Method for extracting multi-component plant beverage through low-temperature pressurization

Abstract

The invention provides a method for extracting a multi-component plant beverage through low-temperature pressurization, and belongs to the technical field of low-temperature pressurization extraction. Rose flowers, Chinese wolfberry fruits, licorice roots, mulberries, Chinese dates, poria cocos, wheat, chrysanthemum flowers and fingered citron are mixed with pure water after being proportioned and crushed; carrying out segmented variable-pressure variable-temperature extraction by adopting a multi-layer paddle combined stirring system and applying ultrasonic assistance; carrying out vacuum degassing, ultrafiltration membrane pretreatment and triple-effect evaporator concentration; seven process parameters including temperature fluctuation amplitude, pressure fluctuation amplitude, ultrasonic cavitation intensity, stirring Reynolds number, dissolved oxygen content after degassing, membrane filtration flux and concentrated solution color difference value are collected and input into an extraction optimization model to calculate a process comprehensive score, and when the score is lower than a standard threshold value, parameters of the next batch are optimized according to an adjustment vector. The technical problem that the quality of products among batches is unstable due to fluctuation of technological parameters in the extraction process of the multi-component plant beverage is solved.

Description

Technical Field

[0001] This invention belongs to the field of low-temperature pressure extraction technology, and more specifically, relates to a method for low-temperature pressure extraction of multi-component plant beverages. Background Technology

[0002] The preparation of multi-component plant-based beverages typically employs hot water extraction or pressure extraction processes to extract the active ingredients from plant raw materials, a method widely used in the food and health beverage production industries. Traditional extraction processes usually utilize fixed temperature and pressure parameters, relying on manual experience to judge the process status and offline monitoring to assess product quality. However, in actual production, temperature fluctuations due to lag in the temperature control system, pressure deviations due to insufficient pressure control precision, and variations in the mixing system's operating status causing differences in mixing uniformity accumulate and are transmitted to subsequent concentration processes. Ultimately, this results in significant differences in the content of active ingredients, color, flavor, and other quality indicators between different batches of product, making it difficult to guarantee product quality stability. In other words, existing technologies suffer from the technical problem of batch-to-batch product quality instability caused by fluctuations in process parameters during the extraction of multi-component plant-based beverages. Summary of the Invention

[0003] In view of this, the present invention provides a method for low-temperature pressure extraction of multi-component plant beverages, which can solve the technical problem in the prior art where fluctuations in process parameters during the extraction of multi-component plant beverages lead to unstable product quality between batches.

[0004] This invention is implemented as follows: A method for low-temperature, pressure-assisted extraction of multi-component plant beverages is provided. Rose, goji berries, licorice, mulberry, jujube, poria cocos, wheat, chrysanthemum, and bergamot are pulverized in a specific ratio and mixed with pure water before being added to an extraction tank. A multi-layer paddle combination stirring system is activated for segmented pressure and temperature-variable extraction with ultrasonic assistance. After extraction, the extract is transferred to a vacuum degassing unit for degassing. The degassed extract is then pretreated with an ultrafiltration membrane to remove enzymes before being concentrated in a triple-effect evaporator. Seven process parameters are collected during the extraction process: temperature fluctuation amplitude, pressure fluctuation amplitude, ultrasonic cavitation intensity, stirring Reynolds number, dissolved oxygen content after degassing, membrane filtration flux, and color difference of the concentrate. These parameters are input into an extraction optimization model to calculate a comprehensive process score. When the comprehensive process score is lower than a standard threshold, the extraction parameters for the next batch are adjusted according to the adjustment vector output by the extraction optimization model, achieving adaptive optimization of process parameters to ensure stable product quality between batches.

[0005] The raw materials are in the following proportions by weight: 2-15 parts rose petals, 10-20 parts mulberry, 10-20 parts poria cocos, 10-30 parts wheat, 5-10 parts licorice root, 10-20 parts jujube, 2-15 parts bergamot, 2-10 parts chrysanthemum, and 10-20 parts goji berries.

[0006] The multi-layer impeller combined mixing system consists of three layers of impellers: a bottom axial flow propeller, a middle turbine impeller, and a top anchor impeller. The speed ratio of the three layers of impellers is set to 1:1.5:0.8, with speeds of 60 rpm, 90 rpm, and 48 rpm, respectively.

[0007] The multi-layer impeller combined stirring system reverses for 5 minutes every 30 minutes, and the three layers of impellers change their rotation direction simultaneously to eliminate dead zones and maintain the Reynolds number between 5000 and 8000.

[0008] The first stage of the segmented pressure and temperature variable extraction process involves extraction for 30 minutes at a pressure of 0.3 MPa and a temperature of 50°C, while simultaneously applying ultrasonic assistance at a frequency of 30 kHz and a power density of 0.8 W / cm³.

[0009] In the second stage of the segmented pressure and temperature variable extraction program, the pressure is increased to 0.6 MPa and the temperature to 58°C for 60 minutes. The transition between the two stages is completed within 5 minutes using a linear temperature and pressure increase method.

[0010] The vacuum degassing unit is subjected to static degassing for 15 minutes at a temperature of 50°C and a vacuum degree of -0.08MPa, while 0.02% by mass of polydimethylsiloxane defoamer is added.

[0011] The ultrafiltration membrane pretreatment uses a 50nm pore size polyethersulfone ultrafiltration membrane to remove polyphenol oxidase and peroxidase, achieving an enzyme removal rate of 88%.

[0012] In the triple-effect evaporator, the first, second, and third effects are concentrated at temperatures of 52°C, 55°C, and 58°C, respectively. Nitrogen gas is introduced for protection to reduce the oxygen volume fraction to 1.5%. The total residence time is controlled within 18 minutes, and the soluble solids mass fraction is concentrated to 45%.

[0013] The extraction optimization model consists of an input layer that receives normalized values ​​of seven process parameters, a first hidden layer containing 128 neurons, a second hidden layer containing 64 neurons, and an attention weight layer that weights the output of the second hidden layer.

[0014] The attention weight coefficient of the attention weight layer is dynamically calculated based on three parameters: dissolved oxygen content after degassing in the current batch, membrane filtration flux, and color difference of the concentrate. This ensures that the higher the three parameters are, the higher the attention weight of the corresponding neuron node.

[0015] The output layer of the extraction optimization model includes one comprehensive scoring node and four adjustment vector nodes. The comprehensive scoring node uses the Sigmoid activation function to output a comprehensive process score between 0 and 1, and the adjustment vector nodes use the hyperbolic tangent activation function to output adjustment coefficients between -1 and 1.

[0016] The steps for establishing the training dataset for the extraction optimization model include conducting 200 batches of extraction experiments, recording process parameters as input features, detecting the final product quality indicators and converting them into a comprehensive process score as output labels, and dividing the data into a training set, a validation set, and a test set in a ratio of 8:1:1.

[0017] The extraction optimization model training step uses the Adam optimization algorithm for gradient descent, with an initial learning rate of 0.001. The learning rate is reduced to 0.9 times the original value every 50 training cycles, and training stops when the validation set loss does not decrease for 20 consecutive cycles.

[0018] The threshold for the comprehensive process score is 0.82, corresponding to the 75th percentile. When the comprehensive process score is lower than 0.82, a parameter adjustment mechanism is triggered.

[0019] The four node values ​​of the adjustment vector correspond to the adjustment coefficients of the first-stage extraction temperature, the second-stage extraction pressure, the ultrasonic power density, and the stirring speed ratio, respectively, with adjustment ranges of ±3℃, ±0.1MPa, ±0.2W / cm³, and ±0.15, respectively.

[0020] This invention establishes an online monitoring system encompassing seven process parameters: temperature fluctuation amplitude, pressure fluctuation amplitude, ultrasonic cavitation intensity, stirring Reynolds number, dissolved oxygen content after degassing, membrane filtration flux, and concentrate color difference. Real-time data from the extraction process is input into an extraction optimization model for comprehensive evaluation. An attention weighting mechanism is used to highlight the impact of three key quality parameters—dissolved oxygen content after degassing, membrane filtration flux, and concentrate color difference—on the process score. When the overall process score falls below a standard threshold, the extraction optimization model outputs adjustment vectors for the first-stage extraction temperature, second-stage extraction pressure, ultrasonic power density, and stirring speed ratio. This enables adaptive optimization of process parameters for the next batch, thereby controlling the impact of process parameter fluctuations on product quality within an acceptable range and ensuring batch-to-batch product quality stability. In summary, this invention solves the technical problem mentioned in the background art where process parameter fluctuations during the extraction of multi-component plant beverages lead to batch-to-batch product quality instability. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention.

[0022] Figure 2This is a schematic diagram of the extraction optimization model.

[0023] Figure 3 This is a graph showing the pressure and temperature changes in a segmented pressure and temperature variable extraction process.

[0024] Figure 4 This is a graph showing the changes in temperature and solids mass fraction of the concentrate in each effect of a triple-effect evaporator.

[0025] Figure 5 Radar plot of attention weight coefficient distribution for extracting and optimizing the model.

[0026] Figure 6 This is a trend chart showing the changes in the overall score of the extraction process for 15 batches of experiments. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below.

[0028] like Figure 1 The diagram shown is a flowchart of a method for low-temperature pressure extraction of multi-component plant beverages provided by the present invention. This method includes the following steps:

[0029] S01. After the raw materials are proportioned according to the mass ratio, they are crushed to a particle size of 80-120μm, mixed with pure water at a material-to-liquid ratio of 1:15, and then put into the extraction tank.

[0030] S02. Start the multi-layer blade combination mixing system. Set the speed ratio of the bottom axial flow propeller, the middle turbine propeller, and the top anchor propeller to 1:1.5:0.8, with speeds of 60 rpm, 90 rpm, and 48 rpm respectively. Reverse the speed for 5 minutes every 30 minutes to maintain the Reynolds number between 5000 and 8000.

[0031] S03. Start the segmented pressure and temperature variable extraction program. In the first stage, extract for 30 minutes at a pressure of 0.3 MPa and a temperature of 50°C, while simultaneously applying ultrasonic assistance at a frequency of 30 kHz and a power density of 0.8 W / cm³. In the second stage, increase the pressure to 0.6 MPa and the temperature to 58°C and extract for 60 minutes.

[0032] S04. After extraction, the extract is transferred to the vacuum degassing unit and allowed to stand for 15 minutes under the conditions of 50℃ and -0.08MPa vacuum. At the same time, 0.02% by mass of polydimethylsiloxane defoamer is added.

[0033] S05. The degassed extract is pretreated by passing it through an ultrafiltration membrane with a pore size of 50 nm to remove polyphenol oxidase and peroxidase. After the enzyme removal rate reaches 88%, it is sent to a triple-effect evaporator for concentration.

[0034] S06. In a triple-effect evaporator, the first, second, and third effects are rapidly concentrated at 52℃, 55℃, and 58℃ respectively. Nitrogen gas is introduced for protection to reduce the oxygen volume fraction to 1.5%. The total residence time is controlled within 18 minutes, and the soluble solids mass fraction is concentrated to 45%.

[0035] S07. Collect seven process parameters during the extraction process, including temperature fluctuation amplitude, pressure fluctuation amplitude, ultrasonic cavitation intensity, stirring Reynolds number, dissolved oxygen content after degassing, membrane filtration flux, and color difference value of concentrate. Input these parameters into the extraction optimization model to calculate the comprehensive process score.

[0036] S08. When the comprehensive process score output by the extraction optimization model is lower than the standard threshold of 0.82, the first-stage extraction temperature, second-stage extraction pressure, ultrasonic power density, and stirring speed ratio of the next batch are adjusted according to the adjustment vector output by the extraction optimization model.

[0037] The raw materials are in the following proportions by weight: 2-15 parts rose petals, 10-20 parts mulberry, 10-20 parts poria cocos, 10-30 parts wheat, 5-10 parts licorice root, 10-20 parts jujube, 2-15 parts bergamot, 2-10 parts chrysanthemum, and 10-20 parts goji berries.

[0038] It should be noted that the adjustment ranges in S08 are ±3℃, ±0.1MPa, and ±0.2W / ±0.15.

[0039] The multi-layer impeller combined mixing system consists of three layers of impellers: a bottom axial flow propeller, a middle turbine impeller, and a top anchor impeller. The bottom axial flow propeller generates an axial circulation flow that causes the material to tumble up and down as a whole. The middle turbine impeller generates a radial shear flow that breaks up high-viscosity agglomerates. The top anchor impeller scrapes off the deposits on the tank wall to prevent localized deposition. The speed ratio of the three layers of impellers is controlled by an independent motor control system to achieve differential operation. Reversing the rotation every 30 minutes for 5 minutes means that the three layers of impellers change their rotation direction simultaneously to eliminate dead zones.

[0040] The segmented pressure and temperature variable extraction program refers to the phased gradient changes in pressure and temperature during the extraction process through a program control system. In the first stage, a lower temperature and pressure are used to avoid thermal degradation of heat-sensitive components such as rose volatile oil. In the second stage, the temperature and pressure are increased to enhance dissolution and diffusion of poorly soluble components such as glycyrrhizin. The transition of pressure and temperature between the two stages is completed within 5 minutes using a linear heating and pressurization method.

[0041] The ultrasonic cavitation intensity refers to the energy density released when cavitation bubbles generated by ultrasound in a liquid collapse. It is characterized by the product of ultrasonic power density and duration of action. An ultrasonic frequency of 30kHz falls within the low-frequency ultrasonic range and can generate relatively large cavitation bubbles, with a power density of 0.8W / m². To ensure that the cavitation effect is fully utilized without causing excessive heat, ultrasound assistance causes mechanical breakage of the cellulose and pectin structure in plant cell walls, reducing the mass transfer resistance of intracellular active ingredients to the solvent.

[0042] The function of the vacuum degassing unit is to remove air dissolved under pressure from the extract. During the pressurized extraction process, the solubility of the gas increases with the increase of pressure. When the gas is concentrated at normal pressure, the dissolved gas is rapidly precipitated to form foam. The vacuum degree of -0.08MPa refers to a vacuum environment with an absolute pressure of 0.02MPa. Under this condition, the escape rate of the dissolved gas increases. After standing for 15 minutes to degas, the gas escape reaches an equilibrium state.

[0043] The ultrafiltration membrane pretreatment uses a polyethersulfone ultrafiltration membrane with a pore size of 50 nm, which can effectively retain enzymes with a molecular weight greater than 50 kDa while allowing small molecule active ingredients to pass through. Polyphenol oxidase and peroxidase are the main enzymes that cause enzymatic browning of plant extracts. An enzyme removal rate of 88% means that the residual enzyme activity is reduced to less than 12% of the original enzyme activity. Membrane filtration flux refers to the volume of liquid passing through a unit membrane area per unit time, and the normal operating range is 80-120 L / ( ·h).

[0044] The triple-effect evaporator refers to a concentration device consisting of three evaporation effects connected in series. The secondary steam generated by the first effect is used as the heating steam for the second effect, and the secondary steam from the second effect is used as the heating steam for the third effect, realizing the cascade utilization of thermal energy. The temperature control of each effect is 52℃, 55℃, and 58℃, all below 60℃ to avoid Maillard reaction and thermal degradation of vitamins. Nitrogen gas is introduced for protection to reduce the oxygen volume fraction from 21% in the air to 1.5%, inhibiting the oxidative browning reaction catalyzed by residual enzymes. The total residence time of 18 minutes refers to the cumulative time from when the extract enters the first effect to when the triple-effect concentration is completed.

[0045] The temperature fluctuation range refers to the maximum deviation of the actual temperature from the set temperature during the extraction process, and the pressure fluctuation range refers to the maximum deviation of the actual pressure from the set pressure during the extraction process. The temperature fluctuation range and pressure fluctuation range reflect the stability of the temperature control system and the pressure control system. Excessive fluctuations will lead to differences in the quality of different batches of products.

[0046] The dissolved oxygen content after degassing refers to the concentration of residual dissolved oxygen in the extract after degassing, which is monitored online using a fluorescence dissolved oxygen sensor, and the unit is mg / L. The lower the dissolved oxygen content after degassing, the weaker the oxidation reaction during subsequent concentration. The color difference value of the concentrate refers to the total color difference between the concentrate and the standard sample in the Lab* color space. The smaller the color difference value of the concentrate, the lighter the degree of browning.

[0047] like Figure 2As shown, the structure of the extraction optimization model is as follows: the input layer receives the normalized values ​​of 7 process parameters; the first hidden layer contains 128 neurons using the modified linear unit activation function; the second hidden layer contains 64 neurons using the modified linear unit activation function; the attention weight layer weights the output of the 64 neurons in the second hidden layer; the attention weight coefficient is dynamically calculated based on three parameters: dissolved oxygen content after degassing, membrane filtration flux, and concentrate color difference value of the current batch; the output layer contains 1 comprehensive scoring node and 4 adjustment vector nodes; the comprehensive scoring node uses the sigmoid activation function to output a comprehensive process score between 0 and 1; and the 4 adjustment vector nodes use the hyperbolic tangent activation function to output adjustment coefficients between -1 and 1.

[0048] The attention weight coefficient of the extraction optimization model is calculated by normalizing the dissolved oxygen content after degassing to a range of 0 to 1, denoted as... Normalizing the membrane filtration flux to the range of 0 to 1 is denoted as The color difference value of the concentrate is normalized to the range of 0 to 1 and denoted as . The formula for calculating the attention weight coefficient of the i-th neuron in the second hidden layer is as follows: The attention weight coefficient is equal to the power of e divided by the sum of the powers of e of all neurons. The exponent part is equal to the output value of the neuron multiplied by the weight factor, and the weight factor is equal to 0.4 multiplied by 1 minus... Add 0.3 multiplied by Add 0.3 multiplied by 1 and subtract The formula states that the lower the dissolved oxygen content after degassing, the higher the membrane filtration flux, and the smaller the color difference value of the concentrate, the higher the attention weight of the corresponding neuron node.

[0049] The steps for establishing the training dataset for the extraction optimization model include conducting 200 batches of extraction experiments on a pilot production line. For each batch, the temperature fluctuation range, pressure fluctuation range, ultrasonic cavitation intensity, stirring Reynolds number, dissolved oxygen content after degassing, membrane filtration flux, and concentrate color difference value are recorded as input features. At the same time, the effective component extraction rate, sensory score, microbial indicators, and shelf life of the final product are detected as quality evaluation indicators. The quality evaluation indicators are converted into a comprehensive process score between 0 and 1 using a weighted average method as the output label. For batches with a comprehensive process score lower than 0.82, the process engineer analyzes the reasons and provides adjustment suggestions for the first-stage extraction temperature, second-stage extraction pressure, ultrasonic power density, and stirring speed ratio as adjustment vector labels. The 200 batches of data are divided into training set, validation set, and test set in a ratio of 8:1:1.

[0050] The training steps of the extraction optimization model include gradient descent using the Adam optimization algorithm, with an initial learning rate of 0.001. Every 50 training cycles, the learning rate is reduced to 0.9 times its original value. The comprehensive score output uses the mean squared error loss function, and the adjustment vector output uses the mean absolute error loss function. The total loss function is the weighted sum of the two loss functions with a weight ratio of 6:4. During training, the performance of the extraction optimization model is evaluated on the validation set every 10 cycles. Training stops when the validation set loss does not decrease for 20 consecutive cycles. The extraction optimization model weight with the smallest validation set loss is selected as the final extraction optimization model. After training, the Pearson correlation coefficient between the process comprehensive score predicted by the extraction optimization model and the actual value is evaluated on the test set. It should be greater than 0.91, and the prediction deviation of the adjustment vector should be less than 15% of the actual adjustment magnitude.

[0051] The method for determining the comprehensive process score threshold of 0.82 is to statistically analyze the distribution of comprehensive process scores in 200 batches of experiments on the pilot production line. Batches with scores below 0.82 are defined as batches that require process adjustment. The standard threshold of 0.82 corresponds to the 75th percentile, which means that when the process quality of the current batch is in the bottom 25%, the parameter adjustment mechanism is triggered.

[0052] The adjustment vectors adjust the process parameters for the next batch by using the four adjustment vector nodes output by the extraction optimization model. These nodes correspond to the adjustment coefficients for the first-stage extraction temperature, the second-stage extraction pressure, the ultrasonic power density, and the stirring speed ratio, respectively. Each adjustment coefficient is between -1 and 1. The actual adjustment is obtained by multiplying the adjustment coefficient by the maximum adjustment range of the corresponding parameter. For example, a maximum adjustment range of ±3℃ for the first-stage extraction temperature means an increase of 1.5℃ when the adjustment coefficient is 0.5; a maximum adjustment range of ±0.1MPa for the second-stage extraction pressure means a decrease of 0.08MPa when the adjustment coefficient is -0.8; and a maximum adjustment range of ±0.2W / ... This means that when the adjustment factor is 1, the power density is increased by 0.2W / The maximum adjustment range of stirring speed ratio is ±0.15, which means that when the adjustment coefficient is -0.6, the speed ratio between the middle turbine propeller and the bottom axial flow propeller is reduced from 1.5 to 1.41.

[0053] The specific implementation methods of the above steps are described in detail below.

[0054] The specific implementation of step S01 is as follows: First, the plant raw materials are weighed. Using an electronic balance, the raw materials are precisely proportioned according to the following mass ratios: 2-15 parts rose, 10-20 parts mulberry, 10-20 parts poria cocos, 10-30 parts wheat, 5-10 parts licorice, 10-20 parts jujube, 2-15 parts bergamot, 2-10 parts chrysanthemum, and 10-20 parts goji berries. The total mass is controlled between 5 kg and 10 kg to meet the needs of single-batch production. Then, the proportioned mixed raw materials are fed into an air jet mill for pulverization. The air jet mill uses high-speed airflow to drive the material particles to collide and shear each other to achieve ultrafine pulverization. The particle size is controlled within the range of 80μm to 120μm. The particle size distribution is detected by a laser particle size analyzer to ensure that more than 90% of the particles meet the particle size requirements. Crushing to the specified particle size range can significantly increase the specific surface area of ​​the plant raw material, thereby increasing the contact area between the effective components and the solvent in the subsequent extraction process. Then, the crushed plant raw material powder is mixed with pure water at a material-to-liquid ratio of 1:15. Water is added in batches and stirred simultaneously to ensure that the powder is fully wetted and avoids agglomeration. After being mixed evenly, it is put into an extraction tank with a volume of 200L to 500L. This step provides pretreated raw materials that meet the particle size and material-to-liquid ratio requirements for the subsequent extraction process.

[0055] The specific implementation of step S02 involves activating the multi-layer impeller combination stirring system within the extraction tank. This system consists of three layers of impellers mounted on the same stirring shaft: a bottom axial-flow propeller positioned 20-30 cm above the bottom of the extraction tank, a middle turbine impeller positioned at the middle of the extraction tank, and a top anchor impeller positioned 10-15 cm below the liquid surface. An independent motor control system drives each of the three impeller layers, setting the speed ratio to 1:1.5:0.8, specifically 60 rpm, 90 rpm, and 48 rpm. The speed ratio setting is based on… Computational fluid dynamics simulation results ensure that a composite flow field of axial main circulation flow and radial shear flow is formed in the extraction tank from bottom to top. The Reynolds number is calculated by multiplying the fluid density by the characteristic velocity, the characteristic length, and then dividing by the dynamic viscosity. Maintaining the Reynolds number between 5000 and 8000 ensures that the fluid is in a fully developed turbulent state. A reversal operation is performed every 30 minutes, in which the three layers of blades simultaneously change their rotation direction and continue for 5 minutes. The reversal operation can eliminate the stirring dead zone and redistribute the material concentration field. The steps achieve uniform mixing of high viscosity phase, suspended phase, and low viscosity phase through the composite stirring flow field.

[0056] The specific implementation of step S03 involves starting the program control system to execute a segmented pressure-temperature variable extraction program. First, the first stage of extraction begins. The pressure inside the extraction tank is adjusted to 0.3 MPa via a pressure control valve, and the temperature inside the extraction tank is controlled at 50°C via a jacketed heating system. These temperature and pressure conditions are suitable for extracting heat-sensitive components such as rose volatile oil. The first stage of extraction lasts for 30 minutes. Simultaneously, an ultrasonic generator is activated to apply an ultrasonic force of 30 kHz and a power density of 0.8 W / m² to the extract. The ultrasonic waves, through cavitation, form tiny bubbles in the liquid, generating localized high temperature and pressure and microjets when the bubbles collapse. These microjets mechanically break down plant cell walls. After the first stage, the program control system automatically starts the second stage of extraction, linearly increasing the pressure to 0.6 MPa and the temperature to 58°C within 5 minutes. The second stage of extraction lasts for 60 minutes. The temperature and pressure conditions are suitable for extracting poorly soluble components such as glycyrrhizin. The segmented extraction strategy achieves efficient extraction of multiple components by matching the optimal dissolution conditions for different components.

[0057] The specific implementation of step S04 involves transferring the extract from the extraction tank to a vacuum degassing unit via a pipeline after extraction. The vacuum degassing unit uses a vacuum pump to reduce the pressure inside the tank to -0.08 MPa (0.02 MPa absolute pressure). Simultaneously, a jacketed heating system maintains the extract temperature at 50°C. Maintaining the temperature at 50°C promotes the escape rate of dissolved gases without causing degradation of heat-sensitive components. The extract is allowed to stand in the vacuum degassing unit for 15 minutes (reference value is 15 to 20 minutes). At the beginning of the standing period, 0.02% (by mass) of polydimethylsiloxane defoamer is added to the extract. This defoamer inhibits foam formation by reducing the surface tension of the liquid. During degassing, a fluorescent dissolved oxygen sensor monitors the dissolved oxygen content in the extract in real time. Degassing is considered complete when the dissolved oxygen content drops below 3 mg / L. This step removes air dissolved under pressure from the extract. To prevent excessive foaming during the subsequent concentration process, the gas is prepared.

[0058] The specific implementation of step S05 involves transporting the degassed extract to an ultrafiltration membrane module for pretreatment via a peristaltic pump. The ultrafiltration membrane module uses a hollow fiber membrane made of polyethersulfone with a pore size of 50 nm. The operating pressure is set to 0.2 MPa to 0.4 MPa. A concentration polarization layer forms on the membrane surface, allowing small molecule active ingredients to pass through the membrane pores into the permeate. Polyphenol oxidase and peroxidase with a molecular weight greater than 50 kDa are retained by the membrane and enter the concentrate. The enzyme activity in the permeate is monitored using an online enzyme activity detector. The ultrafiltration process is stopped when the enzyme removal rate reaches 88%. The reference value for enzyme removal rate is 85% to 92%. The membrane filtration flux is monitored during ultrafiltration, and the normal range for membrane filtration flux is 80 L / (m³). ·h) to 120L / ( ·h), when the membrane filtration flux is less than 80L / ( At h), a membrane cleaning procedure is performed to restore membrane flux. The permeate after ultrafiltration is sent to a triple-effect evaporator for concentration. The above steps remove enzymes that cause enzymatic browning through membrane separation technology.

[0059] The specific implementation of step S06 involves concentrating the ultrafiltration extract by sequentially passing it through the first, second, and third effects of a triple-effect evaporator. The first-effect evaporator uses external steam heating to control the extract temperature at 52°C. The secondary steam generated by the first-effect evaporator, at approximately 48°C, is used as heating steam for the second-effect evaporator. The second-effect evaporator controls the extract temperature at 55°C. The secondary steam generated by the second-effect evaporator is used as heating steam for the third-effect evaporator, which controls the extract temperature at 58°C. This cascaded process utilizes steam heat energy to reduce energy consumption. During the concentration process, nitrogen is introduced into the evaporator through a nitrogen inlet. The oxygen volume fraction in the evaporator was reduced from 21% to 1.5% by volume control. The reference value for oxygen volume fraction is 1% to 2%. The low-oxygen environment inhibits the oxidative browning reaction catalyzed by residual enzymes. By controlling the feed rate and evaporation rate of each effect evaporator, the total residence time of the extract from entering the first effect to completing the third effect concentration was controlled within 18 minutes. The reference value for the total residence time is 15 to 20 minutes. The soluble solids mass fraction of the concentrate was monitored by an online refractometer. Concentration was stopped when the soluble solids mass fraction reached 45%. The above steps reduce Maillard reaction and vitamin thermal degradation through low-temperature rapid concentration.

[0060] The specific implementation of step S07 involves collecting seven process parameters of the extraction process from the distributed control system. The temperature fluctuation amplitude is calculated using the temperature values ​​recorded by the temperature sensor to obtain the maximum deviation of the actual temperature from the set temperature. The pressure fluctuation amplitude is calculated using the pressure values ​​recorded by the pressure sensor to obtain the maximum deviation of the actual pressure from the set pressure. The ultrasonic cavitation intensity is determined by the ultrasonic power density of 0.8 W / The calculation was performed by multiplying the first-stage extraction time by 30 minutes. The stirring Reynolds number was calculated based on fluid density, characteristic velocity, characteristic length, and dynamic viscosity. The dissolved oxygen content after degassing was measured by a fluorescence dissolved oxygen sensor. The membrane filtration flux was calculated using the flow meter and membrane area of ​​the ultrafiltration membrane module. The color difference value of the concentrate was measured by a colorimeter between the concentrate and the standard sample. The total color difference in the color space is obtained. The seven process parameters are normalized to a range of 0 to 1 using the minimum-maximum normalization method. The normalized data is input into the extraction optimization model. The extraction optimization model receives the data through the input layer and performs forward propagation calculations through 128 neurons in the first hidden layer and 64 neurons in the second hidden layer. The attention weight layer dynamically calculates the attention weight coefficient based on the dissolved oxygen content after degassing, membrane filtration flux, and color difference value of the concentrate, and weights the output of the second hidden layer. The comprehensive scoring node of the output layer outputs a comprehensive process score between 0 and 1. The above steps evaluate the process quality of the current batch through the extraction optimization model.

[0061] The specific implementation of step S08 involves comparing the comprehensive process score output by the extraction optimization model with a standard threshold of 0.82. When the comprehensive process score is lower than the standard threshold of 0.82, it is determined that the process quality of the current batch needs improvement. The output layer of the extraction optimization model contains four adjustment vector nodes, which respectively output the adjustment coefficients for the first-stage extraction temperature, the second-stage extraction pressure, the ultrasonic power density, and the stirring speed ratio. These adjustment coefficients are values ​​between -1 and 1. The actual adjustment amount is obtained by multiplying each adjustment coefficient by the maximum adjustment range of the corresponding parameter. The maximum adjustment range for the first-stage extraction temperature is ±3℃, the maximum adjustment range for the second-stage extraction pressure is ±0.1MPa, and the maximum adjustment range for the ultrasonic power density is ±0.2W / The maximum adjustment range of the stirring speed ratio is ±0.15. The actual adjustment is applied to the process parameter setting of the next batch. The above steps achieve adaptive optimization of process parameters through closed-loop feedback control.

[0062] It should be noted that one of the key technical ideas of this invention is a segmented pressure and temperature variable extraction process combined with ultrasonic-assisted technology. Traditional single process conditions are difficult to simultaneously and efficiently extract active ingredients with significantly different dissolution characteristics from multiple plant raw materials. This invention uses a segmented strategy of first-stage low-temperature and low-pressure extraction of heat-sensitive components and second-stage high-temperature and high-pressure extraction of insoluble components to match the optimal dissolution conditions for different components. At the same time, the ultrasonic cavitation effect is introduced in the first stage to mechanically break down the plant cell walls, which compensates for the mass transfer resistance caused by insufficient cell wall breakdown under low-temperature conditions. Compared with the traditional constant temperature and constant pressure extraction method, this technical idea significantly improves the extraction rate of various components and shortens the extraction time. The second key technical approach is the combined application of a vacuum degassing unit and ultrafiltration membrane pretreatment. During pressure extraction, gas solubility increases with pressure, leading to severe foaming during subsequent atmospheric pressure concentration. This invention addresses this by using a vacuum degassing unit to promote the escape of dissolved gases under low pressure, combined with an antifoaming agent to suppress foam formation. Simultaneously, ultrafiltration membranes remove polyphenol oxidase and peroxidase, inhibiting enzymatic browning at its source. This approach, compared to traditional direct concentration methods, avoids concentrated liquid splashing and reduced equipment utilization, while improving product color stability. The third key technical approach is an adaptive process parameter adjustment mechanism based on an extraction optimization model. Traditional experience-based adjustment methods struggle to accurately identify key factors affecting process quality and provide quantitative adjustment schemes. This invention uses an extraction optimization model to learn the nonlinear mapping relationship between process parameters and product quality. An attention weighting mechanism dynamically monitors key quality indicators such as dissolved oxygen content after degassing, membrane filtration flux, and concentrated liquid color difference. When the overall process score falls below a standard threshold, a targeted parameter adjustment vector is automatically output. This approach, compared to traditional manual adjustment methods, achieves closed-loop control of process quality and reduces batch-to-batch variations. The synergistic effect of the above three key technical approaches lies in the fact that segmented extraction and ultrasound assistance improve the extraction rate of effective components, vacuum degassing and membrane pretreatment ensure the stability of the concentration process and the color quality of the product, and the extraction optimization model realizes intelligent optimization of process parameters and closed-loop quality control. Compared with traditional methods, the synergistic effect constructs a full-process quality control system from raw material pretreatment, segmented extraction, degassing and enzyme removal to concentrated finished product, which significantly improves the production efficiency and product stability of multi-component plant beverages.

[0063] It should be noted that this invention also solves the following technical problem: the uneven extraction efficiency caused by differences in extraction conditions for different active ingredients in the extraction process of multi-component plant beverages. This invention designs a segmented pressure-temperature variable extraction program. In the first stage, a pressure of 0.3 MPa and a temperature of 50°C are used for gentle extraction of heat-sensitive components such as rose volatile oil, avoiding thermal degradation caused by high temperatures. Simultaneously, ultrasonic waves at a frequency of 30 kHz are applied to assist in breaking down cell wall structures and reducing mass transfer resistance. In the second stage, the pressure is increased to 0.6 MPa and the temperature to 58°C to enhance the dissolution and diffusion dynamics of poorly soluble components such as glycyrrhizin. Combined with a multi-layer impeller combined stirring system, the bottom axial-flow propeller generates axial circulation flow, the middle turbine impeller generates radial shear flow, and the top anchor impeller scrapes away deposits on the tank wall. This achieves the synergistic effect of different flow modes, allowing heat-sensitive and poorly soluble components to fully dissolve under their respective suitable process conditions, improving the overall extraction efficiency of the multi-component system and solving the technical problem of uneven extraction efficiency caused by differences in extraction conditions for different active ingredients.

[0064] Specifically, the principle of this invention is as follows: This invention constructs a multi-dimensional state feature vector by collecting process parameters from the entire extraction process. It then utilizes a neural network model to learn the nonlinear mapping relationship between process parameters and product quality, enabling accurate prediction of the overall process score for the current batch. An attention weighting mechanism dynamically adjusts the weight coefficients of neuron nodes based on three parameters: dissolved oxygen content after degassing, membrane filtration flux, and concentrate color difference. This allows the model to focus on the process steps that have the most significant impact on product quality, improving the accuracy and sensitivity of the score. When the overall process score is below the standard threshold, it indicates a process deviation in the current batch. The extraction optimization model analyzes the source of the deviation and outputs an adjustment vector to guide the parameter settings for the next batch, forming a closed-loop feedback control mechanism. This data-driven adaptive adjustment strategy can promptly correct process deviations, compensate for equipment response lag and the impact of environmental disturbances, and maintain process parameters within the optimal range, ensuring the stability of product quality between batches and meeting the logical requirements of quality control in industrial production.

[0065] The following provides a specific embodiment 1 of the present invention. The specific implementation of steps S01 and S02 in this embodiment 1 is the same as that described above, and will not be repeated in detail here. The specific implementation of other steps is described in detail below.

[0066] In a specific implementation of step S03, the ultrasonic cavitation intensity The calculation formula is expressed as follows:

[0067] ;

[0068] In the formula, Ultrasonic cavitation intensity, unit: ; The ultrasonic power density is taken as 0.8 in this embodiment. ; The duration of ultrasonic treatment is 1800 in this embodiment. .

[0069] The specific implementation methods of steps S04, S05, and S06 are the same as those described above, and will not be repeated in detail here.

[0070] In the specific implementation of step S07, the seven collected process parameters are normalized. The normalization formula is expressed as follows:

[0071] ;

[0072] In the formula, For the first The normalized values ​​of the process parameters range from 0 to 1. For the first Measured values ​​of the process parameters; For the first The minimum value of a process parameter in the training dataset; For the first The maximum value of each process parameter in the training dataset; The values ​​range from 1 to 7, corresponding to temperature fluctuation amplitude, pressure fluctuation amplitude, ultrasonic cavitation intensity, stirring Reynolds number, dissolved oxygen content after degassing, membrane filtration flux, and concentrate color difference value, respectively. and The Reynolds number was obtained by analyzing 200 batches of training data. The calculation formula is expressed as follows:

[0073] ;

[0074] In the formula, The stirring Reynolds number is dimensionless. This is the density of the extract, in units of... The value is usually 1000. ; The speed of the central turbine propeller is given in units of... In this embodiment, it is 90. ; The diameter of the central turbine propeller, in units of The empirical value is 0.3. ; The extract's dynamic viscosity is expressed in units of... Its value is typically 0.0008 within the temperature range of 50℃ to 58℃. The 60 in the denominator is a conversion factor for rotational speed units, used to convert... Convert to .

[0075] In the specific implementation of step S08, the formula for calculating the attention weight coefficient of the extraction optimization model is as follows:

[0076] ;

[0077] In the formula, For the first The attention weight coefficients of each neuron node, with values ​​ranging from 0 to 1; For the second hidden layer The output value of each neuron node ranges from 0 to positive infinity, determined by the modified linear unit activation function. The weighting factor is dimensionless. The summation index ranges from 1 to 64; For the second hidden layer The output value of each neuron node. Weighting factor. The calculation formula is expressed as follows:

[0078] ;

[0079] In the formula, The normalized value of dissolved oxygen content after degassing is calculated by subtracting the minimum value of the training dataset from the measured value of dissolved oxygen content after degassing and then dividing by the difference between the maximum and minimum values ​​of the training dataset. The value ranges from 0 to 1. This is the normalized value of membrane filtration flux, calculated in the same way as above, and its value ranges from 0 to 1. The normalized value for the color difference of the concentrate is calculated in the same way as above, and its value ranges from 0 to 1. The coefficients 0.4, 0.3, and 0.3 are weighting coefficients, and their empirical values ​​are determined through validation set optimization. The formula for calculating the adjustment amount of the adjustment vector to the process parameters is expressed as follows:

[0080] ;

[0081] In the formula, For the first The actual adjustment amount of each process parameter; The first output of the extraction optimization model Each adjustment vector node has a value ranging from -1 to +1; For the first The maximum adjustment range of each process parameter; The values ​​range from 1 to 4, corresponding to the first-stage extraction temperature, the second-stage extraction pressure, the ultrasonic power density, and the stirring speed ratio, respectively. It is 3℃. It is 0.1 MPa. It is 0.2 , It is 0.15. When When the value is equal to 4, the formula for calculating the adjusted stirring speed ratio is as follows:

[0082] ;

[0083] In the formula, The speed ratio between the middle turboprop and the bottom axial-flow propeller is adjusted, and is dimensionless. This is the base speed ratio, with a default value of 1.5; This represents the speed ratio adjustment, which is dimensionless. (Process overall score) The formula for comparing with the standard threshold is expressed as follows:

[0084] ;

[0085] In the formula, To extract and optimize the comprehensive scoring node of the model output layer The overall process score calculated by the activation function ranges from 0 to 1; The standard threshold is set to 0.82 in this embodiment. When the above inequality is satisfied, the process parameter adjustment mechanism for the next batch is triggered.

[0086] The triple-effect evaporator refers to a concentration device consisting of three evaporation effects connected in series. The secondary steam generated in the first effect is used as heating steam for the second effect, and the secondary steam from the second effect is used as heating steam for the third effect, achieving cascaded utilization of thermal energy. The temperatures of each effect are controlled at 52℃, 55℃, and 58℃, all below 60℃ to avoid Maillard reactions and vitamin thermal degradation. Nitrogen gas is used for protection, reducing the oxygen volume fraction from 21% in the air to 1.5%, inhibiting oxidative browning reactions catalyzed by residual enzymes. The total residence time of 18 minutes refers to the cumulative time from when the extract enters the first effect to when the triple-effect concentration is completed. Soluble solids mass fraction... The calculation formula is expressed as follows:

[0087] ;

[0088] In the formula, This represents the mass fraction of soluble solids, expressed as a percentage. The value represents the mass of soluble solids in the concentrate, expressed in g, and determined by refractive index method. The total mass of the concentrate is expressed in grams; in this embodiment, the concentrate is concentrated to... It reached 45%.

[0089] The color difference value of the concentrate refers to the difference between the concentrate and the standard sample. Total color difference in the color space; the smaller the color difference value of the concentrate, the less browning it indicates. The calculation formula is expressed as follows:

[0090] ;

[0091] In the formula, The total color difference value is dimensionless. The value represents the difference between the concentrate and the standard sample on the lightness axis, calculated as follows: (The value of the concentrate is missing from the original text.) Value minus the standard sample value; The value represents the difference between the concentrate and the standard sample on the red-green axis, calculated as follows: [The value of the concentrate is missing from the original text]. Value minus the standard sample value; The value represents the difference between the concentrate and the standard sample on the yellow-blue axis, calculated as follows: [The value of the concentrate is missing from the original text]. Value minus the standard sample Value; the above value, value, All values ​​were obtained by measuring a colorimeter, among which The value ranges from 0 to 100. The value ranges from -128 to +127. The value ranges from -128 to +127.

[0092] It should be noted that the variables involved in this invention are explained in detail in Table 1.

[0093] Table 1. Variable Explanation Table

[0094]

[0095] To better understand and implement this invention, the following is an example 2 of a specific application scenario: A plant beverage technology team faced technical challenges with traditional hot water extraction processes, including severe loss of heat-sensitive components, insufficient extraction of insoluble components, and large batch-to-batch quality fluctuations. They decided to modify the process using the low-temperature pressure extraction method of this invention. The team first prepared the raw materials according to the following ratio: 2-15 parts rose petals, 10-20 parts mulberry, 10-20 parts poria cocos, 10-30 parts wheat, 5-10 parts licorice root, 10-20 parts jujube, 2-15 parts bergamot, 2-10 parts chrysanthemum, and 10-20 parts goji berries. After processing with an air jet mill to achieve a particle size of 95 μm, the raw materials were mixed with pure water and then added to the extraction tank. After the multi-layer impeller combined mixing system is started, the bottom axial flow propeller generates axial circulation at 60 rpm, the middle turbine propeller generates radial shear flow at 90 rpm, and the top anchor impeller scrapes off the tank wall deposits at 48 rpm. The speed ratio of the three layers of impellers is maintained at 1:1.5:0.8. The system reverses for 5 minutes every 30 minutes to eliminate dead zones, and the measured Reynolds number is stable at 6200.

[0096] After the segmented pressure-switching and temperature-switching extraction program is started, the first stage involves extraction for 30 minutes at a pressure of 0.3 MPa and a temperature of 50°C, while simultaneously applying a frequency of 30 kHz and a power density of 0.8. With ultrasonic assistance, the ultrasonic probe is positioned in the lower part of the extraction tank, and the ultrasonic cavitation intensity reaches 24. Thermosensitive components such as rose volatile oil and wolfberry polysaccharides are protected from thermal degradation under low-temperature conditions. For example... Figure 3 As shown, in the second stage, after a 5-minute linear temperature and pressure increase transition, the pressure was raised to 0.6 MPa and the temperature to 58°C and maintained for 60 minutes. This significantly improved the dissolution and diffusion rates of poorly soluble components such as glycyrrhizin and morin anthocyanins. At the end of the second stage, the soluble solids content of the extract reached 6.8%. After extraction, 18,500 L of extract was transferred to a vacuum degassing unit. The temperature was controlled at 50°C, and the vacuum pump reduced the absolute pressure inside the tank to 0.02 MPa. After standing for 15 minutes for degassing, 370 g of polydimethylsiloxane defoamer was added. The fluorescent dissolved oxygen sensor showed that the dissolved oxygen content decreased from 8.6 mg / L to 1.2 mg / L after degassing.

[0097] The degassed extract is passed through a 50mm pore size. The pretreatment of the polyethersulfone ultrafiltration membrane has an effective filtration area of ​​180. The operating pressure was maintained at 0.25 MPa, and the membrane filtration flux remained stable at 92. The removal rates of polyphenol oxidase and peroxidase reached 88%, and the residual enzyme activity in the permeate decreased to less than 12% of the original enzyme activity. Figure 4 As shown, the pretreated extract enters a triple-effect evaporator for concentration. The evaporation areas of the first, second, and third effects are 45 evaporators, respectively. 38 32 The temperatures were controlled at 52℃, 55℃, and 58℃. The secondary steam generated by the first-effect evaporation was used as the heating steam for the second-effect evaporation, and the secondary steam from the second-effect evaporation was used as the heating steam for the third-effect evaporation, thus realizing the cascade utilization of thermal energy. The nitrogen protection system reduced the oxygen volume fraction in the triple-effect evaporator from 21% to 1.5%. The total residence time of the extract from entering the first effect to completing the triple-effect concentration was 18 minutes. The final concentrate had a soluble solids mass fraction of 45% and was dark reddish-brown. The total color difference between the concentrate and the standard sample was measured to be 3.8 using a colorimeter.

[0098] like Figure 5 As shown, the technical team collected seven process parameters during the extraction process and input them into the extraction optimization model for evaluation. The temperature fluctuation range was 1.2℃, the pressure fluctuation range was 0.04MPa, and the ultrasonic cavitation intensity was 24. The stirring Reynolds number was 6200, the dissolved oxygen content after degassing was 1.2 mg / L, and the membrane filtration flux was 92. The color difference of the concentrate was 3.8. These seven parameters, after normalization, were input into the input layer of the extraction optimization model. The first hidden layer of the extraction optimization model contains 128 neurons and uses a modified linear unit activation function for feature extraction. The second hidden layer contains 64 neurons for feature combination. The attention weight layer calculates the attention weight coefficient of each neuron node based on the normalized value of dissolved oxygen content after degassing (0.15), membrane filtration flux (0.72), and the normalized value of color difference of the concentrate (0.38). The weight factor is 0.68, and the distribution of attention weight coefficients shows that the neuron node corresponding to the dissolved oxygen content after degassing receives a higher weight. The output layer of the extraction optimization model gives a comprehensive process score of 0.79, which is lower than the standard threshold of 0.82. It also outputs four adjustment vector node values ​​of 0.6, -0.4, 0.3, and -0.2, corresponding to the first-stage extraction temperature adjustment coefficient, the second-stage extraction pressure adjustment coefficient, the ultrasonic power density adjustment coefficient, and the stirring speed ratio adjustment coefficient, respectively.

[0099] Based on the adjustment vector of the extraction optimization model, the technical team adjusted the process parameters for the next batch. The extraction temperature in the first stage was increased from 50℃ to 51.8℃, and the extraction pressure in the second stage was decreased from 0.6MPa to 0.56MPa, while the ultrasonic power density was reduced from 0.8... Upgraded to 0.86 The stirring speed ratio between the central turbine propeller and the bottom axial-flow propeller was reduced from 1.5 to 1.47. The adjusted second-batch extraction experiment showed that the overall process score improved to 0.86, higher than the standard value of 0.82; the color difference of the concentrate decreased to 3.2; the dissolved oxygen content after degassing decreased to 0.9 mg / L; and the membrane filtration flux increased to 98. The extraction rate of the effective components and the sensory score of the final product were both improved. The technical team conducted 15 batches of extraction experiments. The extraction optimization model adjusted the operating conditions of the next batch in real time according to the process parameters of each batch. The quality fluctuation between batches was significantly reduced, and the overall process score was stable above 0.85. The process parameters and quality indicators of the 15 batches of extraction experiments are shown in Table 2.

[0100] Table 2. Process parameters and quality indicators for 15 batches of extraction experiments.

[0101]

[0102] Example 3 uses the above process, but the formula has been adjusted as follows: 2-15 parts rose, 10-20 parts mulberry, 10-20 parts poria cocos, 10-30 parts wheat, 5-10 parts licorice, 10-20 parts jujube, 2-15 parts bergamot, 2-10 parts chrysanthemum, 10-20 parts wolfberry, 10-20 parts jujube seed, 10-20 parts lily, and 5-15 parts ganoderma lucidum.

[0103] Example 4 uses the above process with adjustments to the formula, specifically: 2-15 parts rose petals, 10-20 parts mulberry, 10-20 parts poria cocos, 10-30 parts wheat, 5-10 parts licorice root, 10-20 parts jujube, 2-15 parts bergamot, 2-10 parts chrysanthemum, 10-20 parts wolfberry, 10-20 parts sour jujube seed, 10-20 parts lily bulb, 5-15 parts ganoderma lucidum, 10-20 parts codonopsis pilosula, 10-20 parts polygonatum sibiricum, 10-20 parts longan pulp, 10-20 parts gastrodia elata, and 5-15 parts dried tangerine peel.

[0104] In Examples 3 and 4, multiple extraction experiments were also conducted, and the results showed a significant reduction in batch-to-batch quality fluctuations and a stable overall process score of over 0.85.

[0105] This invention represents a significant advancement in technical principles compared to traditional hot water extraction processes. The segmented pressure-temperature-switching extraction procedure designs differentiated extraction conditions based on the physicochemical properties of different components. The first stage uses low temperature and low pressure to prevent thermal degradation of heat-sensitive components, while the second stage uses high temperature and high pressure to promote the dissolution and diffusion of insoluble components, resolving the contradiction in traditional processes where single temperature and pressure conditions cannot simultaneously address the extraction of different components. Ultrasonic-assisted extraction utilizes cavitation to mechanically break down plant cell walls, reducing the mass transfer resistance of intracellular active ingredients to the solvent, thus shortening extraction time and increasing extraction rate. The multi-layered impeller combination stirring system, through the synergistic action of the bottom axial-flow propeller, the middle turbine impeller, and the top anchor impeller, generates axial circulation flow, radial shear flow, and wall scraping effects, eliminating dead zones and preventing localized deposition, ensuring uniform mixing of materials during extraction. The vacuum degassing unit removes pressurized dissolved gases before atmospheric pressure concentration, preventing rapid precipitation and foam formation during concentration, while simultaneously reducing dissolved oxygen content to inhibit oxidative browning reactions. Ultrafiltration membrane pretreatment removes polyphenol oxidase and peroxidase, inhibiting enzymatic browning reactions at the source, rather than relying on high-temperature enzyme inactivation leading to the loss of heat-sensitive components. A triple-effect evaporator enables cascaded utilization of heat energy and rapid concentration under low-temperature nitrogen protection, avoiding Maillard reactions and vitamin thermal degradation. The extraction optimization model dynamically monitors key quality indicators such as dissolved oxygen content after degassing, membrane filtration flux, and concentrate color difference through an attention mechanism. Based on real-time process parameters, it outputs adjustment vectors to achieve adaptive optimization of process parameters between batches, solving the problem of large batch-to-batch quality fluctuations caused by traditional processes relying on manual experience adjustments.

[0106] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for low-temperature pressure extraction of multi-component plant beverages, characterized in that, Rose petals, goji berries, licorice root, mulberry, jujube, poria cocos, wheat, chrysanthemum, and bergamot are pulverized in a specific ratio and mixed with pure water before being added to an extraction tank. A multi-layer paddle combined stirring system is activated for segmented pressure and temperature variable extraction with ultrasonic assistance. After extraction, the extract is transferred to a vacuum degassing unit for degassing. The degassed extract is then pretreated with an ultrafiltration membrane to remove enzymes before being sent to a triple-effect evaporator for concentration. Seven process parameters are collected during the extraction process: temperature fluctuation amplitude, pressure fluctuation amplitude, ultrasonic cavitation intensity, stirring Reynolds number, dissolved oxygen content after degassing, membrane filtration flux, and color difference of the concentrate. These parameters are input into the extraction optimization model to calculate the comprehensive process score. When the comprehensive process score is lower than the standard threshold, the extraction parameters for the next batch are adjusted according to the adjustment vector output by the extraction optimization model. This achieves adaptive optimization of process parameters to ensure stable product quality between batches.

2. The method for low-temperature pressure extraction of multi-component plant beverages according to claim 1, characterized in that, The mass ratio of the raw materials is as follows: 2-15 parts rose petals, 10-20 parts mulberry, 10-20 parts poria cocos, 10-30 parts wheat, 5-10 parts licorice, 10-20 parts jujube, 2-15 parts bergamot, 2-10 parts chrysanthemum, and 10-20 parts wolfberry.

3. The method for low-temperature pressure extraction of multi-component plant beverages according to claim 2, characterized in that, The multi-layer impeller combined mixing system consists of three layers of impellers: a bottom axial flow propeller, a middle turbine propeller, and a top anchor propeller. The speed ratio of the three layers of impellers is set to 1:1.5:0.8, with speeds of 60 rpm, 90 rpm, and 48 rpm, respectively.

4. The method for low-temperature pressure extraction of multi-component plant beverages according to claim 3, characterized in that, The multi-layer impeller combined stirring system reverses for 5 minutes every 30 minutes, and the three layers of impellers change their rotation direction simultaneously to eliminate dead zones and maintain the Reynolds number between 5000 and 8000.

5. The method for low-temperature pressure extraction of multi-component plant beverages according to claim 4, characterized in that, The first stage of the segmented pressure and temperature variable extraction procedure involves extraction for 30 minutes at a pressure of 0.3 MPa and a temperature of 50°C, while simultaneously applying ultrasonic assistance at a frequency of 30 kHz and a power density of 0.8 W / cm³.

6. The method for low-temperature pressure extraction of multi-component plant beverages according to claim 5, characterized in that, The second stage of the segmented pressure and temperature variable extraction program is to increase the pressure to 0.6 MPa and the temperature to 58°C for 60 minutes. The transition between the two stages is completed within 5 minutes using a linear temperature and pressure increase method.

7. The method for low-temperature pressure extraction of multi-component plant beverages according to claim 6, characterized in that, The vacuum degassing unit is allowed to stand for 15 minutes at a temperature of 50°C and a vacuum degree of -0.08MPa, while 0.02% by mass of polydimethylsiloxane defoamer is added.

8. The method for low-temperature pressure extraction of multi-component plant beverages according to claim 7, characterized in that, The ultrafiltration membrane pretreatment uses a 50nm pore size polyethersulfone ultrafiltration membrane to remove polyphenol oxidase and peroxidase, with an enzyme removal rate of 88%.

9. The method for low-temperature pressure extraction of multi-component plant beverages according to claim 8, characterized in that, In the triple-effect evaporator, the first, second, and third effects are concentrated at temperatures of 52°C, 55°C, and 58°C, respectively. Nitrogen gas is introduced for protection to reduce the oxygen volume fraction to 1.5%. The total residence time is controlled within 18 minutes, and the soluble solids mass fraction is concentrated to 45%.

10. The method for low-temperature pressure extraction of multi-component plant beverages according to claim 9, characterized in that, The structure of the extraction optimization model includes an input layer that receives normalized values ​​of seven process parameters, a first hidden layer containing 128 neuron nodes, a second hidden layer containing 64 neuron nodes, and an attention weight layer that weights the output of the second hidden layer.

Images