2g of salt, 16g of egg, 6g of active dry yeast, and 2.5g of trehalose; (2) kneading, proofing, and shaping; (3) wrapping the dough in plastic wrap and transferring it to -18℃ for freezing; (4) restoring the frozen dough to room temperature, fermenting at 25℃ for 30min, shaping at 35℃ for more than 30min, brushing with egg wash, and finally baking in an oven at 180℃ bottom heat and 230℃ top heat until golden brown. The advantage of this invention is that, based on experience, by appropriately adjusting the amount of water added and simultaneously adding the antifreeze agent trehalose, the damage caused by low temperature to the gluten structure and yeast is reduced, effectively improving the quality of non-fermented frozen dough.

[0004] Wang Junping (2024) disclosed a method for preparing buckwheat-quinoa composite bread (publication number: CN118058321A), specifically relating to the field of food processing technology. The bread is made from the following ingredients: 70-80 parts buckwheat flour, 20-30 parts quinoa flour, 2-3 parts yeast, 20-30 parts sugar, 1-2 parts salt, 7-8 parts milk powder, 20-30 parts egg liquid, 20-30 parts butter, and 30-40 parts water. The bread-making conditions are: dough fermentation temperature of 30-35℃, humidity of 80-90%, and fermentation time of 2.5-3 hours. After fermentation, the shaped dough is baked at 120-130℃ for 30-40 minutes. This invention uses buckwheat and quinoa as raw materials, and by optimizing fermentation conditions, improves the product's taste while retaining the nutrients and flavor of the raw materials, resulting in a nutritionally complete, fragrant, and soft whole-grain product that helps improve people's dietary structure and promote health.

[0005] Existing literature discloses methods for developing basic culture medium formulations using artificial intelligence algorithms, the steps of which include establishing a sample formulation database, training a machine learning model, and predicting the formulation effect (Publication No.: CN113450882B); another patent proposes a method for optimizing the preparation of plant extracts based on machine learning and multi-objective optimization algorithms (Publication No.: CN120089243A). However, these technologies are mainly aimed at specific scenarios such as culture media or plant extracts, and there is no targeted solution for the comprehensive optimization of rice and flour product formulations.

[0006] In summary, existing technologies suffer from the following key shortcomings: Current rice and flour product preparation technologies lack a systematic approach to integrating functional plant ingredients with floral aromas, plant pigments, and polyphenols into formula design; they also rely excessively on experience and single experimental methods, lacking AI-driven precision optimization methods. Therefore, a new method is urgently needed that can intelligently construct formulas based on different functional ingredients and combine advanced artificial intelligence algorithms to achieve precise optimization of rice and flour product formulas. Summary of the Invention

[0007] This invention addresses the shortcomings of existing technologies by proposing a method for optimizing the formulation of rice and flour products by combining neural network models and genetic algorithms. This method not only endows the products with a natural and pleasing color, enhancing their visual appeal, but also ensures color stability, effectively suppressing fading during processing and storage, and maintaining the product's attractive appearance.

[0008] The technical solution of the present invention is as follows: A method for precise optimization of rice and flour product formulas based on AI mainly includes the following steps: (1) Identify three categories of functional raw materials: plant materials with fragrance, plant materials with natural pigments, and plant materials rich in polyphenols; (2) The three types of functional raw materials determined in step (1) are freeze-dried and pulverized respectively to obtain functional plant raw material powder; (3) The three types of functional plant raw material powders obtained in step (2) are mixed with rice flour or wheat flour in a preset mass ratio to obtain a composite raw material powder for rice products or flour products; the mass addition ratio of the three types of functional plant raw material powders to rice flour or wheat flour is as follows: 0.1% to 5% of fragrant flower raw material powder, 0.1% to 5% of plant raw material powder with natural pigments, 0.1% to 5% of polyphenol-rich plant raw material powder, and 85% to 99.7% of rice flour or wheat flour, and the total mass ratio of the above three types of functional plant raw material powders to rice flour or wheat flour is 100%; (4) Use the rice or flour composite raw material powder obtained in step (3) to steam food to obtain rice or flour products; (5) Perform sensory evaluation on the quality of the rice and flour products obtained in step (4) and obtain the sensory evaluation results; (6) Using the mass ratio of the three types of functional plant raw material powders to rice flour and wheat flour in step (3) as input variables and the sensory rating results obtained in step (5) as output target variables, construct a sensory rating regression prediction model based on neural networks. (7) Using the sensory rating regression prediction model constructed in step (6) as the objective function, the genetic algorithm is used to perform global optimization on the input variable space, with the goal of maximizing the predicted value of the sensory rating, to solve the optimal quality ratio combination of the three types of functional plant raw material powder with rice flour or wheat flour. (8) Prepare rice and flour products according to the optimal ratio finally obtained in step (7), and perform food steaming processing and sensory evaluation to verify their quality and effect.

[0009] The fragrant flower raw material mentioned in step (1) is one of osmanthus and rose; the plant raw material with natural pigment is one of buddleja officinalis and dried tangerine peel; the plant raw material rich in polyphenols is one of green tea and eucommia leaves.

[0010] The rice products mentioned in steps (4) and (5) are rice cakes, and the flour products are flower-shaped steamed buns.

[0011] The sensory evaluation items mentioned in step (5) include appearance, color, smell and taste. Each item has a full score of 10 points, and each item is divided into three levels: 0-4 points is poor, 4-7 points is good, and 7-10 points is excellent. The sum of the scores of each item constitutes the comprehensive sensory evaluation result of rice and flour products.

[0012] The neural network model described in step (6) is the backpropagation neural network BPNN.

[0013] The beneficial effects of this invention are: (1) This invention provides a method for precise optimization of rice and flour product formulations based on AI technology, and for the first time introduces functional natural plant raw materials with specific floral fragrances, natural pigments and polyphenols into the formulation design. By introducing plant powders rich in natural pigments, not only are the products given a natural and pleasing color and their visual appeal enhanced, but the color is also stable, effectively inhibiting fading during processing and storage, and maintaining the product's attractive appearance.

[0014] (2) This invention establishes a mapping relationship between raw material ratios and sensory scores through neural network modeling, and then uses a genetic algorithm to globally optimize this mapping relationship, thereby achieving multi-objective comprehensive optimization of aroma, color, and taste. Compared with traditional trial-and-error methods and single-factor optimization methods, this invention can obtain rice and flour product formulas that meet specific sensory requirements more efficiently and accurately, satisfying personalized customization needs. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method for precise optimization of rice and flour product formulas based on AI technology, as described in this invention. Figure 2 The image shows the flavor distribution of the osmanthus rice cake recipe in Example 1 before and after optimization. Figure 3 The graphs show the textural analysis of the osmanthus rice cake formula in Example 1 before and after optimization. Figure 4 The image shows the flavor distribution of the rose-shaped steamed bun recipe in Example 2 before and after optimization. Figure 5 The graph shows the texture analysis of the rose-shaped steamed bun recipe in Example 2 before and after optimization. Detailed Implementation

[0016] The technical solution of the present invention will be further described below with reference to specific embodiments.

[0017] Example 1

[0018] A method for precisely optimizing the recipe of osmanthus rice cake with a unique osmanthus fragrance, appealing color, and rich polyphenols is as follows: (1) Identification and processing of plant materials: Fragrant flowers: Osmanthus was selected as the main source of fragrance. Plants with natural pigments: such as Figure 1 As shown in Table 1, from seven pigment-rich plant ingredients (Buddleja officinalis pollen, spinach powder, purple sweet potato powder, dried tangerine peel powder, pumpkin powder, beet powder, and strawberry powder), Buddleja officinalis pollen, with the most stable color, was selected as a natural pigment source to give the rice cake its orange-yellow color. For polyphenol-rich plant ingredients: as shown in Table 1, from nine candidate ingredients (cherry blossom powder, green tea powder, jujube peel powder, blueberry powder, dragon fruit powder, purple cabbage powder, Eucommia ulmoides leaf powder, eggplant peel powder, and purple mustard green powder), green tea powder, which has the highest total polyphenol retention rate in the rice cake, was selected as an additional polyphenol-providing material to enhance the antioxidant properties of the rice cake.

[0019] Table 1. Retention rate (%) of total polyphenols in rice cakes with different plant powders added during frozen storage. (2) Processing of plant raw materials: The above three types of plant raw materials are freeze-dried and then pulverized to 100 mesh fineness for later use.

[0020] (3) Preparation of Osmanthus Rice Cake Raw Material Powder: Staple Food Powder: High-quality rice flour is selected as the main ingredient. Initial Mixing Ratio: The initial setting is 1% for osmanthus, 2% for buddleja officinalis, 1% for green tea, and 96% for rice flour, totaling 100% (Note: This ratio is an initial setting and will be adjusted through subsequent optimization). Mix the above raw materials evenly according to the ratio to obtain Osmanthus Rice Cake Raw Material Powder.

[0021] (4) Osmanthus rice cake production and processing: Add 30% concentration of white sugar syrup to the osmanthus rice cake raw powder, stir, shape, and steam for 30 minutes until cooked through.

[0022] (5) Quality assessment of Osmanthus Rice Cake: Ten professional judges will score the appearance, color, smell and taste of the rice cake, with a maximum score of 10 points for each item.

[0023] Table 2 Sensory Evaluation Criteria (Rice Cake) (6) Neural network modeling: Collect different proportions of formulas and their corresponding sensory scores as data. After 600 rounds of training, establish a sensory score regression prediction model based on backpropagation neural network (BPNN). The 3D response surface and contour map of the interaction between the three functional raw material powders and rice flour and the sensory scores are shown.

[0024] (7) Global optimization using genetic algorithm: Using the neural network model as the objective function, a single-objective genetic algorithm is used for global optimization, with the objective being to maximize the sum of sensory scores. Through iterative search, the optimal proportions of each component of the raw material powder are finally obtained: osmanthus 3.7%, buddleja 3.1%, green tea 2.6%, and rice flour 90.6%.

[0025] (8) Experimental verification: Use the optimal ratio obtained in step (7) to prepare rice cake raw material powder, produce the optimal formula rice cake, and conduct sensory quality evaluation verification.

[0026] Flavor characteristics of rice cake before and after formulation optimization were assessed using 14 flavor sensors (Table 3) from an electronic nose analyzer and a texture property analyzer. Figure 2 ) and texture properties ( Figure 3 The sensory characteristics were characterized. The results showed that the optimal formula rice cake improved in all sensory indicators, especially in aroma and taste, with a richer osmanthus fragrance and a more delicate texture. The sensory scores of the original formula rice cake were: appearance 8.0, color 8.5, aroma 7.6, taste 7.5, and a total score of 31.6. The sensory scores of the optimal formula rice cake predicted by the neural network model were: appearance 8.5, color 8.9, aroma 9.4, taste 9.1, and a total score of 35.9. The actual results of the sensory quality verification of the optimal formula rice cake were: appearance 8.5, color 8.8, aroma 9.5, taste 9, and a total score of 35.8. This indicates that the established neural network model's prediction of the rice cake's sensory quality highly matches the actual verification results, with a total score error of only 0.1 points and deviations of less than 0.3 points for each sub-indicator, demonstrating excellent prediction accuracy.

[0027] Table 3 Array performance of electronic nose sensors Example 2

[0028] A method for precisely optimizing the recipe of flower-shaped steamed buns that are brightly colored, chewy, and rich in polyphenols is as follows: (1) Identification and processing of plant raw materials: Fragrant flowers: Roses were selected as the main source of fragrance. From seven plant raw materials rich in pigments (Buddleja officinalis pollen, spinach powder, purple sweet potato powder, dried tangerine peel powder, pumpkin powder, beet powder, and strawberry powder), dried tangerine peel powder, which has the most stable color, was selected as a natural pigment source to give the flower buns an orange-yellow color. Plant raw materials rich in polyphenols: As shown in Table 4, from nine candidate raw materials (cherry blossom powder, green tea powder, jujube peel powder, blueberry powder, dragon fruit powder, purple cabbage powder, Eucommia ulmoides leaf powder, eggplant peel powder, and purple mustard green powder), Eucommia ulmoides leaf powder, which has the highest total polyphenol retention rate in the flower buns, was selected as an additional source of polyphenols to improve the antioxidant properties of the flower buns.

[0029] Table 4. Retention rate (%) of total polyphenols in flower-shaped steamed buns with different plant powders added during frozen storage. (2) Processing of plant raw materials: The above three types of plant raw materials are freeze-dried and then pulverized to 80 mesh fineness for later use.

[0030] (3) Preparation of raw material powder for flower-shaped steamed buns: Staple food powder: High-quality wheat flour is selected as the main ingredient. Initial mixing ratio: The initial setting is that the addition ratio of rose petals is 1.5%, the addition ratio of dried tangerine peel is 5%, the addition ratio of Eucommia ulmoides leaves is 0.5%, and the addition ratio of wheat flour is 93%, with a total of 100% (Note: This ratio is a preliminary setting and will be adjusted through subsequent optimization). Mix the above raw materials evenly according to the ratio to obtain the raw material powder for flower-shaped steamed buns.

[0031] (4) Flower-shaped steamed bun production and processing: Add an appropriate amount of water to the flower-shaped steamed bun raw powder to form a dough, knead it into shape, and steam it until cooked.

[0032] (5) Quality assessment of flower buns: Ten professional judges will score the appearance, color, smell and taste of the flower buns, with a maximum score of 10 points for each item.

[0033] Table 5 Sensory Evaluation Criteria (Flower-shaped Steamed Buns) (6) Neural network modeling: Collect recipes with different proportions and their corresponding sensory ratings as data, and establish a sensory rating regression prediction model based on backpropagation neural network (BPNN) after 600 rounds of training.

[0034] (7) Global optimization using genetic algorithm: Using the neural network model as the objective function, a single-objective genetic algorithm is used for global optimization, with the objective being to maximize the total sensory score. Through iterative search, the optimal proportions of each component of the raw material powder are finally obtained (3.8% rose petals, 3.9% dried tangerine peel, 2.8% Eucommia ulmoides leaves, and 89.5% wheat flour).

[0035] (8) Experimental verification: Use the optimal ratio obtained in step (7) to prepare the flower bun raw material powder, produce the optimal formula flower bun, and conduct sensory quality scoring verification.

[0036] Flavor characteristics of the steamed buns before and after formula optimization were assessed using 14 flavor sensors (Table 3) from an electronic nose analyzer and a texture property analyzer. Figure 4 ) and texture properties ( Figure 5 The characteristics were analyzed. Results showed that the optimal recipe for steamed flower buns significantly improved in appearance, color, aroma, and taste, with a richer rose aroma and a more appealing color from the dried tangerine peel, resulting in a substantial improvement in overall quality. The sensory scores for the original recipe were: appearance 7.4, color 7.8, aroma 8.2, taste 8.6, and a total of 32.0. The sensory scores for the optimal recipe predicted by the neural network model were: appearance 9.2, color 9.5, aroma 8.8, taste 9.3, and a total of 36.8. The actual results of the sensory quality verification of the optimal recipe were: appearance 9.0, color 9.6, aroma 8.9, taste 9.1, and a total of 36.6. This indicates that the established neural network model's prediction of the sensory quality of the steamed buns is highly consistent with the actual verification results, with a total score error of only 0.2 points and deviations of each sub-index of less than 0.3 points, demonstrating excellent prediction accuracy.

Claims

1. A method for precise optimization of rice and flour product formulas based on AI, characterized in that, The main steps include: (1) Identify three categories of functional raw materials: plant materials with fragrance, plant materials with natural pigments, and plant materials rich in polyphenols; (2) The three types of functional raw materials determined in step (1) are freeze-dried and pulverized respectively to obtain functional plant raw material powder; (3) The three types of functional plant raw material powders obtained in step (2) are mixed with rice flour or wheat flour in a preset mass ratio to obtain a composite raw material powder for rice products or flour products; the mass addition ratio of the three types of functional plant raw material powders to rice flour or wheat flour is as follows: 0.1% to 5% of fragrant flower raw material powder, 0.1% to 5% of plant raw material powder with natural pigments, 0.1% to 5% of polyphenol-rich plant raw material powder, and 85% to 99.7% of rice flour or wheat flour, and the total mass ratio of the above three types of functional plant raw material powders to rice flour or wheat flour is 100%; (4) Use the rice or flour composite raw material powder obtained in step (3) to steam food to obtain rice or flour products; (5) Perform sensory evaluation on the quality of the rice and flour products obtained in step (4) and obtain the sensory evaluation results; (6) Using the mass ratio of the three types of functional plant raw material powders to rice flour and wheat flour in step (3) as input variables and the sensory rating results obtained in step (5) as output target variables, construct a sensory rating regression prediction model based on neural networks. (7) Using the sensory rating regression prediction model constructed in step (6) as the objective function, the genetic algorithm is used to perform global optimization on the input variable space, with the goal of maximizing the predicted value of the sensory rating, to solve the optimal quality ratio combination of the three types of functional plant raw material powder with rice flour or wheat flour. (8) Prepare rice and flour products according to the optimal ratio finally obtained in step (7), and conduct food steaming processing and sensory evaluation to verify their quality effect.

2. The method for precise optimization of rice and flour product formulas based on AI according to claim 1, characterized in that, The fragrant flower raw material mentioned in step (1) is one of osmanthus and rose; the plant raw material with natural pigment is one of buddleja officinalis and dried tangerine peel; the plant raw material rich in polyphenols is one of green tea and eucommia leaves.

3. The method for precise optimization of rice and flour product formulas based on AI according to claim 1, characterized in that, The fineness of the three types of functional plant raw material powders mentioned in step (2) is 50-200 mesh.

4. The method for precise optimization of rice and flour product formulas based on AI according to claim 1, characterized in that, The rice products mentioned in steps (4) and (5) are rice cakes, and the flour products are flower-shaped steamed buns.

5. The method for precise optimization of rice and flour product formulas based on AI according to claim 1, characterized in that, The sensory evaluation items mentioned in step (5) include appearance, color, smell and taste. Each item has a full score of 10 points, and each item is divided into three levels: 0-4 points is poor, 4-7 points is good, and 7-10 points is excellent. The sum of the scores of each item constitutes the comprehensive sensory evaluation result of rice and flour products.

6. The method for precise optimization of rice and flour product formulas based on AI according to claim 1, characterized in that, The neural network model described in step (6) is the backpropagation neural network BPNN.

Images