Method and system for optimizing edible oil sample blending formula and edible oil sample predicted thereby

Abstract

According to an embodiment of the present invention, a method for optimizing a blending formula of an edible oil sample is provided. The method includes: receiving a data set formed by a plurality of numerical values, wherein the numerical values ​​represent at least one indicator type obtained from a plurality of pure oils and / or blended oils; and generating a prediction for an optimized blending formula from the data set formed by the received plurality of numerical values. The step of generating a prediction for an optimized blending formula includes: using a prediction model, the prediction model is configured to generate a plurality of blending formulas based on the data set formed by the received plurality of numerical values, evaluating at least one indicator value for each blending formula, and comparing the evaluated at least one indicator value with at least one predetermined threshold value to determine an optimized blending formula. According to other embodiments, a system for optimizing a blending formula is also provided.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to Singapore Patent Application No. 10202000216P filed on January 9, 2020, the entire contents of which are incorporated herein by reference for all purposes. Technical Field

[0003] Various embodiments relate to methods for optimizing the blending recipe of an edible oil sample, apparatus or systems therefor, and edible oil samples having a blending recipe predicted using the methods of various embodiments. Background Art

[0004] Blended oil formulations for various culinary purposes have been extensively explored.

[0005] The most commonly used frying oils include pure oils, such as palm oil, sunflower oil, soybean oil, high oleic sunflower oil, etc. Due to their specific chemical and physical properties, pure vegetable oils have limited industrial applications. Therefore, blended oils formed by mixing different types of oils are commonly used (especially in the field of fried foods). Frying oils are generally required to be stable for a long time at high temperatures, produce as few toxic chemicals as possible at high temperatures, provide good flavor, and be reasonably priced or affordable. However, providing a blended oil that meets all of the above requirements has always been difficult and challenging.

[0006] Conventional approaches to creating or adjusting oil formulations involve laboratory experiments or calculations using spreadsheets (e.g., in Microsoft Excel). However, such approaches can result in a high degree of inaccuracy and can be inefficient. Computational methods for optimization have also been explored, but only a few prior publications have addressed the application of deep learning to design and optimize blended oil formulations. The types of algorithms explored are limited and can differ in many factors, such as the input information, constraints, and / or desired outputs during the optimization process.

[0007] Previous studies have reported some characteristics of blended oils based on oil availability, cost, and simple calculations of their fatty acid profiles. However, these studies have not considered the scope for blending large quantities of pure oils and / or starting blends, nor have they considered limitations relevant to the deep-frying industry, which uses total polar compounds (TPC) as a measure of suitability for sustained frying.

[0008] Therefore, there is a need for a method and system for optimizing the blending formula of edible oil samples to at least solve the above problems and comply with the restrictions imposed on frying oil parameters by industry needs. Summary of the Invention

[0009] According to one embodiment, a method for optimizing a blending recipe of an edible oil sample is provided. The method may include: receiving a plurality of numerical values ​​to form a data set, wherein the numerical values ​​represent at least one index type and are derived from a plurality of pure oils and / or blended oils; and generating a prediction for an optimized blending recipe based on the data set formed from the received numerical values. The step of generating the prediction for the optimized blending recipe includes: using a prediction model configured to generate a plurality of blending recipes based on the received data set of the plurality of numerical values, evaluating at least one index value in each blending recipe, and comparing the evaluated at least one index value to at least one predetermined threshold value to determine an optimized blending recipe from the generated plurality of blending recipes, wherein the at least one predetermined threshold value and the at least one index value have the same index type, and the at least one predetermined threshold value is industry-recognized.

[0010] According to one embodiment, edible oil samples predicted using the methods of various embodiments are provided.

[0011] According to one embodiment, a computer-readable storage medium is provided, which contains computer-readable instructions, which, when executed by a computer, are operable to optimize a blending recipe of an edible oil sample. The computer-readable instructions can be configured to perform the methods of various embodiments.

[0012] According to one embodiment, a device or system is provided. The device or system may include: a receiving unit configured to receive a plurality of numerical values ​​forming a data set, wherein the numerical values ​​represent at least one indicator type and are derived from a plurality of pure oils and / or blended oils; a memory configured to store a prediction model, wherein the prediction model is configured to generate a plurality of blending recipes based on the received data set formed by the numerical values, evaluate at least one indicator value in each blending recipe, and compare the evaluated at least one indicator value with at least one predetermined threshold value to determine an optimized blending recipe from the generated plurality of blending recipes, wherein the at least one predetermined threshold value and the at least one indicator value have the same indicator type, and the at least one predetermined threshold value is industry-recognized; and a processor configured to access the prediction model stored in the memory to execute the steps of the method of various embodiments to generate a prediction of an optimized blending recipe for an edible oil sample. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Like reference numerals generally refer to like parts between the different views in the drawings. The drawings are not necessarily to scale, but emphasis is placed upon generally illustrating the principles of the invention. In the following description, various embodiments of the invention are described with reference to the following drawings, in which:

[0014] Figure 1A A flow chart is shown illustrating a method for optimizing a blending recipe of an edible oil sample, according to various embodiments.

[0015] Figure 1B A schematic diagram of an apparatus or system 120 for optimizing a blending recipe of an edible oil sample is shown, according to various embodiments.

[0016] Figure 2A A schematic flow chart is shown illustrating an overview method for designing or optimizing an edible oil blend, according to various embodiments.

[0017] Figure 2B The flowchart shown illustrates the workflow of the genetic algorithm according to various embodiments.

[0018] Figure 3A The chart shown illustrates how different types of neat oils perform in frying.

[0019] Figure 3B Shows Figure 3A Representative diagram of pure oils, indicating corresponding frying life, saturated fatty acid content, and price according to various examples.

[0020] Figure 4A The graph shown illustrates experimentally measured TPC values ​​for blends that meet predicted frying performance, according to various embodiments.

[0021] Figure 4B The graph shown illustrates experimentally measured TPC values ​​of blended oils according to various embodiments, using palm oil as a positive control and a negative control.

[0022] Figure 4C Shown is a graph illustrating the relationship between the acid value of blended oils and time, according to various embodiments, compared to a control.

[0023] 5A to 5I The graphs shown illustrate the relationships between different oxidation indicators according to various embodiments. DETAILED DESCRIPTION

[0024] The following detailed description refers to the accompanying drawings, which show, by way of illustration, specific details and embodiments in which the present invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the present invention. Other embodiments may be utilized and changes may be made without departing from the scope of the present invention. The various embodiments are not necessarily mutually exclusive, as some embodiments may be combined with one or more other embodiments to form new embodiments.

[0025] Embodiments described in the context of one of the method or apparatus are analogously valid for the other method or apparatus / device. Similarly, embodiments described in the context of the method are analogously valid for the apparatus, and vice versa.

[0026] Features described in the context of a particular embodiment may be applied accordingly to the same or similar features in other embodiments. Features described in the context of a particular embodiment may be applied accordingly to other embodiments even if not explicitly described in these other embodiments. Furthermore, additions and / or combinations and / or alternatives described for a feature in the context of a particular embodiment may be applied accordingly to the same or similar features in other embodiments.

[0027] In the context of various embodiments, the articles "a," "an," and "the" when used with a feature or element include reference to one or more features or elements.

[0028] In the context of various embodiments, the phrase "substantially" can include "exactly" and reasonable variations.

[0029] In the context of the various embodiments, the terms "about" or "approximately" as applied to numerical values ​​encompass the exact value and a reasonable variance.

[0030] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0031] As used herein, a phrase of the form "at least one of A or B" may include A or B or both A and B. Accordingly, a phrase of the form "at least one of A or B or C," or including further listed items, may include any and all combinations of one or more of the associated listed items.

[0032] Various embodiments may provide methods and systems for optimizing edible oil blends.

[0033] The method for optimizing edible blended oil and the device or system used therefor according to various embodiments can also provide a formula of blended oil (especially frying blended oil) that meets several standards.For example, one or more blended oils can be provided.

[0034] The method may include: receiving the fatty acid composition of pure oil and / or blended oil, the price of pure oil and / or blended oil, and the total polar compound (TPC) value of the pure oil and / or blended oil during a 24-hour heating period; and providing a prediction model capable of generating at least one blended oil.

[0035] The device or system may include: a receiving unit configured to receive the fatty acid composition of pure oil and / or blended oil, the price of pure oil and / or blended oil and / or the TPC value of pure oil and / or blended oil during a heating period (e.g., 24 hours); a memory for storing a prediction model, which is capable of generating a prediction for optimizing the blended oil; and a processor configured to access the prediction model stored in the memory to perform the steps of the method according to various embodiments, for generating a prediction for optimizing the blended oil (e.g., frying oil), and according to various embodiments, for generating a blended oil recipe.

[0036] Figure 1A The flowchart shown illustrates a method for optimizing a blending recipe of an edible oil sample 100, according to various embodiments. Figure 1A In the process, a plurality of values ​​are received to form a data set, wherein the values ​​represent at least one indicator type obtained from a plurality of pure oils and / or initial blended oils. In step 104, a prediction for an optimized blending recipe is generated based on the data set formed by the received values.

[0037] The method 100 may also include generating predictions of parameters of an optimized blend recipe for the edible oil sample.

[0038] In other words, method 100 can be a method for predicting an optimized blending formula of an edible oil sample using a prediction model, wherein the numerical values ​​processed by the prediction model in the prediction represent one or more indicators of multiple pure oils and / or blended oils. Parameters of the optimized blending formula can also be predicted using the prediction model.

[0039] In various embodiments, the predicted parameters can represent predicted blend properties within a specific timeframe. The predicted parameters can also indicate the quality level of the edible oil sample. For example, the parameters can include the frying life of an optimized blend recipe. The frying life can be verified by conducting laboratory experiments.

[0040] In various embodiments, generating a prediction for an optimized blend recipe at step 104 may include: using a prediction model configured to generate a plurality of blend recipes based on a data set formed from received numerical values, evaluating at least one indicator value in each blend recipe, and comparing the evaluated at least one indicator value with at least one predetermined threshold value to determine an optimized blend recipe from the generated plurality of blend recipes. The at least one predetermined threshold value and the at least one indicator value may be of the same indicator type. The at least one predetermined threshold value may be industry-recognized. For example, the at least one predetermined value may be an industry standard value or an industry recommended value.

[0041] In various embodiments, the prediction model can be a genetic model constructed by at least one feature vector. Each feature can correspond to an initial index value obtained from a specific pure oil or blended oil in multiple pure oils and / or blended oils, or an index value in at least one index value. The initial index value may include the frying life of a specific pure oil or blended oil, the fatty acid content of a specific pure oil or blended oil, the price of a specific pure oil or blended oil, and at least one of the total polar compounds (TPC) values ​​produced by a specific pure oil or blended oil during cooking. In other words, the prediction model can be optimized by at least one feature vector, or, in the case where there are multiple vectors, the prediction model can be trained by a feature matrix. The index value in the at least one index value can be based on data samples (data examples), which may include at least one of the total polar compounds (TPC) values ​​produced by the blended oil of prediction, the fatty acid content of the blended oil of prediction, the price of the blended oil of prediction, and the total polar compounds (TPC) value produced by the blended oil of prediction during cooking.

[0042] The prediction model may be based on a non-hybridized, independent genetic algorithm. In various embodiments, the genetic model may include an input portion configured to receive a data set formed of a plurality of numerical values, a genetic algorithm configured to determine an optimal blend recipe based on an objective function, and an output portion comprising a plurality of outputs configured to present the optimal blend recipe as an optimized blend recipe. The genetic algorithm may also be configured to evaluate parameters of the optimized blend recipe, and the plurality of outputs may be configured to present the parameters substantially simultaneously at the output portion.

[0043] In various embodiments, at least one type of indicator may indicate characteristics of a plurality of pure oils and / or blended oils. For example, the type of indicator may include an oxidation indicator, a nutritional indicator, or a market indicator.

[0044] Oxidation indices may include: total polar compound (TPC) value, fry life, anisidine value, carbonyl value, thiobarbituric acid level, conjugated dienoic acid value, conjugated trienoic acid value, or combined conjugated dienoic and conjugated trienoic acid value, reduced polyunsaturated fatty acid content, polymer byproduct value. For example, the TPC value may refer to the TPC value over a 24-hour heating period.

[0045] Nutritional indices may include: fatty acid content or structured lipid levels.

[0046] The market index may include prices. It should be recognized and understood that the prices shown in Table 2 below are provided as examples only and should not be interpreted in a limiting sense.

[0047] For example, an optimized blend formula meets at least one of the following requirements:

[0048] (i) lowest price;

[0049] (ii) constraints imposed by at least one predetermined threshold; and

[0050] (iii) Longest frying life.

[0051] In other words, the optimized blend recipe may have the lowest price, or may comply with constraints imposed by at least one predetermined threshold, or may have the longest fry life, or any combination thereof.

[0052] In various embodiments, method 100 may further include generating a data set formed from a plurality of values. Generating the data set may include: linearly combining first indicator values ​​obtained from a plurality of pure oils and / or blended oils to form a landscape of second indicator values ​​for the plurality of pure oils and / or blended oils; applying a sampling algorithm to the landscape to sample the second indicator values; and forming at least a portion of the data set based on the sampled second indicator values. The first indicator value and the second indicator value may be of at least one indicator type, or may be based on the same indicator type.

[0053] For example, the sampling algorithm may be based on the Dirichlet distribution.

[0054] The step of generating a data set formed of numerical values ​​may be performed before step 102 .

[0055] In various embodiments, the step of receiving a data set formed by numerical values ​​representing at least one indicator type obtained from multiple pure oils and / or blended oils in step 102 may include: receiving a data set representing numerical values ​​of more than one indicator type obtained from multiple pure oils and / or blended oils, wherein the more than one indicator type includes an oxidation index and a nutritional index, or an oxidation value, a nutritional index and a market index.

[0056] The steps of generating the prediction at step 104 and generating the prediction for the parameters of the optimized blend recipe may be in real time.

[0057] Although the above method is illustrated and described as a series of steps or events, it should be understood that any order of these steps or events should not be interpreted in a restrictive sense. For example, in addition to those illustrated and / or described herein, some steps can occur in different orders and / or occur simultaneously with other steps or events. In addition, it is not required that all steps of the description be performed to perform one or more aspects or embodiments described herein. Similarly, one or more steps described herein can be performed with one or more separate operations and / or stages.

[0058] Various embodiments may provide edible oil samples predicted using the method 100 according to various embodiments. The edible oil samples may be blended oil, blended frying oil, or blended seasoning oil.

[0059] Various embodiments may also provide a computer-readable storage medium containing computer-readable instructions that, when executed by a computer, are operable to optimize a blending recipe for an edible oil sample. According to various embodiments and / or as described herein, the computer-readable instructions are configured to perform method 100.

[0060] Figure 1B A schematic diagram of an apparatus or system 120 for optimizing a blending recipe of an edible oil sample is shown in accordance with various embodiments. Figure 1B In the embodiment, the device or system 120 includes: a receiving unit 122 configured to receive a data set formed by a plurality of numerical values, wherein the numerical values ​​represent at least one indicator type, which are obtained from a plurality of pure oils and / or blended oils; a memory 124 for storing a prediction model; and a processor 126 configured to access the prediction model stored in the memory 124 to execute the method 100 ( Figure 1A ) to generate a prediction of an optimized blending recipe for the edible oil sample. Processor 126 may also be configured to access a prediction model stored in memory 124 to execute steps of method 100 to further generate a prediction of parameters of the optimized blending recipe for the edible oil sample. The predicted parameters may represent predicted post-blending properties within a specific timeframe or may indicate a quality level of the edible oil sample.

[0061] The receiving unit 122, the memory 124, and the processor 126 may communicate with each other as shown by lines 128, 130. This communication may be bidirectional.

[0062] Device or system 120 may include Figure 1A Those same or similar elements or components as those described in the method 100, whereby similar elements may be described in Figure 1A The present invention is not described in the context of method 100, and thus the corresponding description is omitted here.

[0063] In various embodiments, the apparatus or system 120 may further include a pre-processor unit configured to: perform a linear combination of first indicator values ​​obtained from a plurality of neat oils and / or blended oils to form a distribution relationship graph comprising second indicator values ​​for the plurality of neat oils and / or blended oils; and apply a sampling algorithm to the distribution relationship graph to sample the second indicator values ​​to form at least a portion of a numerical data set based on the sampled second indicator values. The first indicator value and the second indicator value have at least one indicator type.

[0064] For example, the pre-processor unit may form an integral part of the processor and the memory may be further configured to store the sampling algorithm.

[0065] Examples are described below in the form of experiments that were performed to better understand the method 100 and the apparatus or system 120 .

[0066] Designing or optimizing blended oils for different cooking purposes is a challenging task. Various embodiments may provide for the integration of deep learning and various laboratory techniques in the field of cooking oils to design and optimize blended oils.

[0067] Figure 2A A schematic flow chart 200 is shown illustrating an overview method for designing or optimizing an edible oil blend 208 according to various embodiments. The method includes steps 201, 202 and / or 203 and / or 204, 205, 206, and 207.

[0068] Figure 2A The overview method can be compared with Figure 1A The method 100 is similarly described.

[0069] exist Figure 2A In step 201, raw data are used as input to the algorithm, and the raw data include: the fatty acid composition of the pure oil and / or blended oil, the price of the pure oil and / or blended oil, and the total polar compound (TPC) value of the pure oil and / or blended oil during the 24-hour heating period.

[0070] Figure 3AGraph 300 is shown illustrating the performance of different types of neat oils for frying. "Value" represents the percentage of TPC produced after the oil has been heated for approximately 24 hours. "Time" is measured in hours. A linear model is used to estimate the TPC value during frying. Frying life is determined as the length of time it takes for the oil to reach a TPC value of 27%, which is generally considered the threshold for unhealthy frying oils. In the examples, RSO represents rapeseed oil; RBO represents rice bran oil; POL represents palm oil; SFO represents sunflower oil, and HOSFO represents high oleic sunflower oil; SBO represents soybean oil; and MZO represents corn oil (maize oil).

[0071] Figure 3B A representative schematic diagram of pure oil (in a bottle) is shown, with the corresponding frying life in hours (i.e., the number of hours it takes for the oil to reach 27% TPC when heated at 180°C), the corresponding saturated fatty acid content (SFA) in percentage, and the price in Chinese Yuan (CNY). It should be understood that the critical values ​​for temperature and TPC should not be construed as limiting. Oils priced below 8,000 CNY / MT (RMB / metric ton) have a lower frying life, less than 20 hours, and generally have a lower SFA. Oils with longer frying times (over 20 hours) are much more expensive, at around 14,000 CNY / MT. For example, palm oil has a relatively long frying life of 25 hours and is relatively inexpensive. However, palm oil has a high SFA of 44%.

[0072] Figure 2A The purpose of the method described in is to determine the blend oil "B" (such as Figure 3B shown) and their frying life, SFA, and price.

[0073] Reference Figure 2A, step 202, the space of blended oil is evaluated for the price of pure oil or blended oil. Using a sampling algorithm based on Dirichlet distribution, the distribution relationship diagram of blended oil price is investigated in the form of a linear combination of pure oil or blended oil price. The numerals at 202 in the figure show the price space of blending possibility, with bright areas representing high-price blending and dark areas representing low-price blending. In step 203, the space of fatty acid composition (FAC) of all pure oils or blended oils is evaluated. Using a sampling algorithm based on Dirichlet distribution, the distribution relationship diagram of blended oil FAC is investigated in the form of a linear combination of pure oil or blended oil fatty acid composition. In the figure, 203 places show the bright area representing pure oil or blended oil with high saturated fatty acids and the dark area representing pure oil or blended oil with low saturated fatty acids. It should be understood that other types of fatty acids, for example, monounsaturated, polyunsaturated, n-3 and n-6 ​​fatty acids also can be evaluated in this step. In step 204, the TPC values ​​of the pure oils or blends after 24 hours of heating are first fitted to a linear model to estimate the frying life. Frying life is defined as the length of time it takes for the oil to reach a TPC value of 27. This can be easily estimated using a linear model. Subsequently, the space of blends is evaluated for frying life. A sampling algorithm based on the Dirichlet distribution is used to examine a frying life distribution graph representing linear combinations of the frying lives of the pure oils. The graph at 204 shows bright areas representing pure oils or blends with high frying lives and dark areas representing pure oils or blends with low frying lives.

[0074] A sampling algorithm is used to achieve sampling of the search space, which is typically large, as shown at 202, 203 and 204, and may conflict with each other, such as low price corresponding to an undesirable saturated fat content.

[0075] In step 205, the subsamples of the blended oils in 202, 203 and 204 are evaluated to see whether they meet industrial requirements, such as controlled saturated fatty acids (SFA), medium-chain unsaturated fatty acids (MUFA) and polyunsaturated fatty acids (PUFA) percentages. Industrial requirements provide constraints related to frying oil parameters. In other words, initial suggestions that meet the constraints related to frying oil parameters given by industrial requirements can be provided by sampling the search space. For example, it is generally believed that SFA is not good or unhealthy, while the ratio of MUFA:PUFA is also important for nutritional balance. Certain fatty acids can bring adverse flavors and thus may also be controlled. Frying life can be controlled by monitoring the TPC of the blended oil.

[0076] The blend at 205 may meet the requirements, but is not optimal in terms of price difference. However, the blend at 205 may provide a starting point for optimization.

[0077] The data at 202-206 is used as input to a genetic algorithm at 207 to find the optimal blend based on an objective function that combines the qualities of price, SFA, MUFA, PUFA, and frying life. The optimal blend at 208 (and as described herein) is the blend that has the lowest price, meets the SFA / MUFA / PUFA requirements, and has the longest frying life.

[0078] In other words, further improvements to the oil recipe (e.g., the blend at 205) are optimized based on a genetic algorithm (e.g., 207), which iteratively mutates each recipe to find the best solution over several generations. The best solution also maintains frying quality the longest over several generations, as indicated by the time to high TPC values. This can be verified experimentally or through quick testing, such as with a separate deep learning NIR (near infrared) model.

[0079] The genetic algorithm (e.g. 207) can be based on Figure 2B The workflow 220 shown in the flowchart is operated. Figure 2B As shown, in general, workflow 220 may begin with an initial population 222. At 224, the fitness of the population, which may include the initial population 222 and / or the resulting new population 232, may be determined using an objective function, which will be described in detail below. Mate selection and crossover may be performed at 226 and 228, respectively. This may be followed by mutation at 230, which may then create a new population 232. If the new population 232 does not meet the stopping criteria 234, the new population 232 may be fed back to the adaptation step at 224. This process may continue until the new population 232 eventually meets the stopping criteria 234, and a result 236 (in this case, a predicted blending recipe) may be obtained.

[0080] Figure 1A and 2A The method described can be a method for designing or optimizing more than one edible oil blend. In other words, the modeling of multiple blend results can be provided in a single model.

[0081] The predictive model, or interchangeably referred to as a genetic model, may utilize an objective function wherein the formula reflects the best fit of the blend, wherein the higher the best fit value, the closer it is to the optimal blend as reflected by having the lowest price, meeting the fatty acid requirements, and having the longest frying life. More specifically, the model evaluates generations of blends, wherein the initial population may be constructed from recommendations (e.g., based on Figure 2A202-204). In each generation, the blend with the highest fitness, as assessed by a formula reflecting the lowest price, satisfying the SFA / MUFA / PUFA constraints, and having the longest frying life, is derived into the next generation. In this case, the blend ratio is slightly altered to allow for the generation of offspring from the most adaptable blend of the previous generation. Offspring here refers to a blend that is a mixture of the two parent blends. At generation n, the model can identify the optimal blend based on the defined objective function.

[0082] The objective function may consist of multiple parts that must be optimized.

[0083] Using Equation 1, it may be necessary to minimize the total cost of a blended oil formulation, where the cost of each oil type can be predefined:

[0084]

[0085] It may be necessary to minimize saturated fat by using Formula 2:

[0086]

[0087] It may be necessary to maximize monounsaturated fat by using Formula 3:

[0088]

[0089] It may be necessary to minimize polyunsaturated fats by using Formula 4:

[0090]

[0091] The c18:3 fatty acids that contribute to the unpleasant fried taste can be minimized by using Formula 5:

[0092]

[0093] It may be necessary to use Equation 6 to maximize the fry life defined as the time taken to reach a TPC value of 27%, where and Indicates different TPC-related indicators:

[0094]

[0095] In each of Formulas 1 to 6, N represents the total number of oil types, and x represents a feature corresponding to an initial index value obtained from a specific pure oil or blended oil or an index value obtained from a predicted blending recipe. For example, the initial index value may include at least one of the frying life of a specific pure oil or blended oil, the fatty acid content of a specific pure oil or blended oil, the price of a specific pure oil or blended oil, and the total polar compound (TPC) value produced by the specific pure oil or blended oil during cooking, while the index value may include at least one of the predicted frying life of the blended oil, the predicted fatty acid content of the blended oil, the predicted price of the blended oil, and the predicted total polar compound (TPC) value produced by the blended oil during cooking. It should be recognized and understood that the initial index value may change over time, which is described herein as an example.

[0096] The boundary conditions can be modified as required. The boundary condition for fat can be, for example, S(x i )<25,M(x i )<50, P(x i ) < 25. In other examples, the boundary condition for the inventory sample can be the total amount of oil used to scale up from the blending ratio to the inventory amount. These boundary conditions may vary with oil price fluctuations and are described here as an example. It should be appreciated and understood that other boundary conditions may be considered, even if not described here, and that these boundary conditions may also vary with oil price fluctuations.

[0097] Experimental Group I

[0098] The characteristics of the frying oil recipes obtained by running the above genetic model are shown in Table 1, and:

[0099] (i) the percentage of each oil type may vary within the range of 5%, 4%, 3%, 2%, 1%, 0.5% or 0.1% by weight; and / or

[0100] (ii) These formulations can meet the set standards in line with industry requirements.

[0101] In one embodiment, the properties of the optimized formulations are shown in Table 1 below, where B1 to B4 refer to four different blending oil formulations.

[0102] Table 1

[0103]

[0104] Table 2 below lists the blending formulas of B1 to B4, compared with the negative controls N1 to N6, as well as the content of each oil component in each blending oil formula of B1 to B4 and N1 to N6 (the sum is 1), and the price (in CNY / MT), SFA, MUFA, PUFA, and frying life (in hours) of the resulting blending oil. Table 2 also reflects the price (in CNY / MT) of each oil component.

[0105] Table 2

[0106]

[0107] Laboratory experiments were conducted to examine the frying performance of formulations B1, B2, B3 and B4. The results are shown in Figure 4A middle. Figure 4A A graph 400 is shown illustrating experimentally measured TPC values ​​for blends (B1 to B4) that met predicted frying performance. Palm oil was used as a control. The heating time corresponded to the frying time.

[0108] like Figure 4A As shown, formulations B1, B2, B3, and B4 showed similar frying performance to palm oil and had a better frying base than other blends in the prior art. This suggests that the model can find blends with similar performance at a much lower price. Figure 4B A graph 402 is shown illustrating experimentally measured TPC values ​​for blends (B1 to B4) with negative controls (N1 to N6). The model calculation formula is then used to randomly select the TPC that was observed to increase faster to provide N1 to N6. Figure 4C A graph 404 is shown, illustrating the relationship between the acid value and heating time of blended oils (B1 to B4), with N1 to N6 used as a control. It is observed that B1-B4 have similar performance to palm oil and exhibit better performance than N1-N4, which have a shorter pot life. It is also observed that B1-B4 have similar performance to N5 and N6, but are priced lower. Figure 4B and Figure 4C Based on experimental results.

[0109] Experimental Group II

[0110] As a measure of the total polar compound (TPC) value of pure oil and / or blended oil (e.g. Figure 2A Alternatively to the method of measuring the oxidation stability of the polymer, different oxidation indices may be used, for example, acid value (AC), peroxide value (PV), p-anisidine value (PAV), color (COL), conjugated dienes (CD), conjugated trienes (CT), viscosity, carbonyl groups, polymer (or polymer by-products) and oxidative stability index (OSI) may be considered.

[0111] The basis for expanding the application of different oxidation indices is the existence of linear relationships between them. For example, in the frying test, the relationship between different oxidation indices was evaluated. Figure 5A Shows a graph illustrating the relationship between carbonyl groups and polymer byproducts, Figure 5B Displays a graph illustrating the relationship between PAV and polymer byproducts, Figure 5C Displays a graph illustrating the relationship between TPC and AC, Figure 5D Display a graph illustrating the relationship between polymer and AC, Figure 5E Show a graph illustrating the relationship between viscosity and AC, Figure 5F Display a graph illustrating the relationship between CD and PAV, Figure 5G Show a diagram illustrating the relationship between CD and carbonyl groups, Figure 5H Displays a graph illustrating the relationship between CD and polymer, Figure 5I Shows a diagram illustrating the relationship between OSI and TPC.

[0112] Table 3 shows the Figures 5A to 5I The linear correlation regression coefficients (R 2 ).

[0113] Table 3

[0114]

[0115] Further experiments were conducted using soybean oil and blended frying oils, taking into account different oxidation indices. The frying test results in Tables 4 to 6 below demonstrate a linear correlation between the various oxidation indices during the actual frying process. This pattern was observed not only in pure oils but also in blended frying oils. Therefore, those skilled in the art will appreciate that any one oxidation index, or a combination of multiple oxidation indices, can be used in the present invention to characterize the degree of oxidation and are within the scope of this invention.

[0116] Those skilled in the art will also appreciate that the model of the present invention can be used to optimize blending of different pure oils. It can also be used to optimize an initial semi-finished blended oil, allowing it to be blended with other pure oils or other blended oils to produce an optimized blended oil. Table 4 shows the different oxidation indices of soybean oil when used to cook French fries, noodles, chicken wings, and noodles (with oil added).

[0117] Table 4

[0118]

[0119]

[0120] Table 5 shows the different oxidation indices of blended oil.

[0121] Table 5

[0122]

[0123] Table 6 shows the different oxidation indices of five other blended oil samples, including the oxidation state index.

[0124] Table 6

[0125]

[0126]

[0127] Although the present invention has been particularly shown and described with reference to certain embodiments, it will be understood by those skilled in the art that various changes in form and details may be made without departing from the spirit and scope of the invention as defined in the appended claims. The scope of the present invention is therefore indicated by the appended claims, and it is therefore intended to include all variations that fall within the equivalent meaning and range of the claims.

Claims

1. A method for optimizing the blending formula of an edible oil sample, the method comprising: receiving a data set formed of a plurality of values, wherein the values ​​are obtained from a plurality of pure oils and / or blended oils and represent more than one indicator type, wherein the more than one indicator type comprises at least an oxidation indicator and a market indicator, wherein the oxidation indicator comprises at least frying life, and the market indicator comprises price; generating a plurality of blending recipes based on the data set using a prediction model, wherein the prediction model is a genetic model constructed from at least one feature vector, the feature vector corresponding to at least an initial index value obtained from a specific pure oil or blended oil among the plurality of pure oils and / or blended oils, the initial index value including at least a frying life and a price of the specific pure oil or blended oil; and At least one indicator value in each of the blending recipes is evaluated and compared with at least one predetermined threshold value to determine an optimized blending recipe from the generated plurality of blending recipes, wherein the at least one predetermined threshold value and the at least one indicator value have the same indicator type and are industry-recognized, and the optimized blending recipe at least meets the minimum price and the longest frying life.

2. The method according to claim 1, wherein The feature vector also corresponds to an index value among the at least one index value.

3. The method according to claim 2, wherein: The initial index value also includes a value for the total polar compounds produced during cooking of the particular neat oil or blend.

4. The method according to claim 2 or 3, wherein: The genetic model includes an input portion configured to receive a data set of values, configured as a genetic algorithm to determine an optimal blending recipe based on an objective function; and an output portion including a plurality of outputs configured to present the optimal blending recipe as an optimized blending recipe.

5. The method according to claim 4, wherein: The genetic algorithm is further configured to evaluate parameters of the optimized blend recipe, the evaluated parameters representing predicted blended properties within a specific timeframe, and the plurality of outputs are further configured to present the parameters at the output portion substantially simultaneously.

6. The method of claim 1 , further comprising generating a prediction of a parameter of the optimized blend recipe for the edible oil sample, the predicted parameter representing a predicted post-blending property within a specified time frame.

7. The method of claim 1, wherein: The more than one indicator types also include nutritional indicators.

8. The method of claim 7, wherein: The oxidation index also includes: a total polar compound value, an anisidine value, a carbonyl value, a thiobarbituric acid level, a conjugated dienoic acid value, a conjugated trienoic acid value, a combined value of conjugated dienoic acid and conjugated trienoic acid, a reduction in polyunsaturated fatty acid content, or a polymer byproduct value.

9. The method of claim 7, wherein: The nutritional indicators include: fatty acid content, or structured lipid level.

10. The method of claim 1, wherein: The optimized blend recipe further satisfies the following requirement: a constraint imposed by the at least one predetermined threshold.

11. The method of claim 1 , further comprising: Generate a plurality of data sets of values, wherein the step of generating the plurality of data sets of values ​​comprises: linearly combining first index values ​​obtained from the plurality of pure oils and / or blended oils to form a distribution relationship graph comprising second index values ​​of the plurality of pure oils and / or blended oils, wherein the first index values ​​and the second index values ​​have the more than one index type; applying a sampling algorithm to the distribution relationship graph to sample from the second indicator value; and At least a portion of the data set is formed based on the sampled second indicator value.

12. The method of claim 11, wherein: The sampling algorithm is based on Dirichlet distribution.

13. The method of claim 1, wherein: The step of generating the prediction is real-time.

14. A computer-readable storage medium comprising computer-readable instructions operable, when executed by a computer, to optimize a blending recipe of an edible oil sample, the computer-readable instructions being configured to perform the method of any one of claims 1-13.

15. A device or system comprising: A receiving unit configured to receive a data set formed by a plurality of numerical values, wherein: The numerical value is obtained from a plurality of pure oils and / or blended oils and represents more than one indicator type, wherein the more than one indicator type includes at least an oxidation indicator and a market indicator, the oxidation indicator includes at least a frying life, and the market indicator includes a price; a memory for storing a prediction model, wherein the prediction model is a genetic model constructed by at least one feature vector, the feature vector corresponding to at least an initial index value obtained from a specific pure oil or blended oil among the plurality of pure oils and / or blended oils, the initial index value including at least a frying life and a price of the specific pure oil or blended oil; and A processor configured to access the prediction model stored in the memory to perform the steps of the method according to any one of claims 1 to 13 for generating a prediction of an optimized blending recipe for an edible oil sample.

16. The device or system according to claim 15, further comprising a pre-processing unit, wherein the pre-processing unit is configured to: Performing a linear combination of first index values ​​obtained from a plurality of pure oils and / or blended oils to form a distribution relationship graph containing second index values ​​of the plurality of pure oils and / or blended oils, wherein The first indicator value and the second indicator value have more than one indicator type; Applying a sampling algorithm to the distribution relationship graph to sample from the second indicator value; and At least a portion of a data set of values ​​is formed based on the sampled second indicator values.

Images