A method for preparing a cream based on biphasic extraction

Abstract

The present application relates to a kind of based on two-phase extraction cream preparation method, the present application is based on typical raw material to establish the calibration mapping of dielectric parameter and extraction efficiency, the real part, imaginary part and frequency spectrum curvature of complex dielectric constant three indicators are compressed into three-dimensional dielectric fingerprint vector by principal component regression, quantitatively represent the water content, oil content and cell structure characteristics of raw material, extraction feed line integrated microwave resonant cavity sensor, realize the real-time collection of raw material slurry dielectric parameters, combined with fast vector fitting and filtering algorithm to obtain stable real-time dielectric fingerprint, find matching in calibration atlas by K near neighbor algorithm, the table lookup result is weighted interpolation as temperature gradient slope, flow rate ratio and pH three dynamic adjustment instructions, drive actuator to carry out closed-loop optimization adjustment, and there is drift calibration mechanism to maintain the long-term stability of parameter mapping.The present application improves the adaptability of extraction process to raw material difference, realizes parameter real-time optimization and stable control of product extraction rate.

Description

Technical Field

[0001] This invention relates to the field of cosmetic industrial process control and extraction engineering technology, and in particular to a method for preparing face cream based on two-phase extraction. Background Technology

[0002] In existing face cream manufacturing processes, two-phase extraction technology is the mainstream method for extracting and separating active ingredients from natural plants, achieving synergistic control of efficient extraction of effective components and protection of sensitive components in the cosmetics industry. Mainstream technical routes typically rely on setting fixed extraction process parameters (such as temperature, flow rate ratio, pH, etc.) and combining solvent system optimization, homogenization and dispersion method improvement, or back-end component detection feedback to regulate extraction efficiency and component stability. With the diversification of natural plant raw material types and suppliers, providing a dynamically consistent extraction process for industrial production has become an important industry demand. Currently, most production lines adopt empirical presets or static process modeling for parameter setting, while some processes introduce offline HPLC and other analyzers of the extracted products to correct parameters, but all suffer from critical limitations in adaptability.

[0003] In related applications, temperature control typically relies on zoned electric heating elements, flow rate adjustment is mainly achieved through high-precision pump control, and pH fine-tuning depends on liquid addition or electrode regulation. These methods primarily rely on "passive monitoring" or "stage-by-stage correction" of process variables, failing to achieve real-time front-end identification and proactive parameter adaptation for sudden changes in raw material properties (such as rapid changes in moisture content, differences in cell wall integrity, or lipid migration). Furthermore, while in-situ ultrasonic monitoring, online image analysis, and near-infrared spectroscopy methods have found applications in tracking raw material properties, they are limited by signal penetration capabilities, system response speed, and the complexity of multi-variable coordinated control, and cannot yet meet the dynamic response requirements of "sensorless feedback delay and real-time fully automatic parameter mapping" in continuous industrial production.

[0004] Currently, the industry-standard adaptive parameter control logic mainly relies on backend product testing feedback for trend-tracking or intermittent manual adjustments. For example, sample testing results are used as a basis for process traceability, and parameters such as extraction temperature, flow rate, and pH are manually fine-tuned. This approach has limitations such as long response cycles, significant adjustment lag, and large process fluctuations. When there are significant changes in the physical properties of raw material batches or unexpected adulteration occurs, fixed parameters cannot adaptively compensate for individual differences between raw materials, easily leading to quality risks such as decreased extraction rates of effective components, degradation of sensitive components, or batch-specific inconsistencies. Known technologies have not yet achieved an "integrated intelligent adaptive parameter control" paradigm based on the intrinsic physical properties of raw materials. The key bottleneck lies in developing a technical system capable of capturing macroscopic property variations of raw materials in real time and non-invasively and automatically reflecting them in process parameter settings.

[0005] The requirements in the cosmetics manufacturing industry are becoming increasingly stringent. The market's demand for the stability, batch-to-batch consistency, and high-throughput industrialization of natural active ingredients is constantly increasing, placing higher standards on the intelligence, real-time performance, and robustness of production processes. How to achieve an adaptive parameter adjustment system based on the intrinsic physical properties of raw materials without relying on slow backend detection or costly sensor monitoring has become a pressing challenge in current extraction process engineering technology. Summary of the Invention

[0006] This invention provides a method for preparing face cream based on biphasic extraction, aiming to solve the problems existing in the prior art mentioned in the background section.

[0007] The technical solution of this invention is: a method for preparing a face cream based on biphasic extraction, comprising the following steps: S1: Construct a dielectric fingerprint vector for the raw materials based on experimental data of dielectric properties of typical plant raw material slurry samples; S2: Based on the correlation between the dielectric fingerprint vector of raw materials and the two-phase extraction efficiency, a pre-calibration spectrum is established; S3: An embedded microwave resonant cavity sensor is integrated into the extraction feed pipeline to collect the S11 reflection coefficient frequency response curve of the raw material slurry; S4: Based on the frequency response curve of the reflection coefficient of S11, calculate the real-time dielectric parameters of the raw material slurry; extract the resonant frequency shift and half-power bandwidth through fast vector fitting, and output stable real and imaginary values ​​of the complex dielectric constant after sliding window averaging and outlier removal. S5: Substitute the real-time dielectric parameters into the material dielectric fingerprint vector to generate the real-time dielectric fingerprint vector of the current material. S6: Input the real-time dielectric fingerprint vector into the pre-calibrated spectrum, perform parameter mapping lookup; use the K-nearest neighbor algorithm to match the nearest neighbor sample cluster, perform weighted inverse distance interpolation on the parameter combination within the cluster, and output a dynamic adjustment instruction set; S7: Drives the actuator based on a dynamic adjustment instruction set; S8: Implement drift compensation calibration. When switching raw material batches, inject standard glycerol aqueous solution for zero-point calibration. After continuous operation for a preset time, dynamically fine-tune the mapping weight coefficient of the pre-calibrated spectrum according to the deviation of the product component extraction rate to maintain the long-term stability of parameter mapping.

[0008] Preferably, in step S1, the experimental data includes three indicators: the real part amplitude, the imaginary part amplitude, and the spectral curvature of the complex permittivity measured at a frequency of 1.2 GHz. These three indicators are integrated into a three-dimensional vector through principal component regression analysis, serving as intrinsic physical characteristics that characterize the moisture content, oil content, and cell wall integrity of the raw materials.

[0009] Preferably, in step S2, the correlation is obtained through orthogonal experiments, and a mapping relationship is established between the three-dimensional dielectric fingerprint vector and the optimal extraction parameter combination. The optimal extraction parameter combination includes the temperature gradient slope, the ratio of organic phase to aqueous phase flow rate, and the interface pH offset.

[0010] Preferably, in step S3, a microwave excitation signal is transmitted through a microstrip feeding structure; the reflected signal is received and the frequency domain response is recorded; the original data is subjected to sliding window averaging filtering to remove high-frequency noise; and the S11 reflection coefficient frequency response curve data is output.

[0011] Preferably, in step S7, the dynamic adjustment instruction set is converted into actuator drive signals through digital-analog conversion. The temperature gradient slope signal is mapped to 0–5, the flow rate ratio signal is mapped to 0–3, the pH offset is mapped to 0–1, the signal quantization step size is 0.01, and CRC checksum is used to ensure data packet integrity.

[0012] The beneficial technical effects of this invention are as follows: 1) By constructing a mapping relationship between the dielectric fingerprint of raw materials and the dual-phase extraction efficiency, this invention realizes the autonomous setting of parameters based on the intrinsic physical properties of the front-end materials, which significantly improves the response sensitivity and control accuracy of the extraction process. 2) This invention selects the amplitude, phase difference and spectral curvature of the complex dielectric constant in the 1.2 GHz band as the three-dimensional dielectric fingerprint vector, and combines it with principal component regression analysis to establish an offline calibration spectrum. This allows different plant raw materials to predict the optimal process window through their electromagnetic properties before entering the extraction system, without waiting for the terminal component analysis results. This significantly shortens the system response time to hundreds of milliseconds and effectively ensures the single-pass extraction rate and structural integrity of highly active ingredients. 3) This invention adopts a technical approach that combines miniaturized embedded microwave resonant cavity sensors with real-time calculation algorithms, which significantly enhances the adaptive capability and operational robustness of the process system; 4) The U-shaped flow channel resonant cavity of the present invention has a higher dielectric residence time and signal-to-noise ratio. Combined with the nano-silicon nitride anti-corrosion coating, it can maintain stable performance under harsh working conditions of long-term contact with polyphenolic plant components. At the same time, by extracting the S11 reflection coefficient features through fast vector fitting, and combining sliding filtering and anomaly rejection mechanism, the dielectric fingerprint vector output is ensured to be continuous and reliable. 5) This invention introduces a K-nearest neighbor matching and inverse distance weighted interpolation strategy, enabling the system to achieve smooth parameter transitions for atypical raw material samples and avoid control mutations caused by database coverage blind spots. Thus, without relying on complex optimization algorithms, it achieves coordinated dynamic adjustment of temperature gradient slope, two-phase flow rate ratio, and interface pH offset. The overall control accuracy reaches ±0.1℃, 0.01 mL / min, and ±0.05 pH units, respectively, fully meeting the stringent requirements for consistency and repeatability in the extraction of high-value face cream raw materials. Attached Figure Description

[0013] Figure 1 A flowchart illustrating the method of this invention; Figure 2 This is a schematic diagram of the acquisition process of the S11 reflection coefficient frequency response curve in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the process of calculating the real and imaginary parts of the complex permittivity in an embodiment of the present invention. Detailed Implementation

[0014] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0015] The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0016] like Figure 1 As shown, this embodiment provides a method for preparing a face cream based on biphasic extraction, specifically including the following steps: S1: Based on the experimental data of dielectric properties of typical plant raw material slurry samples, a dielectric fingerprint vector of the raw material is constructed. The experimental data includes three indicators: the amplitude of the real part of the complex dielectric constant, the amplitude of the imaginary part, and the spectral curvature measured at a frequency of 1.2 GHz. The three indicators are integrated into a three-dimensional vector through principal component regression analysis, which serves as an intrinsic physical feature identifier characterizing the moisture content, oil content, and cell wall integrity of the raw material. S2: Based on the correlation between the dielectric fingerprint vector of the raw material and the two-phase extraction efficiency, a pre-calibration spectrum is established; the correlation is obtained through orthogonal experiments, and a mapping relationship is established between the three-dimensional dielectric fingerprint vector and the optimal extraction parameter combination, which includes the temperature gradient slope, the ratio of organic phase to aqueous phase flow rate and the interface pH offset. S3: An embedded microwave resonant cavity sensor is integrated in the extraction feed pipeline. The sensor operates at a center frequency of 1.2 GHz and the cavity length is one-quarter of the waveguide wavelength. The frequency response curve of the S11 reflection coefficient of the raw material slurry is acquired through a microstrip feeding structure. S4: Based on the frequency response curve of the reflection coefficient in S11, calculate the real-time dielectric parameters of the raw material slurry; extract the resonant frequency shift and half-power bandwidth through fast vector fitting, and output stable real and imaginary values ​​of the complex dielectric constant after sliding window averaging and outlier removal. S5: Substitute the real-time dielectric parameters into the dielectric fingerprint vector of the raw material to generate the real-time dielectric fingerprint vector of the current raw material; the generation process is based on the three-dimensional vector structure defined in step S1, and combines the real part amplitude, imaginary part amplitude and spectral curvature of the filtered complex dielectric constant into a vector input; S6: Input the real-time dielectric fingerprint vector into the pre-calibrated spectrum and perform parameter mapping lookup; use the K-nearest neighbor algorithm to match the nearest neighbor sample cluster, perform weighted inverse distance interpolation on the parameter combination within the cluster, and output a dynamic adjustment instruction set, which includes temperature gradient slope instruction, flow rate ratio instruction and pH offset instruction. S7: The actuator is driven based on the dynamic adjustment instruction set; the temperature gradient slope instruction triggers independent temperature control of the zoned heating zone, the flow rate ratio instruction adjusts the dual-channel high-precision peristaltic pump, and the pH offset instruction links the micro pH fine-tuning module to realize closed-loop execution of extraction parameters; S8: Implement drift compensation calibration. When switching raw material batches, inject standard glycerol aqueous solution for zero-point calibration. After continuous operation for a preset time, dynamically fine-tune the mapping weight coefficient of the pre-calibrated spectrum according to the deviation of the product component extraction rate to maintain the long-term stability of parameter mapping.

[0017] In this embodiment, in step S1, a dielectric fingerprint vector of the raw material is constructed based on experimental data of the dielectric properties of typical plant raw material slurry samples. The experimental data includes three indicators: the amplitude of the real part of the complex dielectric constant, the amplitude of the imaginary part, and the spectral curvature measured at a frequency of 1.2 GHz. The three indicators are integrated into a three-dimensional vector through principal component regression analysis, serving as intrinsic physical feature identifiers characterizing the moisture content, oil content, and cell wall integrity of the raw material. Specifically, the steps are as follows: S1.1: Perform pulverization and homogenization pretreatment on typical plant raw material slurry samples to obtain standard samples; based on the 0.3–3 GHz microwave frequency band scanning experiment, collect the original data of the real part amplitude and the imaginary part amplitude of the complex dielectric constant at the 1.2 GHz frequency point as the basic input conditions for dielectric property experiments; S1.2: Perform spectral curvature calculation processing on the original data of the real part amplitude of the complex dielectric constant and the original data of the imaginary part amplitude of the complex dielectric constant to quantify the nonlinear characteristics of the dielectric parameter as a function of frequency; generate a structured experimental dataset based on a three-variable orthogonal experimental design of moisture content, oil content and particle size distribution. This dataset contains spectral curvature index values ​​and controls the combination of raw material characteristic parameters. S1.3: Based on the structured experimental dataset, principal component regression analysis algorithm is applied to perform dimensionality reduction to extract key features; the principal component loading coefficients of the three indices of real part magnitude of complex permittivity, imaginary part magnitude of complex permittivity, and spectral curvature are calculated, and the normalized feature vector component weights are output. S1.4: The principal component loading coefficients and the spectral curvature index values ​​are linearly combined to generate a standardized three-dimensional dielectric fingerprint vector. This vector consists of the normalized real part amplitude component of the complex dielectric constant, the imaginary part amplitude component of the complex dielectric constant, and the spectral curvature component, serving as a quantitative identifier of the intrinsic physical characteristics of the raw material. S1.5: Perform Pearson correlation verification on the three-dimensional dielectric fingerprint vector with the moisture content, oil content and cell wall integrity of the raw material to confirm its characterization effectiveness; output the verified dielectric fingerprint vector of the raw material based on the correlation coefficient analysis results to ensure that the vector serves as a reliable input for subsequent parameter mapping.

[0018] In this embodiment, in step S2, a pre-calibration spectrum is established based on the correlation between the dielectric fingerprint vector of the raw material and the two-phase extraction efficiency. The correlation is obtained through orthogonal experiments, and a mapping relationship is established between the three-dimensional dielectric fingerprint vector and the optimal extraction parameter combination. The optimal extraction parameter combination includes the temperature gradient slope, the organic phase / aqueous phase flow rate ratio, and the interface pH shift. Specifically, the steps are as follows: S2.1: Multi-factor orthogonal experimental design was carried out on the key variables of raw material characteristics, such as moisture content, oil content and particle size distribution. The range of level combinations of each variable was determined based on the L36 orthogonal table to generate an orthogonal experimental matrix covering typical raw material characteristics. S2.2: Based on the orthogonal experimental matrix, perform the two-phase extraction experimental process, collect the dielectric fingerprint vector of the raw material and the two-phase extraction efficiency index data for each group of experimental samples, so as to obtain a complete experimental dataset containing input variables and output efficiency. S2.3: Perform principal component regression analysis on the experimental dataset, extract the principal component feature parameters of the dielectric fingerprint vector of the raw material, and establish a mathematical correlation model between it and the two-phase extraction efficiency index to quantitatively characterize the influence of raw material characteristics on the extraction effect. S2.4: Based on the mathematical correlation model, perform parameter optimization and solution processing to determine the optimal combination of temperature gradient slope, organic phase / aqueous phase flow rate ratio and interface pH offset that maximizes the two-phase extraction efficiency, so as to generate the optimal extraction parameter instruction set for different raw material characteristics; S2.5: The optimal extraction parameter instruction set and the corresponding raw material dielectric fingerprint vector are mapped and stored to construct a pre-calibrated spectrum database, so as to realize the rapid lookup of the table to obtain the optimal parameter instruction during real-time extraction.

[0019] In this embodiment, step S3: An embedded microwave resonant cavity sensor is integrated into the extraction feed pipeline. The sensor operates at a center frequency of 1.2 GHz, and the cavity length is one-quarter of the waveguide wavelength. The frequency response curve of the S11 reflection coefficient of the raw material slurry is acquired through a microstrip feeding structure, as shown below. Figure 2 As shown, the specific steps include the following: S3.1: Based on the target center frequency of 1.2 GHz and the dielectric constant range of the raw material slurry, calculate the guided wave wavelength; determine the cavity length to be one-quarter of the guided wave wavelength according to the guided wave wavelength; design the rectangular cavity geometry to match the feed pipe diameter; integrate a U-shaped flow channel structure to extend the medium residence time of the raw material slurry in the cavity; apply a nano-scale silicon nitride anti-corrosion coating to the cavity surface; generate the design specifications for the embedded microwave resonant cavity sensor as the basis for subsequent physical integration; Calculate the guided wave wavelength using the dielectric waveguide propagation formula. The theoretical basis for cavity design is as follows: in, For the guided wave wavelength, For the speed of light in free space, For operating frequency, is the relative permittivity of the medium.

[0020] S3.2: Based on the design specifications of the embedded microwave resonant cavity sensor and the fluid dynamics parameters of the feed pipeline, perform mechanical coupling installation of the sensor and the pipeline; use flange connection and sealing ring to ensure no leakage; calibrate the sensor position to align with the microwave feed point; output the integrated pipeline-sensor assembly to provide the physical basis for the microstrip feed structure configuration; S3.3: Based on the integrated pipeline-sensor assembly and vector network analyzer, connect to the SMA interface and set the frequency band scanning range to cover 1.0 GHz to 1.4 GHz; perform open circuit, short circuit, and load calibration to eliminate background interference; configure the data acquisition rate and trigger conditions; output the configured S11 reflection coefficient acquisition system to ensure signal transmission stability; S3.4: Based on the configured S11 reflection coefficient acquisition system and the flowing raw material slurry, a microwave excitation signal is sent through a microstrip feeding structure; the reflected signal is received and the frequency domain response is recorded; the raw data is subjected to sliding window averaging filtering to remove high-frequency noise; and the S11 reflection coefficient frequency response curve data is output to provide input for real-time dielectric parameter calculation. S3.5: Based on the S11 reflection coefficient frequency response curve data, calculate the resonant cavity Q value and evaluate the noise level; check whether the resonance peak characteristics meet the preset threshold; if the data quality is not up to standard, trigger the re-acquisition process; output the verified S11 reflection coefficient frequency response curve to ensure that the data can be used for subsequent dielectric fingerprint vector generation.

[0021] In this embodiment, in step S4, based on the frequency response curve of the reflection coefficient in S11, the real-time dielectric parameters of the raw material slurry are calculated; the resonant frequency shift and half-power bandwidth are extracted through fast vector fitting, and after sliding window averaging filtering and outlier removal, stable real and imaginary values ​​of the complex dielectric constant are output, such as... Figure 3 As shown, the specific steps include the following: S4.1: An optimized implementation based on a vector fitting algorithm is constructed by pre-setting the initial pole distribution and iteratively minimizing the fitting error mechanism. The input is the frequency response curve of the S11 reflection coefficient, and the technical processing method is solving the linear equation system and iteratively updating the poles. The output is the resonant frequency offset and half-power bandwidth value, so as to achieve efficient parameter extraction under high spectral resolution. S4.2: Execute a fast vector fitting algorithm on the collected S11 reflection coefficient frequency response curve to extract the resonant frequency offset and half-power bandwidth; the input is the S11 reflection coefficient frequency response curve, the technical processing method is the application of the fast vector fitting algorithm, and the output is the resonant frequency offset and half-power bandwidth values, which serve as the basic input for reconstructing the complex permittivity frequency response characteristics. S4.3: Based on the extracted resonant frequency offset and half-power bandwidth, the complex dielectric constant frequency response characteristics of the raw material slurry are reconstructed using the pre-calibrated calibration coefficient relationship, and the real and imaginary part amplitudes of the complex dielectric constant are sampled at a frequency of 1.2 GHz; the inputs are the resonant frequency offset and half-power bandwidth values, the technical processing method is frequency response characteristic reconstruction and frequency sampling, and the outputs are the complex dielectric constant frequency response characteristics, the real part amplitude of the complex dielectric constant, and the imaginary part amplitude of the complex dielectric constant; S4.4: Based on the reconstructed frequency response characteristics of the complex permittivity, a quadratic polynomial local fitting is performed in the neighborhood of the 1.2 GHz frequency point to calculate the spectral curvature; the input is the frequency response characteristics of the complex permittivity, the technical processing method is the local polynomial fitting algorithm, and the output is the spectral curvature value, which serves as a key dimension parameter of the dielectric characteristics of the raw material.

[0022] Furthermore, by using the second derivative calculation method, the second derivative value of the fitted curve at the target frequency point is extracted as a direct quantitative index of the spectral curvature, and curvature numerical data is generated.

[0023] Furthermore, normalization is employed to make the curvature data dimensionless, thereby improving the comparability of curvature between different batches of raw materials.

[0024] Furthermore, statistical denoising is used to eliminate any abnormal curvature results that may occur during the fitting process, ensuring the stability and reliability of the output spectrum curvature.

[0025] S4.5: Perform sliding window averaging filtering and 3σ statistical outlier removal on the calculated real part magnitude, imaginary part magnitude, and spectral curvature values ​​of the complex permittivity. The input is the real part magnitude, imaginary part magnitude, and spectral curvature values ​​of the complex permittivity calculated in real time. The technical processing method is time-domain filtering and statistical anomaly detection. The output is the stable real part magnitude, imaginary part magnitude, and spectral curvature values ​​of the complex permittivity.

[0026] In this embodiment, in step S5, the real-time dielectric parameters are substituted into the dielectric fingerprint vector of the raw material to generate the real-time dielectric fingerprint vector of the current raw material. The generation process is based on the three-dimensional vector structure defined in step S1, and combines the real part amplitude, imaginary part amplitude, and spectral curvature of the filtered complex dielectric constant into a vector input, specifically including the following steps: S5.1: Perform range verification processing on the real part amplitude of the filtered complex permittivity output in step S4 to confirm that its value is within the preset range determined by the calibration experiment, and output the real part amplitude that has passed the verification. S5.2: Perform minimum-maximum normalization processing on the real part amplitude that has passed the verification, map the value to the standard interval [0,1], and generate the normalized real part amplitude; S5.3: Perform sliding window averaging filtering on the magnitude of the imaginary part of the complex permittivity output from S4. The window size is 5 seconds, and the output is a smooth imaginary part magnitude. In an online monitoring scenario for face cream ingredients based on the 0.5–2.5 GHz frequency band, the sampling system acquires real-time data of the imaginary part amplitude at a frequency response curve sampling rate of once per second. Assuming a sliding window width of 5 seconds, a time series of length 5 is formed, with sampled values ​​of 2.15, 2.18, 2.22, 2.17, and 2.20. The weighted moving average is calculated using a weight vector [0.05, 0.1, 0.15, 0.25, 0.45]. The formula is as follows: in, For the first The weight of each sampling point Substituting the sampled values ​​corresponding to the imaginary part amplitude into the weights yields a weighted average of 2.19. Then, the exponential smoothing formula is applied: ; in, The smoothing coefficient is set to 0.3. The current weighted average is 2.19. Given the previous smoothing value of 2.16, the current smoothing value is calculated. It is 2.17.

[0027] Finally, the output range is limited to [2.0, 2.3] by a threshold constraint function, generating a final smooth imaginary part amplitude of 2.17, which serves as the input for subsequent S5.4 drift compensation. Under different raw material batch conditions, it exhibits the characteristics of smooth curve without jumps and rapid response.

[0028] S5.4: Based on the smoothed imaginary part amplitude, a nonlinear correction function is applied to compensate for the long-term drift of the sensor and generate the corrected imaginary part amplitude; S5.5: Combine the normalized real part amplitude, the corrected imaginary part amplitude, and the spectral curvature output from S4 according to the three-dimensional vector structure defined in step S1 to generate a real-time dielectric fingerprint vector.

[0029] In this embodiment, in step S6, the real-time dielectric fingerprint vector is input into the pre-calibrated spectrum, and parameter mapping lookup is performed; the K-nearest neighbor algorithm is used to match the nearest neighbor sample clusters, and weighted inverse distance interpolation is performed on the parameter combinations within the clusters to output a dynamic adjustment instruction set. The instruction set includes temperature gradient slope instructions, flow rate ratio instructions, and pH offset instructions, specifically including the following steps: S6.1: Based on the spatial distribution characteristics of sample points in the pre-calibrated map, configure... Neighborhood parameters of the nearest neighbor algorithm =5; Based on the inverse square distance relationship, the weight calculation rule for weighted inverse distance interpolation is defined as the weight being inversely proportional to the square of the Euclidean distance, in order to generate the local sensitivity weight coefficients of the parameter mapping; S6.2: Input the real-time dielectric fingerprint vector as the query vector into the pre-calibrated graph, and execute the configured K-nearest neighbor algorithm search operation to match... The nearest neighbor sample points are used to determine the candidate sample cluster with the optimal combination of extraction parameters; S6.3: For each sample point in the candidate sample cluster, calculate its Euclidean distance to the query vector; based on the Euclidean distance, apply the defined weighted inverse distance interpolation rule to calculate the weighted average of the temperature gradient slope, the organic phase / water phase flow rate ratio, and the interface pH offset to generate the interpolation parameter vector. S6.4: Convert the interpolation parameter vector into specific control instructions to generate a dynamic adjustment instruction set including temperature gradient slope instructions, organic phase / aqueous phase flow rate ratio instructions and interface pH offset instructions, so as to adapt to the input interface requirements of the actuator. S6.5: Output the dynamic adjustment instruction set to the PID controller to drive the partitioned heating belt, dual-channel high-precision peristaltic pump and micro pH fine-tuning module to perform closed-loop adjustment, so as to realize real-time dynamic optimization of extraction parameters.

[0030] In this embodiment, in step S7, the actuator is driven based on the dynamic adjustment instruction set; the temperature gradient slope instruction triggers independent temperature control of the zoned heating zone, the flow rate ratio instruction adjusts the dual-channel high-precision peristaltic pump, and the pH offset instruction links the micro pH fine-tuning module to achieve closed-loop execution of extraction parameters, specifically including: S7.1: Perform PID parameter parsing on the temperature gradient slope command in the dynamic adjustment command set to generate zone heating zone temperature control commands; execute independent temperature control of the zone heating zone based on the temperature control commands to achieve the target temperature gradient slope; collect feedback data from the zone temperature sensors in real time to calculate the deviation between the actual temperature gradient slope and the target value; perform PID dynamic compensation adjustment based on the deviation to optimize the temperature control response speed and accuracy; output a temperature gradient slope execution confirmation signal to the closed-loop execution monitoring module. S7.2: Servo control parsing of the flow rate ratio commands in the dynamic adjustment command set to generate dual-channel high-precision peristaltic pump drive commands; adjusting the pump speed ratio of the organic phase and water phase channels based on the drive commands to achieve the target flow rate ratio; real-time monitoring of the dual-channel flow meter output data to calculate the deviation between the actual flow rate ratio and the command value; executing a proportional-integral-derivative compensation algorithm based on the deviation to stabilize the flow rate ratio and reduce fluctuations; outputting a flow rate ratio execution confirmation signal to the closed-loop execution monitoring module. A servo control parsing module is used to implement the flow rate ratio command decoding function, which decomposes the target flow rate ratio of the organic phase and water phase channels into a set of corresponding pump speed setpoints.

[0031] Furthermore, a drive instruction generation algorithm is used to generate PWM drive signals for a dual-channel high-precision peristaltic pump, and to obtain PWM signals for the organic phase channel and the aqueous phase channel.

[0032] Furthermore, the real-time flow acquisition function after driving is realized through the flow monitoring module, and the real-time flow values ​​of the organic phase channel and the aqueous phase channel are generated.

[0033] Furthermore, the real-time flow rate ratio is calculated using the flow rate ratio calculation formula: in, This represents the real-time flow rate value of the organic phase channel. This represents the real-time flow rate of the water phase channel. This represents the real-time flow rate ratio.

[0034] Furthermore, the deviation between the actual flow rate ratio and the commanded value is quantified through the deviation calculation formula: in, Set the target flow rate ratio. This represents the flow rate ratio deviation.

[0035] S7.3: Perform current pulse parsing on the pH offset commands in the dynamic adjustment command set to generate control signals for the micro pH fine-tuning module; drive the solid-state H based on the control signals. + Ion migration through the exchange membrane is used to achieve the target interface pH shift; real-time acquisition of pH sensor measurements is used to calculate the deviation between the actual pH shift and the commanded value; dynamic compensation with fine-tuning current is performed based on the deviation to maintain pH stability; and a pH shift execution confirmation signal is output to the closed-loop execution monitoring module.

[0036] In this embodiment, in step S8, drift compensation calibration is performed. A standard glycerol aqueous solution is injected for zero-point calibration during raw material batch switching. After continuous operation for a preset time, the mapping weight coefficients of the pre-calibrated spectrum are dynamically fine-tuned based on the deviation in product component extraction rate to maintain the long-term stability of parameter mapping. Specifically, this includes the following steps: S8.1: Automatically inject standard glycerol aqueous solution to calibrate the zero-point offset of the embedded microwave resonant cavity sensor and obtain the calibrated sensor reference parameters; S8.2: Perform continuous runtime monitoring based on the system runtime timer to determine whether the preset calibration trigger duration has been reached, and output the runtime calibration signal; S8.3: Based on the long-term target signal during runtime, perform high-performance liquid chromatography detection data acquisition and processing on the extracted product to analyze the actual extraction rate value of the target component and generate actual component extraction rate data; S8.4: Perform deviation calculation processing based on the actual component extraction rate data and the preset target extraction rate value to quantify the product component extraction rate deviation and output the extraction rate deviation value; S8.5: Based on the difference between the extraction rate deviation value and the preset deviation threshold, the condition judgment and dynamic adjustment of the mapping weight coefficient are performed to correct the parameter mapping relationship of the corresponding region in the pre-calibrated map and maintain the long-term stability of the parameter mapping.

[0037] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this invention.

[0038] Unless otherwise defined, the technical or scientific terms used herein should be understood in their ordinary sense by one of ordinary skill in the art to which this invention pertains. The terms “first,” “second,” “third,” and similar terms used in this patent application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms “comprising” or “including” and similar terms mean that the elements or objects preceding “comprising” or “including” encompass the elements or objects listed following “comprising” or “including” and their equivalents, and do not exclude other elements or objects. The “multiple” involved in the embodiments of this invention refers to two or more. A and / or B indicate three possibilities: A; B; and A and B.

[0039] The above description is merely an exemplary embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for preparing a face cream based on two-phase extraction, characterized in that, Includes the following steps: S1: Construct a dielectric fingerprint vector for the raw materials based on experimental data of dielectric properties of typical plant raw material slurry samples; S2: Based on the correlation between the dielectric fingerprint vector of raw materials and the two-phase extraction efficiency, a pre-calibration spectrum is established; S3: An embedded microwave resonant cavity sensor is integrated into the extraction feed pipeline to collect the S11 reflection coefficient frequency response curve of the raw material slurry; S4: Based on the frequency response curve of the reflection coefficient of S11, calculate the real-time dielectric parameters of the raw material slurry; extract the resonant frequency shift and half-power bandwidth through fast vector fitting, and output stable real and imaginary values ​​of the complex dielectric constant after sliding window averaging and outlier removal. S5: Substitute the real-time dielectric parameters into the material dielectric fingerprint vector to generate the real-time dielectric fingerprint vector of the current material. S6: Input the real-time dielectric fingerprint vector into the pre-calibrated spectrum, perform parameter mapping lookup; use the K-nearest neighbor algorithm to match the nearest neighbor sample cluster, perform weighted inverse distance interpolation on the parameter combination within the cluster, and output a dynamic adjustment instruction set; S7: Drives the actuator based on a dynamic adjustment instruction set; S8: Implement drift compensation calibration. When switching raw material batches, inject standard glycerol aqueous solution for zero-point calibration. After continuous operation for a preset time, dynamically fine-tune the mapping weight coefficient of the pre-calibrated spectrum according to the deviation of the product component extraction rate to maintain the long-term stability of parameter mapping.

2. The method for preparing a face cream based on biphasic extraction according to claim 1, characterized in that, In step S1, the experimental data includes three indicators: the real part amplitude, the imaginary part amplitude, and the spectral curvature of the complex permittivity measured at a frequency of 1.2 GHz. The three indicators are integrated into a three-dimensional vector through principal component regression analysis, which serves as an intrinsic physical characteristic identifier for characterizing the moisture content, oil content, and cell wall integrity of the raw materials.

3. The method for preparing a face cream based on biphasic extraction according to claim 1, characterized in that, In step S2, the correlation is obtained through orthogonal experiments, and a mapping relationship is established between the three-dimensional dielectric fingerprint vector and the optimal extraction parameter combination. The optimal extraction parameter combination includes the temperature gradient slope, the ratio of organic phase to aqueous phase flow rate, and the interface pH offset.

4. The method for preparing a face cream based on biphasic extraction according to claim 3, characterized in that, In step S2, a multi-factor orthogonal experimental design is performed on the key variables of raw material characteristics, namely moisture content, oil content and particle size distribution. The range of level combinations of each variable is determined based on the L36 orthogonal table, and an orthogonal experimental matrix is ​​generated. The two-phase extraction experimental process is executed based on the orthogonal experimental matrix. The dielectric fingerprint vector of the raw material and the two-phase extraction efficiency index data are collected for each experimental sample to obtain the experimental dataset. Principal component regression analysis was performed on the experimental dataset to extract the principal component feature parameters of the dielectric fingerprint vector of the raw materials, and a mathematical correlation model between the parameters and the two-phase extraction efficiency index was established. Based on the mathematical correlation model, the optimal combination of temperature gradient slope, organic phase, aqueous phase flow rate ratio and interface pH offset is determined to maximize the two-phase extraction efficiency, thus obtaining the optimal extraction parameter instruction set.

5. The method for preparing a face cream based on biphasic extraction according to claim 1, characterized in that, In step S3, a microwave excitation signal is transmitted through a microstrip feeding structure; the reflected signal is received and the frequency domain response is recorded. The original data is subjected to sliding window averaging filter to remove high-frequency noise; the frequency response curve data of the S11 reflection coefficient is output. Based on the frequency response curve data of the S11 reflection coefficient, the Q value of the resonant cavity is calculated and the noise level is evaluated. Check if the resonance peak characteristics meet the preset threshold; if the data quality is substandard, trigger the re-acquisition process. Output the frequency response curve of the S11 reflection coefficient that has passed verification.

6. The method for preparing a face cream based on biphasic extraction according to claim 1, characterized in that, In step S4, specifically: A fast vector fitting algorithm is constructed by pre-setting the initial pole distribution and iteratively minimizing the fitting error mechanism; A fast vector fitting algorithm is applied to the frequency response curve of the S11 reflection coefficient to extract the resonant frequency offset and half-power bandwidth. Based on the resonant frequency offset and half-power bandwidth, the frequency response characteristics of the complex dielectric constant of the raw material slurry are reconstructed using the pre-calibrated calibration coefficient relationship, and the amplitudes of the real and imaginary parts of the complex dielectric constant at a frequency of 1.2 GHz are sampled. Based on the reconstructed frequency response characteristics of the complex permittivity, a quadratic polynomial local fitting is performed in the neighborhood of the 1.2 GHz frequency point to calculate the spectral curvature; Sliding window averaging filtering and 3D filtering are applied to the real part magnitude, imaginary part magnitude, and spectral curvature values ​​of the complex permittivity. Outlier removal process; The magnitudes of the real part and imaginary part of the complex permittivity and the spectral curvature are obtained.

7. The method for preparing a face cream based on biphasic extraction according to claim 1, characterized in that, Step S5 specifically includes: The range verification process is performed on the real part magnitude of the filtered complex permittivity, and the real part magnitude that passes the verification is output. Based on the verified real part amplitude, perform minimum-maximum normalization to map the value to the standard interval [0,1] and generate the normalized real part amplitude. The magnitude of the imaginary part of the filtered complex permittivity is processed by a sliding window averaging filter to output a smooth imaginary part magnitude. Based on the smoothed imaginary part amplitude, a nonlinear correction function is applied to compensate for the long-term drift of the sensor and generate a corrected imaginary part amplitude. The normalized real part amplitude, the corrected imaginary part amplitude, and the spectral curvature are combined to generate a real-time dielectric fingerprint vector.

8. The method for preparing a face cream based on biphasic extraction according to claim 1, characterized in that, In step S6, based on the spatial distribution characteristics of sample points in the pre-calibrated map, the neighborhood parameters of the K-nearest neighbor algorithm are configured; according to the inverse square distance relationship, the weight calculation rule of the weighted inverse distance interpolation is defined as the weight being inversely proportional to the square of the Euclidean distance, so as to generate the local sensitive weight coefficients of the parameter mapping.

9. A method for preparing a face cream based on biphasic extraction according to claim 8, characterized in that, In step S6, the real-time dielectric fingerprint vector is used as the query vector input to the pre-calibrated map, and the configured K-nearest neighbor algorithm search operation is performed to match the K nearest neighbor sample points in order to determine the candidate sample cluster with the optimal combination of extraction parameters. For each sample point in the candidate sample cluster, calculate its Euclidean distance to the query vector; based on the Euclidean distance, apply the defined weighted inverse distance interpolation rule to calculate the weighted average of the temperature gradient slope, the ratio of organic phase to aqueous phase flow rate, and the interface pH offset to generate the interpolation parameter vector. The interpolation parameter vector is converted into specific control commands, generating a dynamic adjustment command set that includes commands for temperature gradient slope, organic phase and aqueous phase flow rate ratio, and interface pH offset.

10. A method for preparing a face cream based on biphasic extraction according to claim 9, characterized in that, In step S7, the dynamic adjustment instruction set is converted into actuator drive signals through digital-analog conversion. The temperature gradient slope signal is mapped to 0–5, the flow rate ratio signal is mapped to 0–3, the pH offset is mapped to 0–1, the signal quantization step size is 0.01, and CRC checksum is used to ensure data packet integrity.

Images