Artemisia argyi essential oil extraction equipment control method and system

Abstract

The present application belongs to the technical field of equipment control, and relates to a kind of wormwood essential oil extraction equipment control method and system.The method collects the temperature of multiple positions in the kettle during distillation, heating steam pressure, condenser outlet liquid temperature and acoustic emission signal in real time, comprehensively senses the distillation state;Then the signal is input into a double-channel gated recurrent network model that has been dynamically modulated by trajectory entropy and can fuse physical model for Bayesian correction under stable working conditions to accurately measure the soft measurement value of the wormwood essential oil oil yield;After that, based on the oil yield deviation, a multi-objective optimization function containing the square of the deviation, the control amount change rate, the thermal shock gradient and the boiling instability risk is solved to obtain the control sequence;Finally, the control sequence is decoupled to determine the precise allocation strategy for heating steam pressure and flow and execute.The present application can maximize the efficiency of wormwood essential oil extraction while ensuring smooth and safe operation.

Description

Technical Field

[0001] This invention belongs to the field of equipment control technology, specifically relating to a control method and system for an artemisia essential oil extraction device. Background Technology

[0002] Steam distillation is the mainstream technology for industrial extraction of Artemisia argyi essential oil. Its principle involves using steam to carry volatile oily components from Artemisia argyi cells, which are then separated by condensation to obtain the essential oil. In this process, the oil yield, product quality, and energy consumption are directly determined by the precision of controlling key process parameters such as the temperature field inside the distillation vessel and the steam supply. Therefore, accurately controlling these parameters is crucial for ensuring the effectiveness of Artemisia argyi essential oil extraction.

[0003] However, in actual production, the essential oil yield, a core controlled variable, cannot be directly measured online via sensors like parameters such as temperature and pressure. Its acquisition typically relies on offline analysis or manual observation, resulting in significant lag in control feedback and hindering real-time regulation. Although the industry has adopted soft sensor technology to estimate the yield in real time by establishing mathematical models of easily measurable parameters (such as temperature and pressure) and essential oil yield, existing soft sensor models, whether based on simplified mechanisms or conventional data-driven models, struggle to adapt to the complex characteristics of the distillation process of Artemisia argyi, such as time-varying mass transfer resistance and nonlinear coupling of thermodynamic states caused by changes in material state. This leads to insufficient estimation accuracy and fails to provide a reliable basis for control strategies.

[0004] At the control strategy level, current methods for extracting Artemisia argyi essential oil rely heavily on manual experience to control single parameters such as the temperature at the top of the vessel. This approach overlooks two key constraints in the production process: First, improper control of the heating rate can easily trigger severe thermal shock, leading to excessive temperature gradients in the material layer within the vessel. This can not only cause the decomposition of heat-sensitive active ingredients in the essential oil, reducing product quality, but also damage the equipment. Second, unstable steam supply or excessive heating intensity can easily lead to violent boiling instability or even bumping of the liquid within the vessel, severely deteriorating the heat and mass transfer efficiency between the vapor and liquid phases and posing safety hazards. Therefore, it is evident that existing technologies struggle to maximize the extraction efficiency of Artemisia argyi essential oil while ensuring a stable and safe process. Summary of the Invention

[0005] Therefore, the purpose of this invention is to provide a control method and system for mugwort essential oil extraction equipment, so as to solve the technical problem that the existing technology is unable to maximize the extraction efficiency of mugwort essential oil while ensuring stability and safety.

[0006] To solve the above problems, the technical solution of the control method for artemisia essential oil extraction equipment provided by the present invention is as follows:

[0007] A method for controlling an artemisia essential oil extraction device includes the following steps:

[0008] S1, real-time acquisition of multiple temperatures, heating steam pressures, condenser outlet liquid phase temperature and acoustic emission signals distributed along the central axis of the distillation vessel;

[0009] S2, the collected signal is input to the corresponding channel of the dual-channel gated loop network model, and the trajectory entropy calculated by heating steam pressure and the top temperature of the kettle among multiple temperatures is used to nonlinearly modulate the gate unit of the dual-channel gated loop network model; the dual-channel gated loop network model outputs the direct output value of the oil yield of Artemisia argyi essential oil and a mass transfer resistance parameter in parallel. When the rate of change of the mass transfer resistance parameter is lower than a preset threshold, the direct output value of the dual-channel gated loop network model and the output value of the Fick's law diffusion model corrected by the mass transfer resistance parameter are fused in Bayesian to obtain the soft measurement value of the oil yield of Artemisia argyi essential oil;

[0010] S3. Based on the deviation between the soft measurement value and the preset target oil yield, solve the multi-objective optimization function that includes the square of the deviation, the rate of change of the control quantity, the thermal shock gradient and the boiling instability risk to obtain the control sequence.

[0011] S4. Decouple the control sequence based on the trajectory entropy, determine the control allocation strategy for the pressure and flow rate of the heating steam, and execute the corresponding control commands.

[0012] Furthermore, the dual-channel gated loop network model includes a first input channel and a second input channel. The first input channel processes signals reflecting the macroscopic thermodynamic state of the extraction process, and the second input channel processes signals characterizing the microscopic dynamics of the extraction process. The time-series signals of multiple temperatures, heating steam pressures, and condenser outlet liquid phase temperatures are input to the first input channel; the time-series signal of the acoustic emission signal is input to the second input channel.

[0013] Further, in step S2, the method for nonlinearly modulating the gating unit of the dual-channel gated recurrent network model includes:

[0014] S21, construct a two-dimensional phase space from the time series data of the reactor top temperature and the heating steam pressure;

[0015] S22, the two-dimensional phase space is divided into 10×10 rectangular regions and symbolized, and the trajectory is mapped into a symbol sequence;

[0016] S23, Calculate the trajectory entropy by counting the occurrence probability of a symbol sequence of length 3 using the Shannon entropy formula;

[0017] S24 uses the trajectory entropy as a modulation factor and performs element-wise multiplication with the outputs of the update gate and reset gate in the dual-channel gated recurrent network to achieve nonlinear modulation.

[0018] Furthermore, the method for Bayesian fusion of the direct output value of the dual-channel gated recurrent network model and the output value of the Fick's law diffusion model corrected using the mass transfer resistance parameter includes:

[0019] S25, calculate in real time the linear regression slope of the mass transfer resistance parameter output by the dual-channel gated loop network model over the past 10 minutes, and use the linear regression slope as the rate of change;

[0020] S26, When the absolute value of the rate of change is lower than a preset absolute value threshold, Bayesian fusion is initiated;

[0021] S27, estimate the prediction variance of the output value of the dual-channel gated recurrent network model and the output value of the corrected Fick's law diffusion model respectively;

[0022] S28. According to the Bayesian estimation formula, the output values ​​of the two models are weighted and averaged according to the reciprocal of the predicted variance of the output values ​​to obtain the soft measurement value of the oil yield of Artemisia argyi essential oil.

[0023] Furthermore, the thermal shock gradient is calculated based on the second spatial derivatives of the plurality of temperatures, and the calculation method includes:

[0024] S31, assuming distribution from top to bottom along the central axis of the distillation vessel. There are 1 temperature measurement points, and the temperature at each temperature measurement point is recorded as . The location of the temperature measurement point is recorded as ,in, ;

[0025] S32, the second spatial derivative of each temperature measurement point is calculated using the central difference formula. The formula is as follows:

[0026] ;in, For the first The temperature at each temperature measurement point For the first The temperature at each temperature measurement point For the first The location of each temperature measurement point For the first The location of each temperature measurement point;

[0027] S33, take the maximum absolute value of the second spatial derivative of all temperature measurement points as the thermal shock gradient at the current moment.

[0028] Furthermore, the boiling instability risk is calculated based on the high-frequency energy spectral entropy of the acoustic emission signal, and the calculation method includes:

[0029] S34 extracts information of the acoustic emission signal in the 100kHz-300kHz frequency band and filters out interference noise from all other frequency bands;

[0030] S35. Apply a Hanning window to the filtered acoustic emission signal, perform a 2048-point fast Fourier transform, and calculate the power spectral density.

[0031] S36, normalize the power spectral density values ​​at each frequency point into probabilities. The normalization formula is:

[0032] ,in, For the first Each frequency point, For the first Each frequency point, For the first Frequency points The power spectral density, For the first Frequency points The power spectral density, For the first Normalized probability values ​​for each frequency point;

[0033] S37. The energy spectrum entropy is calculated using the Shannon entropy formula and serves as a quantitative indicator of boiling instability risk. The formula for calculating the energy spectrum entropy is as follows: ,in, This represents the entropy of the energy spectrum.

[0034] Preferably, in step S34, a Butterworth bandpass filter is used to filter the acoustic emission signal.

[0035] Furthermore, the multi-objective optimization function is:

[0036] ;

[0037] in, To predict discrete time steps in the time domain; For the first The deviation between the measured soft measurement value and the preset target oil yield. For the first The rate of change of the control quantity at each step For the first thermal shock gradient of the step, For the first The risk of boiling instability in the step; These are the preset weighting coefficients for positive constants; For prediction in the time domain;

[0038] In step S3, the process of solving the multi-objective optimization function that includes the integral of the deviation, the rate of change of the control quantity, the thermal shock gradient, and the boiling instability risk is to solve the function in the prediction time domain. The multi-objective optimization function within The minimum value.

[0039] Further, in step S4, the method for determining the control allocation strategy for the pressure and flow rate of the heating steam by decoupling the control sequence based on the trajectory entropy is as follows:

[0040] Based on the preset upper and lower limits of trajectory entropy, the real-time calculated trajectory entropy is converted into a normalized dynamic factor.

[0041] Based on the normalized dynamic factor, the pressure control weight and flow control weight for heating steam are determined. When the normalized dynamic factor approaches 0, the pressure control weight is higher; when the normalized dynamic factor approaches 1, the flow control weight is higher.

[0042] The technical solution of the control system for the artemisia essential oil extraction equipment provided by this invention is as follows:

[0043] A control system for a mugwort essential oil extraction device includes a processor and a memory. The memory stores a computer program, and the processor executes the computer program to implement a control method for a mugwort essential oil extraction device as described in the above-mentioned technical solutions.

[0044] The beneficial effects of this invention are:

[0045] This invention constructs a dual-channel gated cyclic network model, processing macroscopic state signals such as temperature and pressure separately from high-frequency microscopic signals such as acoustic emission, enabling a more comprehensive capture and extraction of process dynamics. The invention utilizes the trajectory entropy calculated from the heating steam pressure and the vessel top temperature to nonlinearly modulate the gated units of the dual-channel gated cyclic network model, allowing the model to dynamically adjust its memory and update rates according to process complexity. Furthermore, this invention combines the dual-channel gated cyclic network model, adept at handling complex dynamics, with a physical mechanism-based Fick's law model using a Bayesian fusion strategy. This ensures that the estimation results adapt to both dynamic process changes and are constrained by physical laws under stable operating conditions, improving the accuracy and reliability of online estimation of Artemisia argyi essential oil yield.

[0046] Compared with existing technologies, this invention constructs a multi-objective optimization function and innovatively introduces two risk quantification indicators: calculating the thermal shock gradient based on the second-order spatial derivative of multi-point temperature and calculating the boiling instability risk based on the high-frequency energy spectrum entropy of acoustic emission signals. While pursuing the oil yield target, these two risk quantification indicators are incorporated as penalty terms into the multi-objective optimization function. This allows the control system to proactively avoid operations that might lead to product quality degradation or safety issues when making each decision, maximizing extraction efficiency while ensuring process stability and safety. Furthermore, this invention decouples the control sequence based on trajectory entropy, dynamically allocating control weights for steam pressure and flow rate, further ensuring control accuracy and process stability. Ultimately, it achieves a balance between extraction efficiency, product quality, and process safety in the extraction of Artemisia argyi essential oil. Attached Figure Description

[0047] Figure 1 This is a flowchart illustrating the steps of a method for controlling an artemisia essential oil extraction device according to the present invention. Detailed Implementation

[0048] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0049] A specific embodiment of the control method for artemisia essential oil extraction equipment of the present invention:

[0050] like Figure 1 As shown, a control method for an artemisia essential oil extraction device includes the following steps:

[0051] S1 collects multiple temperatures, heating steam pressures, condenser outlet liquid phase temperature, and acoustic emission signals distributed along the central axis of the distillation vessel in real time.

[0052] In this step, PT100 platinum resistance temperature sensors are arranged along the central axis at the top, middle, and bottom of the distillation vessel to measure the temperatures at the top, middle, and bottom of the vessel. A pressure transmitter is installed on the inlet pipe of the heating steam to measure the heating steam pressure. A thermocouple is installed at the cooling water outlet of the condenser to measure the liquid phase temperature at the condenser outlet. A broadband piezoelectric acoustic emission sensor is attached to the lower middle part of the outer wall of the distillation vessel to collect acoustic wave signals with a frequency band of 20kHz to 1MHz generated during the boiling process inside the vessel. The temperature and pressure signals are sampled at a frequency of 10Hz using a data acquisition card, and the acoustic emission signals are collected at a sampling rate of 2.5MHz.

[0053] S2, the collected signal is input to the corresponding channel of the dual-channel gated loop network model, and the trajectory entropy calculated by the heating steam pressure and the top temperature of the kettle among multiple temperatures is used to nonlinearly modulate the gate unit of the dual-channel gated loop network model; the dual-channel gated loop network model outputs the direct output value of the oil yield of Artemisia argyi essential oil and a mass transfer resistance parameter in parallel. When the rate of change of the mass transfer resistance parameter is lower than a preset threshold, the direct output value of the dual-channel gated loop network model and the output value of the Fick's law diffusion model corrected by the mass transfer resistance parameter are fused in Bayesian order to obtain a soft measurement value of the oil yield of Artemisia argyi essential oil.

[0054] This step requires establishing a dual-channel gated recurrent network model, which includes a first input channel, a second input channel, a feature fusion layer, and an output layer. The first input channel is specifically used to process low-frequency time-series signals characterizing the macroscopic thermodynamic state of the extraction process, such as time-series signals of multiple temperatures distributed along the central axis of the distillation vessel, heating steam pressure, and condenser outlet liquid phase temperature. The second input channel is specifically used to process signals characterizing the microscopic dynamics of the extraction process, particularly high-frequency acoustic emission signals of the boiling state. Both the first and second input channels are composed of one or more stacked gated recurrent unit layers, and the two input channels process their respective input signals in parallel. The feature fusion layer concatenates the feature vectors output from the two input channels or merges them using other weighting methods to obtain a global feature vector. This global feature vector is input into a fully connected layer as the output layer, and two key outputs are calculated using a nonlinear mapping: a direct prediction of the oil yield of Artemisia argyi essential oil, and a parameter representing the mass transfer efficiency of the current distillation extraction process, i.e., the mass transfer resistance parameter.

[0055] In the network structure design, besides the main output terminal used to predict the oil yield of Artemisia argyi essential oil, a parallel output terminal is obtained to output an implicit mass transfer resistance parameter. The change in the mass transfer resistance parameter per unit time is calculated. When this change is less than a preset threshold, such as 0.01, it indicates that the distillation extraction process has entered a stable late stage dominated by diffusion, at which point the Bayesian fusion mechanism is activated. Fick's law diffusion model states that under steady-state conditions, the diffusion flux of a substance is proportional to the concentration gradient of the substance. ,in, Let D be the diffusion flux, and D be the diffusion coefficient. Let R be the concentration gradient. On one hand, the classical Fick's law diffusion model is corrected using the mass transfer resistance parameter, denoted as R. In practical applications, the diffusion coefficient D in the Fick's law diffusion model is preferably corrected to D / R or to... , and It is a predicted value based on a physical mechanism, obtained from a normal value pre-calibrated using experimental data. On the other hand, the direct output value of the two-channel gated recurrent network model is retained, denoted as... Assuming both output values ​​follow a Gaussian distribution, the variance is determined based on their respective historical prediction errors, and then Bayes' theorem is used to... and Weighted fusion is performed to obtain the optimal estimate with the minimum variance, which serves as the soft measurement value for the oil yield of Artemisia argyi essential oil. In this field, the soft measurement value is obtained by using one or more process variables that are easily measured online in real time, and then indirectly estimating or inferring a parameter that is difficult to measure directly through an established mathematical model.

[0056] In an optional embodiment, the method for nonlinearly modulating the gating units of the dual-channel gated recurrent network model includes:

[0057] S21, construct a two-dimensional phase space from the time series data of the reactor top temperature and the heating steam pressure;

[0058] S22, the two-dimensional phase space is divided into 10×10 rectangular regions and symbolized, and the trajectory is mapped into a symbol sequence;

[0059] S23, Calculate the trajectory entropy by counting the occurrence probability of a symbol sequence of length 3 using the Shannon entropy formula;

[0060] S24 uses the trajectory entropy as a modulation factor and performs element-wise multiplication with the outputs of the update gate and reset gate in the dual-channel gated recurrent network to achieve nonlinear modulation.

[0061] For example, suppose the real-time data points of the vessel top temperature and heating steam pressure form a trajectory in a two-dimensional phase space. Assume the symbolic region sequence traversed by the trajectory over the past five minutes is ACDCAB. By statistically analyzing the frequency of all subsequences of length three, such as ACA and CDC, the trajectory entropy value can be calculated using the Shannon entropy formula. A high trajectory entropy value indicates that the system state fluctuates wildly and is disordered. Within the dual-channel gated recurrent network model, the calculated trajectory entropy value is directly multiplied by the activation values ​​of the update and reset gates. This enhances the sensitivity of the dual-channel gated recurrent network model to the current input when the system is unstable, reduces its reliance on historical information, and allows the model to adapt to changes in operating conditions more quickly.

[0062] In an optional embodiment, the method for Bayesian fusion of the direct output of the dual-channel gated recurrent network model with the output of the Fick's law diffusion model corrected using the mass transfer resistance parameter includes:

[0063] S25, calculate in real time the linear regression slope of the mass transfer resistance parameter output by the dual-channel gated loop network model over the past 10 minutes, and use the linear regression slope as the rate of change;

[0064] S26, When the absolute value of the rate of change is lower than a preset absolute value threshold, Bayesian fusion is initiated;

[0065] S27, estimate the prediction variance of the output value of the dual-channel gated recurrent network model and the output value of the corrected Fick's law diffusion model respectively;

[0066] S28. According to the Bayesian estimation formula, the output values ​​of the two models are weighted and averaged according to the reciprocal of the predicted variance of the output values ​​to obtain the soft measurement value of the oil yield of Artemisia argyi essential oil.

[0067] S3: Based on the deviation between the soft measurement value and the preset target oil yield, solve a multi-objective optimization function that includes the square of the deviation, the rate of change of the control quantity, the thermal shock gradient, and the boiling instability risk to obtain a control sequence; wherein, the thermal shock gradient is calculated based on the second spatial derivative of the multiple temperatures, and the boiling instability risk is calculated based on the high-frequency energy spectrum entropy of the acoustic emission signal.

[0068] In an optional embodiment, the multi-objective optimization function is:

[0069] ;

[0070] in, To predict discrete time steps in the time domain; For the first The deviation between the measured soft measurement value and the preset target oil yield. For the first The rate of change of the control quantity at each step For the first thermal shock gradient of the step, For the first The risk of boiling instability in the step; These are the preset weighting coefficients for positive constants; For prediction in the time domain.

[0071] The multi-objective optimization function is the core of advanced control strategies; it achieves optimal control by balancing multiple conflicting performance metrics. Solving the multi-objective optimization function involves solving for the function in the prediction time domain. Multi-objective optimization function within The minimum value is then used to obtain the future. The control sequence within each control cycle ensures product quality while avoiding drastic operations and mitigating equipment safety risks. The first term in the multi-objective optimization function includes the square of the oil yield deviation to ensure tracking accuracy; the second term includes the rate of change of the control quantity, i.e., the rate of change of heating steam pressure and flow rate, to ensure control stability; the third term includes the thermal shock gradient, which reflects the most severe temperature distribution non-uniformity and is a key parameter for preventing equipment damage due to excessive thermal stress; the fourth term includes the boiling instability risk, indicating the degree of safety risk.

[0072] In an optional embodiment, the thermal shock gradient is calculated based on the second spatial derivative of the plurality of temperatures, and the calculation method includes:

[0073] S31, assuming distribution from top to bottom along the central axis of the distillation vessel. There are 1 temperature measurement points, and the temperature at each temperature measurement point is recorded as . The location of the temperature measurement point is recorded as ,in, ;

[0074] S32, the second spatial derivative of each temperature measurement point is calculated using the central difference formula. The formula is as follows:

[0075] ;in, For the first The temperature at each temperature measurement point For the first The temperature at each temperature measurement point For the first The location of each temperature measurement point For the first The location of each temperature measurement point;

[0076] S33, take the maximum absolute value of the second spatial derivative of all temperature measurement points as the thermal shock gradient at the current moment.

[0077] In an optional embodiment, the boiling instability risk is calculated based on the high-frequency energy spectral entropy of the acoustic emission signal, and the calculation method includes:

[0078] S34, extract information of the acoustic emission signal in the frequency band of 100kHz-300kHz and filter out interference noise in all other frequency bands; in this step, it is preferable to use a Butterworth bandpass filter to filter the acoustic emission signal;

[0079] S35. Apply a Hanning window to the filtered acoustic emission signal, perform a 2048-point fast Fourier transform, and calculate the power spectral density.

[0080] S36, normalize the power spectral density values ​​at each frequency point into probabilities. The normalization formula is:

[0081] ,in, For the first Each frequency point, For the first Each frequency point, For the first Frequency points The power spectral density, For the first Frequency points The power spectral density, For the first Normalized probability values ​​for each frequency point;

[0082] S37. The energy spectrum entropy is calculated using the Shannon entropy formula and serves as a quantitative indicator of boiling instability risk. The formula for calculating the energy spectrum entropy is as follows: ,in, This represents the entropy of the energy spectrum.

[0083] Energy spectral entropy is used to assess the stability of boiling processes. In stable nucleate boiling, bubble generation and collapse are random and widespread, resulting in a uniform distribution of acoustic emission signal energy across a wide frequency band, leading to a high calculated energy spectral entropy value. However, when the process becomes unstable, such as with film boiling or bubble coalescence, the acoustic emission signal energy concentrates at a few specific frequencies. After filtering and Fourier transform, the power spectrum is found to be mainly concentrated in two narrow bands at 150kHz and 250kHz, with very low energy at other frequencies. This non-uniform energy distribution causes a decrease in the calculated energy spectral entropy value.

[0084] S4: Decouple the control sequence based on the trajectory entropy, determine the control allocation strategy for the pressure and flow rate of the heating steam, and obtain and execute the corresponding control commands.

[0085] In this step, when the real-time calculated trajectory entropy value is low, it indicates that the distillation extraction process is stable. At this time, the decoupling strategy will allocate higher weights from the unified control sequence output by the multi-objective optimization function to the pressure controller of the heating steam, achieving rapid adjustment of heating efficiency and thus quickly tracking the target oil yield. Conversely, when the trajectory entropy value rises and exceeds a certain threshold, it indicates that the distillation extraction process is becoming unstable. The decoupling strategy will then allocate more control weights to the flow controller of the heating steam, stabilizing the energy input through a smoother, stronger integral effect adjustment method, while suppressing drastic fluctuations in pressure control, thereby executing control commands while ensuring safety and stability.

[0086] Specifically, obtain the lower limit of trajectory entropy under highly stable conditions. Upper limit of trajectory entropy in highly unstable states Based on the upper and lower limits of trajectory entropy, the real-time calculated trajectory entropy is converted into a normalized dynamic factor. A normalized dynamic factor approaching 0 indicates a very stable distillation extraction process with a high weight for pressure control; a normalized dynamic factor approaching 1 indicates a very unstable distillation extraction process with a high weight for flow rate control. According to... Calculate pressure control weight ,according to Calculate flow control weights The pressure control command increment and flow control command increment are obtained by using weights. The calculated command increment is added to the set value of the previous moment to obtain the control value of the current moment.

[0087] The control method for the artemisia essential oil extraction equipment of the present invention achieves synergistic optimization of artemisia essential oil yield, product quality and process safety, and improves the level of process automation.

[0088] A specific embodiment of the control system for a mugwort essential oil extraction device provided by the present invention:

[0089] The control system for the artemisia essential oil extraction equipment includes a processor and a memory. The memory stores a computer program, and when the computer program is executed by the processor, it can realize the artemisia essential oil extraction equipment control method in the above embodiments.

[0090] The control system of the artemisia essential oil extraction equipment also includes other components well known to those skilled in the art, such as communication buses and communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0091] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as resistive random access memory (RRAM), dynamic random access memory (DRAM), static random access memory (SRAM), enhanced dynamic random access memory (EDRAM), high-bandwidth memory (HBM), hybrid memory cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented by computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0092] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. An artemisia oil extraction apparatus control method, characterized by, The method comprises the following steps: S1, collecting multiple temperatures distributed along the central axis direction of the distillation kettle, heating steam pressure, condenser outlet liquid phase temperature and acoustic emission signals in real time; S2, inputting the collected signals into the corresponding channels of the dual-channel gated recurrent network model, and performing nonlinear modulation on the gating unit of the dual-channel gated recurrent network model by using the trajectory entropy calculated from the heating steam pressure and the kettle top temperature among the multiple temperatures; The dual-channel gated recurrent network model outputs a direct output value of the artemisia oil yield and a mass transfer resistance parameter in parallel, when the change rate of the mass transfer resistance parameter is lower than a preset threshold, the direct output value of the dual-channel gated recurrent network model is fused with the output value of the Fick's law diffusion model corrected by the mass transfer resistance parameter to obtain a soft measurement value of the artemisia oil yield; S3, according to the deviation of the soft measurement value and the preset target oil yield, a multi-objective optimization function containing the square of the deviation, the change rate of the control amount, the thermal shock gradient and the boiling instability risk is solved to obtain the control sequence; S4, decoupling the control sequence according to the trajectory entropy, determining the control allocation strategy of the pressure and flow of the heating steam, and executing the corresponding control instructions; The method for nonlinear modulation on the gating unit of the dual-channel gated recurrent network model comprises: S21, constructing a two-dimensional phase space with the time series data of the kettle top temperature and the heating steam pressure; S22, dividing the two-dimensional phase space into 10×10 rectangular regions and symbolizing, and mapping the trajectory into a symbol sequence; S23, calculating the trajectory entropy according to the Shannon entropy formula by counting the occurrence probability of the symbol sequence with a length of 3; S24, taking the trajectory entropy as a modulation factor, and performing element-by-element multiplication operation with the outputs of the update gate and the reset gate in the dual-channel gated recurrent network to realize nonlinear modulation.

2. The control method of the equipment for extracting artemisia oil according to claim 1, characterized in that, The dual-channel gated recurrent network model comprises a first input channel and a second input channel, the first input channel processes signals reflecting the macroscopic thermodynamic state of the distillation extraction process, and the second input channel processes signals representing the microcosmic dynamics of the distillation extraction process; the time series signals of the multiple temperatures, the heating steam pressure and the condenser outlet liquid phase temperature are used for input into the first input channel; The time series signal of the acoustic emission signal is used for input into the second input channel.

3. The control method of the equipment for extracting artemisia oil according to claim 1, characterized in that, The method for Bayesian fusion of the direct output value of the dual-channel gated recurrent network model and the output value of the Fick's law diffusion model corrected by the mass transfer resistance parameter comprises: S25, calculating the linear regression slope of the mass transfer resistance parameter output by the dual-channel gated recurrent network model in the past 10 minutes in real time, and taking the linear regression slope as the change rate; S26, when the absolute value of the change rate is lower than a preset absolute value threshold, starting Bayesian fusion; S27, respectively estimating the prediction variances of the output value of the dual-channel gated recurrent network model and the output value of the corrected Fick's law diffusion model; S28, according to the Bayesian estimation formula, weighting and averaging the output values of the two models according to the inverses of the prediction variances of the output values to obtain the soft measurement value of the artemisia oil yield.

4. The control method of the equipment for extracting artemisia oil according to claim 1, characterized in that, The thermal shock gradient is calculated based on the second-order spatial derivative of the plurality of temperatures, and the calculation method comprises: S31, set along the distillation kettle center axis direction from top to bottom distribution temperature measurement points, the temperature of the temperature measurement points is recorded as , the position of the temperature measurement point is recorded as , wherein ; S32, the second-order spatial derivative of each temperature measurement point is calculated by using a central difference formula, and the calculation formula is: ; wherein, is the temperature of the temperature measurement point, is the temperature of the temperature measurement point, is the position of the temperature measurement point, is the position of the temperature measurement point; S33, the maximum value of the absolute values of the second-order spatial derivatives of all temperature measurement points is taken as the thermal shock gradient at the current moment.

5. The control method of the equipment for extracting artemisia oil according to claim 1, characterized in that, The boiling instability risk is calculated based on the high-frequency energy spectrum entropy of the acoustic emission signal, and the calculation method comprises: S34, information of the acoustic emission signal in the frequency band of 100kHz-300kHz is extracted, and interference noise in all other frequency bands is filtered out; S35, a Hanning window is applied to the filtered acoustic emission signal, and fast Fourier transform of 2048 points is performed to calculate the power spectrum density; S36, the values of the power spectrum density at each frequency point are normalized as probabilities, and the normalization formula is: wherein, is the power spectral density at the th frequency point, is the power spectral density at the th frequency point, is the power spectral density at the th frequency point is the power spectral density at the th frequency point is the power spectral density at the th frequency point, is the normalized probability value at the th frequency point. S37, according to the Shannon entropy formula, the energy spectrum entropy is calculated as a quantitative indicator of the risk of boiling instability, and the calculation formula of the energy spectrum entropy is: wherein, denotes the energy spectrum entropy.

6. The method of claim 5, wherein the method further comprises: In step S34, a Butterworth band-pass filter is used to filter the acoustic emission signal.

7. The control method of the equipment for extracting artemisia oil according to claim 1, characterized in that, The multi-objective optimization function is: ； in, To predict discrete time steps in the time domain; For the first The deviation between the measured soft measurement value and the preset target oil yield. For the first The rate of change of the control quantity of the step For the first thermal shock gradient of the step, For the first The risk of boiling instability in the step; These are the preset weighting coefficients for positive constants; For prediction in the time domain; In step S3, the process of solving the multi-objective optimization function of the integral of the deviation, the control variable rate of change, the thermal shock gradient and the boiling instability risk comprises solving the minimum value of the multi-objective optimization function in the prediction time domain . ​ 8. The control method of the equipment for extracting artemisia oil according to claim 1, characterized in that, In step S4, the method for determining the control distribution strategy for the pressure and flow of the heating steam according to the trajectory entropy decoupling of the control sequence is: According to the preset upper limit value and lower limit value of the trajectory entropy, the trajectory entropy calculated in real time is converted into a normalized dynamic factor; According to the normalized dynamic factor, the pressure control weight and the flow control weight of the heating steam are determined, wherein when the normalized dynamic factor tends to 0, the pressure control weight is higher; and when the normalized dynamic factor tends to 1, the flow control weight is higher.

9. An artemisia oil extraction apparatus control system, characterized by, The device comprises a processor and a memory, the memory stores a computer program, and the processor executes the computer program to implement the control method of the wormwood essential oil extraction equipment according to any one of claims 1-8.

Images