A multi-parameter linkage control method for the extraction stage of coix seed oil

Abstract

This invention discloses a multi-parameter linkage control method for the extraction stage of Coix seed oil, belonging to the field of Coix seed oil extraction technology. Specifically, it includes: collecting the cumulative flow rate of solvent entering and leaving the extraction tank and calculating the real-time solvent holding amount; collecting the overall weight of the extraction tank, subtracting the empty weight and solvent holding amount to obtain the real-time dry basis weight; collecting the absorbance of the extract and multiplying it by the solvent holding amount, inputting it into a dissolution quality prediction model to obtain the predicted value of the dissolved Coix seed oil quality, and then calculating the predicted dissolution rate; when the real-time dry basis weight is greater than the lower limit threshold of dry basis weight and the predicted dissolution rate is less than the lower limit threshold of dissolution rate, calculating the change in solvent holding amount, and discharging or replenishing solvent according to the positive or negative change, until the solvent holding amount recovers and the predicted dissolution rate is recalculated; repeating the above steps until the real-time dry basis weight is less than or equal to the lower limit threshold of dry basis weight, and then discharging the residue. This invention achieves dynamic adjustment of solvent holding amount, avoids starch gelatinization, and improves the extraction rate.

Description

Technical Field

[0001] This invention relates to the field of coix seed oil extraction technology, specifically to a multi-parameter linkage control method for the coix seed oil extraction stage. Background Technology

[0002] Coix seed oil is the main medicinal active ingredient extracted from coix seeds and is widely used in the production of traditional Chinese medicine preparations and health products. Currently, the industrial extraction of coix seed oil mainly employs solvent extraction. This involves placing the coix seed raw material in an extraction tank, adding an organic solvent, and then heating and stirring to dissolve the oil from the raw material cells into the solvent phase. The oil-containing solvent is then separated from the residue, and the solvent is recovered to obtain crude oil. The extraction process involves multiple process parameters, such as solvent dosage, extraction temperature, extraction time, and stirring intensity. These parameters have complex coupling relationships, collectively affecting extraction efficiency and product quality.

[0003] In existing technologies, the control methods for the extraction process of coix seed oil mostly adopt a fixed process parameter mode, that is, the solvent dosage, extraction temperature, and extraction time are preset according to the batch of raw materials, and the extraction process is carried out at a constant temperature and time according to the preset parameters. Some improved methods implement feedback control based on changes in the concentration of the extract in the later stages of extraction. For example, when the concentration of the extract detected online is lower than a certain threshold, the heating temperature is increased or the extraction time is extended. Other methods determine the extraction endpoint by detecting changes in the pressure or stirring power in the extraction tank. When the pressure drops or the stirring power decreases to the set value, the extraction is stopped and the residue is discharged.

[0004] However, since Job's tears raw material contains a large amount of starch and protein in addition to oil, as the extraction process proceeds, the raw material particles gradually expand and break down. The dissolution of starch increases solvent viscosity and mass transfer resistance. When the oil dissolution rate decreases in the later stages of extraction, existing methods often enhance mass transfer by increasing the temperature. However, operators cannot determine whether the decrease in the dissolution rate is due to a reduction in the oil content of the raw material or due to starch gelatinization hindering mass transfer. If it is the latter, increasing the temperature will accelerate starch gelatinization, leading to material clumping in the extraction tank, blockage of the slag discharge valve, and in severe cases, even scrapping the entire batch of raw material. Secondly, the solvent holding amount is not a constant value during extraction. As the raw material absorbs water and expands, and the solvent evaporates, the actual amount of solvent participating in mass transfer in the tank is dynamically changing. Existing methods set the solvent usage as a fixed parameter, failing to adjust the solvent holding amount according to the real-time dry weight and dissolution state of the raw material, resulting in the mass transfer driving force not being maintained at an optimal level. When the solvent holding amount is too high, the extract is diluted, increasing the energy consumption for subsequent concentration; when the solvent holding amount is too low, the mass transfer driving force is insufficient, prolonging the extraction time. These two problems are coupled, making it difficult for existing control methods to further improve the extraction rate while avoiding gelatinization. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-parameter linkage control method for the extraction stage of coix seed oil, and to solve the following technical problems: Existing methods cannot distinguish whether the decrease in dissolution rate is due to a reduction in oil content or starch gelatinization. Furthermore, by setting the solvent amount as a fixed parameter, they fail to adjust the solvent holding amount based on real-time dry weight and dissolution status, thus making it impossible to improve the extraction rate while avoiding gelatinization.

[0006] The objective of this invention can be achieved through the following technical solutions: A multi-parameter linkage control method for the extraction stage of coix seed oil includes the following steps: S1. Continuously collect the cumulative flow of solvent entering the extraction tank and the cumulative flow of solvent leaving the extraction tank, and subtract the cumulative flow from the cumulative flow to obtain the real-time solvent holding amount in the extraction tank; S2. Obtain the total weight of the extraction tank, subtract the empty weight of the extraction tank from the total weight to obtain the total weight of the material, and subtract the real-time solvent holding amount from the total weight of the material to obtain the real-time dry weight of the coix seed raw material. S3. Collect the absorbance of the extract at the characteristic absorption wavelength, and input the product of the absorbance and the real-time solvent holding amount into the dissolution quality prediction model to obtain the predicted value of the dissolved coix seed oil quality at the current moment. S4. The difference between the predicted mass of dissolved coix seed oil at the current time and the previous time is taken as the predicted increment of dissolution mass. The predicted dissolution mass increment is divided by the sampling period duration to obtain the predicted dissolution rate. S5. When the real-time dry basis weight is greater than the lower limit threshold of the dry basis weight and the predicted dissolution rate is less than the lower limit threshold of the dissolution rate, calculate the difference between the current real-time solvent holding amount and the real-time solvent holding amount at the previous moment as the change in solvent holding amount. S6. If the change in solvent holding is positive, the solvent in the extraction tank will be discharged to the external temporary storage tank. If it is negative, the solvent will be replenished from the external temporary storage tank to the extraction tank until the real-time solvent holding is restored to the value of the previous moment, and then the predicted dissolution rate will be recalculated. S7. Repeat S5 and S6 until the real-time dry basis weight is less than or equal to the lower limit threshold of the dry basis weight, stop the solvent holding amount adjustment operation and open the slag discharge valve at the bottom of the extraction tank.

[0007] As a further aspect of the present invention: the specific process for establishing the dissolution quality prediction model in step S3 is as follows: Samples were taken from the raw coix seed to be extracted, and the initial moisture content and breakage index of the samples were measured. The samples were placed in a micro-extraction column, and dissolution tests were conducted at multiple preset temperature gradients. At each temperature gradient, solvent was continuously introduced into the micro-extraction column, and the absorbance of the eluent was continuously collected. The overall weight change of the micro-extraction column was weighed and converted into the real-time dry basis weight of the sample. The product of absorbance and real-time solvent holding amount was used as the model input value, and the mass of dissolved coix seed oil in the micro-extraction column was used as the model output value. The initial moisture content, breakage index, and temperature value were used as model covariates. The particle swarm optimization algorithm was used to optimize the parameters of the multivariate nonlinear regression equation containing covariates. The coefficients of the optimized regression equation were solidified into the controller as internal parameters of the dissolution quality prediction model.

[0008] As a further aspect of the present invention: the specific process for determining the lower limit threshold of dry basis weight in step S5 is as follows: Acoustic emission sensors are installed on the outer wall of the extraction tank. During the extraction process, broadband acoustic emission signals generated by the collision of coix seed raw material with the inner wall of the extraction tank are continuously collected. After wavelet packet decomposition of the broadband acoustic emission signals, the energy values ​​of multiple characteristic frequency bands are extracted. The energy values ​​of multiple characteristic frequency bands are weighted and fused to obtain the comprehensive crushing energy value. The second derivative of the curve of the comprehensive crushing energy value over time is used to obtain the energy change acceleration. When the energy change acceleration changes from negative to positive at the zero-crossing point, the real-time dry basis weight at the zero-crossing point is recorded as the critical weight for cell crushing. The critical weight for cell crushing is multiplied by the hardness coefficient of the current batch of raw material to obtain the lower limit threshold of dry basis weight. The hardness coefficient is determined in advance by crushing test of the raw material.

[0009] As a further aspect of the present invention: the specific calculation process for the lower limit threshold of the dissolution rate in step S5 is as follows: The instantaneous liquid-to-solid ratio is obtained by dividing the real-time solvent holding amount by the real-time dry basis weight. The instantaneous liquid-to-solid ratio is then subjected to an exponentially weighted moving average with the average liquid-to-solid ratio of the previous sampling period to obtain a smoothed liquid-to-solid ratio value. Simultaneously, the stirring motor current value and circulating solvent temperature value in the current extraction tank are acquired. The smoothed liquid-to-solid ratio value, stirring motor current value, and circulating solvent temperature value are input into a pre-established fuzzy inference system. The fuzzy inference system outputs a lower limit threshold for dissolution rate based on a preset membership function and a fuzzy rule base. The fuzzy rule base is established through correlation analysis of dissolution rate and various influencing factors in historical extraction batches.

[0010] As a further aspect of the present invention: in step S6, the specific method for synchronously adjusting the circulation flow rate during the solvent discharge or replenishment process is as follows: When the solvent in the extraction tank is discharged to the external storage tank, a first adjustment factor is calculated based on the ratio of the absolute value of the change in solvent holding to the current real-time dry weight, and the circulation pump frequency is reduced to the value obtained by multiplying the original frequency by the first adjustment factor. When the solvent is replenished from the external storage tank to the extraction tank, a second adjustment factor is calculated based on the ratio of the absolute value of the change in solvent holding to the current real-time solvent holding, and the circulation pump frequency is increased to the value obtained by multiplying the original frequency by the second adjustment factor. After the real-time solvent holding is restored to the value of the previous moment, the circulation pump frequency is restored to the original frequency. The first adjustment factor is less than 1 and decreases as the ratio increases, while the second adjustment factor is greater than 1 and increases as the ratio increases.

[0011] As a further aspect of the present invention: in step S6, the specific handling process when the pressure exceeds the limit during the real-time solvent holding capacity recovery process is as follows: During the process of restoring the real-time solvent holding to the value of the previous moment, the pressure value inside the extraction tank is monitored simultaneously, the rate of change of the pressure value over time is calculated, and the pressure value and the rate of change of pressure are input into the pre-stored pressure control decision table. The corresponding pressure relief valve opening degree and pressure relief duration are read from the decision table, and the vent valve at the top of the extraction tank is opened according to the pressure relief valve opening degree to relieve pressure. After the pressure relief duration ends, the vent valve is closed, and the pressure change rate is recalculated. If the pressure value still exceeds the pressure upper limit threshold, the pressure relief operation is repeated until the pressure value is lower than the pressure upper limit threshold, and then the solvent holding adjustment operation continues.

[0012] As a further aspect of the present invention: the specific process of judging the adjustment effect after recalculating the predicted dissolution rate in step S6 is as follows: The recalculated predicted dissolution rate is compared with the predicted dissolution rate at the time of the last S5 trigger to obtain the dissolution rate recovery value. The dissolution rate recovery value is divided by the absolute value of the change in solvent holding amount to obtain the dissolution rate responsiveness per unit solvent adjustment amount. The dissolution rate responsiveness is compared with the dynamic responsiveness benchmark obtained based on the moving average of historical responsiveness over the previous five sampling periods. When the dissolution rate responsiveness is less than 30% of the dynamic responsiveness benchmark, it is determined that the current solvent holding amount adjustment operation is insufficient. When S5 is executed again in subsequent iterations, the calculation method of the change in solvent holding amount is adjusted from the difference between the current real-time solvent holding amount and the real-time solvent holding amount at the previous moment to the difference between the current real-time solvent holding amount and the real-time solvent holding amount at the moment before that. At the same time, the change in solvent holding amount at the current moment is multiplied by the gain coefficient and used as the actual adjustment amount.

[0013] As a further aspect of the present invention: In step S7, after opening the slag discharge valve at the bottom of the extraction tank, the moisture content spectral characteristics of the discharged residue are continuously collected by a near-infrared spectral sensor, and the dielectric constant of the residual coix seed oil in the residue is continuously collected by a microwave resonance sensor. The moisture content spectral characteristics and dielectric constant are compared with preset discharge qualification thresholds. When the moisture content spectral characteristics or dielectric constant exceeds the discharge qualification threshold, the real-time dry basis weight, predicted dissolution rate, and solvent holding amount change at each trigger of step S5 during the current batch extraction process are extracted as abnormal batch feature data. The abnormal batch feature data are input into the online learning module of the dissolution quality prediction model. The online learning module uses the recursive least squares method to incrementally update the regression equation coefficients of the dissolution quality prediction model.

[0014] The beneficial effects of this invention are: This invention solves the technical problems of existing methods being unable to distinguish the causes of dissolution rate decline and dynamically adjust solvent holding capacity by constructing a real-time monitoring and linkage control system. First, the inflow and outflow flow rates and extraction tank weight are continuously collected to calculate the real-time solvent holding capacity and the real-time dry basis weight of the Job's tears raw material, providing a basis for judging the dissolution status. The product of absorbance and solvent holding capacity is input into the dissolution quality prediction model to obtain the predicted value of the dissolved Job's tears oil quality and derive the predicted dissolution rate. When the real-time dry basis weight is greater than the lower limit threshold and the predicted dissolution rate is less than the lower limit threshold, the change in solvent holding capacity is calculated. Based on its positive or negative sign, the solvent is either discharged or replenished until the solvent holding capacity recovers to the value of the previous moment, at which point the dissolution rate is recalculated.

[0015] The above methods avoid relying solely on heating to address dissolution decline. By adjusting the solvent holding capacity, the mass transfer driving force is altered, preventing starch gelatinization and clumping, and ensuring that mass transfer conditions match the real-time state of the raw materials. Specifically, the lower limit threshold for dry weight is dynamically determined by acoustic emission monitoring of cell breakage, while the lower limit threshold for dissolution rate is calculated using fuzzy inference based on the liquid-to-solid ratio, stirring current, and temperature, improving control precision and raw material adaptability. The circulation pump frequency is simultaneously changed during solvent holding capacity adjustment, with the adjustment coefficient adaptively calculated based on the ratio of solvent holding capacity change to dry weight or solvent holding capacity, enhancing mass transfer regulation. After slag discharge, abnormal data is fed back to the model for online learning by detecting residue quality, continuously optimizing model parameters. This invention transforms the extraction process from fixed-parameter control to multi-parameter linkage control based on real-time dissolution status, significantly improving the extraction rate while avoiding gelatinization. Attached Figure Description

[0016] The invention will now be further described with reference to the accompanying drawings.

[0017] Figure 1 This is a flowchart illustrating the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 As shown, this invention provides a multi-parameter linkage control method for the extraction stage of coix seed oil, comprising the following steps: In the solvent extraction process of Coix seed oil, the extraction tank is equipped with a solvent inlet pipe at the top and a solvent outlet pipe and slag discharge valve at the bottom. A first mass flow meter is installed on the solvent inlet pipe, and a second mass flow meter is installed on the solvent outlet pipe. The two mass flow meters are used to continuously collect the cumulative flow rate of solvent entering and leaving the extraction tank, respectively. The controller reads the instantaneous flow rate values ​​of the first and second mass flow meters every second and integrates the instantaneous flow rate values ​​over time to obtain the cumulative flow rate value.

[0020] S1. Subtract the cumulative outflow value from the cumulative inflow value to obtain the real-time solvent holding amount in the extraction tank. This real-time solvent holding amount represents the mass of solvent participating in mass transfer in the extraction tank at the current moment.

[0021] Three weighing sensors are installed at the bottom of the extraction tank, evenly distributed around its circumference. The output values ​​of the three sensors are added together to obtain the total weight of the extraction tank. Before extraction begins, with no material inside the tank, the total weight collected by the three weighing sensors is stored in the controller as the empty weight of the extraction tank. During extraction, the controller collects the total weight once per second, subtracts the empty weight of the extraction tank from the total weight to obtain the total weight of the material, which includes the total mass of the Job's tears raw material and the solvent inside the tank.

[0022] S2. Subtract the real-time solvent holding amount from the total weight of the material to obtain the real-time dry weight of the Job's tears raw material. This real-time dry weight represents the mass of the undissolved Job's tears solid material in the extraction tank at the current moment.

[0023] A sampling branch is installed on the liquid phase circulation line of the extraction tank. This branch diverts a portion of the extract from the liquid phase circulation line, filters it, and then returns it to the liquid phase circulation line. A near-infrared spectroscopy probe is installed on the sampling branch, and the near-infrared spectroscopy analyzer acquires the transmission spectrum of the extract every 30 seconds. The controller performs baseline correction and first derivative processing on the transmission spectrum, finds the wavelength corresponding to the zero-crossing point in the derivative spectrum, compares this wavelength with the characteristic absorption wavelength of trioleic acid glycerides in coix seed oil, and reads the absorbance value at that wavelength upon successful matching.

[0024] S3. Multiply the absorbance value by the real-time solvent holding amount, input the product into the dissolution quality prediction model pre-stored in the controller, and the dissolution quality prediction model outputs the current time's predicted value of dissolved coix seed oil quality.

[0025] S4. The controller subtracts the predicted mass of dissolved coix seed oil at the current moment from the predicted mass of dissolved coix seed oil at the previous moment to obtain the predicted increment of dissolution mass. The predicted increment of dissolution mass is divided by the sampling period of 30 seconds to obtain the predicted dissolution rate. The predicted dissolution rate represents the mass of coix seed oil dissolved from the raw material into the solvent phase per unit time.

[0026] The controller pre-stores a dry basis weight lower limit threshold and a dissolution rate lower limit threshold. The dry basis weight lower limit threshold is determined based on the degree of cell disruption monitored by acoustic emission; for example, for the current batch of raw materials, the dry basis weight lower limit threshold is set to 120 kg. The dissolution rate lower limit threshold is dynamically calculated based on the liquid-to-solid ratio, the stirring motor current, and the circulating solvent temperature; for example, the dissolution rate lower limit threshold calculated at the current moment is 0.5 kg per minute.

[0027] S5. The controller compares the real-time dry basis weight with the lower limit threshold of dry basis weight and compares the predicted dissolution rate with the lower limit threshold of dissolution rate. When the real-time dry basis weight is greater than 120 kg and the predicted dissolution rate is less than 0.5 kg / min, the controller calculates the difference between the real-time solvent holding amount at the current moment and the real-time solvent holding amount at the previous moment, and uses this difference as the change in solvent holding amount. For example, if the real-time solvent holding amount at the current moment is 800 kg and the real-time solvent holding amount at the previous moment is 790 kg, then the change in solvent holding amount is +10 kg.

[0028] S6. If the change in solvent holding is positive, it indicates that the solvent holding at the current moment has increased compared to the previous moment. The controller opens the connecting valve between the bottom of the extraction tank and the external temporary storage tank to discharge the solvent from the extraction tank to the external temporary storage tank. During the discharge process, the real-time solvent holding is monitored until it recovers to the previous moment's value of 790 kg, at which point the connecting valve is closed. If the change in solvent holding is negative, it indicates that the solvent holding at the current moment has decreased compared to the previous moment. The controller opens the connecting valve between the external temporary storage tank and the top of the extraction tank to replenish solvent from the external temporary storage tank to the extraction tank until the real-time solvent holding recovers to the previous moment's value, at which point the connecting valve is closed. After the real-time solvent holding recovers to the previous moment's value, the controller re-executes S3 and S4, collects absorbance, and calculates the new predicted dissolution rate.

[0029] S7. The controller repeats S5 and S6. That is, each time the real-time dry weight is greater than the lower limit threshold of dry weight and the predicted dissolution rate is less than the lower limit threshold of dissolution rate, the change in solvent holding is calculated and the operation of discharging or replenishing solvent is performed according to the sign of the change, until the real-time dry weight is less than or equal to the lower limit threshold of dry weight of 120 kg. At this time, the controller stops the adjustment operation of solvent holding, closes all valves connected to the external temporary storage tank, opens the slag discharge valve at the bottom of the extraction tank, and discharges the residue and remaining solvent in the extraction tank.

[0030] In a preferred embodiment of the present invention, the specific process of establishing the dissolution quality prediction model in step S3 is as follows: First, a sample of 500 grams was taken from the raw coix seed to be extracted. The sample was divided into two equal portions: one for determining the initial moisture content and the other for determining the breakage index. The initial moisture content was determined using an oven drying method, where the sample was dried to constant weight at 105°C, and the percentage of weight loss was calculated to obtain the initial moisture content; for example, the initial moisture content was measured to be 12.5%. The breakage index was determined using a sieving method, where the sample was sieved through standard sieves with apertures of 2 mm, 1 mm, 0.5 mm, and 0.25 mm. The mass of sample remaining on each sieve was weighed, and the mass percentage of each particle size range was calculated. The mass percentage of particles smaller than 0.5 mm was defined as the breakage index; for example, the breakage index was measured to be 18.3%.

[0031] The sample, after moisture content and breakage index determination, was placed in a micro-extraction column with an inner diameter of 20 mm and a height of 200 mm. The column had a solvent inlet and distribution plate at the bottom and a solvent outlet at the top. The micro-extraction column was mounted on an electronic balance with an accuracy of 0.01 g, used to continuously measure the overall weight change of the micro-extraction column. A fiber optic near-infrared spectral probe was installed at the solvent outlet of the micro-extraction column, connected to a near-infrared spectrometer for continuous acquisition of the transmission spectrum of the effluent.

[0032] Dissolution experiments were conducted at multiple preset temperature gradients: 40°C, 50°C, 60°C, 70°C, and 80°C. At each temperature gradient, the micro-extraction column was placed in a constant-temperature water bath. After the temperature stabilized, solvent (95% ethanol by volume) was continuously bubbled into the micro-extraction column at a flow rate of 2 mL / min. During the dissolution process, the transmission spectrum of the eluent was acquired every 10 seconds, and the readings of the electronic balance were recorded simultaneously. Acquisitions were continued until the absorbance of the eluent stabilized. The dissolution experiment at each temperature gradient lasted for 120 minutes.

[0033] The acquired transmission spectra were processed to extract the absorbance value of coix seed oil at the characteristic absorption wavelength, which is 5800 cm⁻¹. The absorbance value was multiplied by the real-time solvent holding capacity, calculated based on the solvent volume and density within the micro-extraction column. This product was used as the model input. The real-time dry basis weight of the sample was calculated based on the overall weight change of the micro-extraction column recorded by an electronic balance. Specifically, the current total weight was subtracted from the empty weight of the micro-extraction column and the current solvent holding capacity. The mass of dissolved coix seed oil within the micro-extraction column was used as the model output. This mass was calculated based on the coix seed oil concentration in the effluent and the cumulative effluent solvent volume.

[0034] The collected data were organized into a sample set, with each sample containing model input values, model output values, and corresponding initial moisture content, fragmentation index, and temperature values. Initial moisture content, fragmentation index, and temperature values ​​were used as model covariates to construct a multiple nonlinear regression equation containing covariates. The form of the regression equation is: output value equals input value multiplied by the first regression coefficient plus the square of the input value multiplied by the second regression coefficient plus the input value multiplied by the temperature multiplied by the third regression coefficient plus the input value multiplied by the initial moisture content multiplied by the fourth regression coefficient plus the input value multiplied by the fragmentation index multiplied by the fifth regression coefficient plus the intercept term.

[0035] A particle swarm optimization (PSO) algorithm was used to optimize the parameters of each regression coefficient in the regression equation. The PSO population size was set to 50 particles, the maximum number of iterations to 200, the learning factor to 2, and the inertia weight to 0.8. The root mean square error between the calculated output value and the measured output value was used as the fitness function, and the fitness function was minimized through iterative optimization. After optimization, the determined regression coefficients were embedded into the controller as internal parameters of the dissolution quality prediction model. For example, the optimized first regression coefficient was 0.85, the second regression coefficient was -0.02, the third regression coefficient was 0.03, the fourth regression coefficient was -0.15, the fifth regression coefficient was -0.08, and the intercept term was 0.5.

[0036] In another preferred embodiment of the present invention, the specific process for determining the lower limit threshold of dry basis weight in step S5 is as follows: An acoustic emission sensor is installed on the outer wall of the extraction tank. This wideband sensor has a frequency response range of 20 kHz to 200 kHz. The sensor is fixed to the lower part of the outer wall of the extraction tank using a magnetic base. Vacuum grease is applied between the sensor and the outer wall to enhance signal coupling. The output signal from the acoustic emission sensor is amplified by a preamplifier and then input to a data acquisition card. The sampling frequency of the data acquisition card is set to 500 kHz.

[0037] During the extraction process, the Job's tears raw material continuously collides with the inner wall of the extraction tank due to the stirring motion, generating broadband acoustic emission signals. A data acquisition card continuously acquires these acoustic emission signals, with each acquisition lasting 0.1 seconds, and 10 acquisitions per second. Wavelet packet decomposition is performed on the acquired broadband acoustic emission signals using the db4 wavelet basis function, with a decomposition level of 5. The decomposition yields energy values ​​for 32 frequency bands, each corresponding to a different frequency range. Based on the correlation between raw material particle size and acoustic emission frequency, four characteristic frequency bands with high correlation to the raw material particle size distribution are selected, with frequency ranges of 40 kHz to 50 kHz, 70 kHz to 80 kHz, 110 kHz to 120 kHz, and 150 kHz to 160 kHz.

[0038] The energy value of each characteristic frequency band is extracted, and the energy values ​​of the four characteristic frequency bands are weighted and fused to obtain the comprehensive crushing energy value. The weight of each characteristic frequency band is determined according to the correlation coefficient between the energy of that frequency band and the degree of raw material crushing. For example, the weight of the 40 kHz to 50 kHz band is 0.3, the weight of the 70 kHz to 80 kHz band is 0.4, the weight of the 110 kHz to 120 kHz band is 0.2, and the weight of the 150 kHz to 160 kHz band is 0.1. The energy value of each frequency band is multiplied by its corresponding weight and then summed to obtain the comprehensive crushing energy value.

[0039] The curve of the overall crushing energy value versus time was smoothed with a smoothing window width of 10 seconds. The second derivative of the smoothed curve was then calculated to obtain the energy change acceleration. The energy change acceleration represents the trend of the rate of change of the overall crushing energy value. The moment the energy change acceleration changes from negative to positive was recorded. This moment represents the turning point where the crushing rate of the raw material particles changes from deceleration to acceleration, corresponding to the moment when the degree of cell breakage reaches a critical value.

[0040] Extract the real-time dry basis weight at the zero-crossing point from the controller and use this weight as the critical weight for cell disruption. For example, if the real-time dry basis weight at the zero-crossing point is 150 kg, then the critical weight for cell disruption is 150 kg. Perform a crushing test on the current batch of raw materials. Place the raw material sample in a pressure testing machine and apply pressure at a loading rate of 5 mm / min. Record the pressure value when the raw material breaks down, and divide the pressure value by the sample cross-sectional area to obtain the hardness coefficient. For example, the measured hardness coefficient is 1.2 MPa. Multiply the critical weight for cell disruption by the hardness coefficient to obtain the lower limit threshold of the dry basis weight. For example, 150 kg multiplied by 1.2 equals 180 kg. MPa is a pressure unit and cannot be multiplied by a mass unit. This is an example and needs correction. The actual hardness coefficient is a dimensionless coefficient. For example, if the hardness coefficient determined through comparative testing is 0.8, then the lower limit threshold of the dry basis weight is 150 kg multiplied by 0.8, which equals 120 kg.

[0041] In another preferred embodiment of the present invention, the specific calculation process for the lower limit threshold of dissolution rate in step S5 is as follows: First, calculate the instantaneous liquid-to-solid ratio by dividing the current real-time solvent holdings by the real-time dry weight. For example, if the current real-time solvent holdings are 800 kg and the real-time dry weight is 200 kg, the instantaneous liquid-to-solid ratio is 4.0. Then, perform an exponentially weighted moving average (TLA) on the instantaneous liquid-to-solid ratio and the average liquid-to-solid ratio from the previous sampling period to obtain the smoothed liquid-to-solid ratio. The formula for the exponentially weighted moving average is: the current smoothed value equals the smoothing coefficient multiplied by the current instantaneous value plus a minus the smoothing coefficient multiplied by the previous smoothed value. The smoothing coefficient is set to 0.3. If the average liquid-to-solid ratio from the previous sampling period is 3.9, then the current smoothed liquid-to-solid ratio is 0.3 multiplied by 4.0 plus 0.7 multiplied by 3.9, which equals 3.93.

[0042] Simultaneously, the current value of the stirring motor and the temperature of the circulating solvent in the extraction tank are acquired. The stirring motor current value is acquired through a current transformer installed on the stirring motor's power line, outputting a standard signal of 4 mA to 20 mA. This signal is converted into a digital value by an analog input module before the current value is read. For example, the current stirring motor current is 15 amps. The circulating solvent temperature value is acquired through a platinum resistance temperature sensor installed on the liquid phase circulation pipeline of the extraction tank. The sensor outputs a resistance signal, which is converted into a standard signal by a temperature transmitter before the temperature value is read. For example, the current circulating solvent temperature is 55 degrees Celsius.

[0043] The smoothed liquid-to-solid ratio, the stirring motor current, and the circulating solvent temperature are used as input variables and input into a pre-established fuzzy inference system. The fuzzy inference system employs Mamdani-type fuzzy inference, containing three input variables and one output variable: the lower limit threshold of the dissolution rate. Membership functions are defined for each input variable. The universe of discourse for the smoothed liquid-to-solid ratio is set to 2 to 6, divided into three fuzzy sets: low, medium, and high. The membership function uses a triangular function, with the vertex of the low set at 2, the medium set at 4, and the high set at 6. The universe of discourse for the stirring motor current is set to 5 to 25 amperes, divided into three fuzzy sets: small, medium, and large. The vertex of the small set is at 5 amperes, the medium set at 15 amperes, and the large set at 25 amperes. The universe of discourse for the circulating solvent temperature is set to 30 to 80 degrees Celsius, divided into three fuzzy sets: low, medium, and high. The vertex of the low set is at 30 degrees Celsius, the medium set at 55 degrees Celsius, and the high set at 80 degrees Celsius. The universe of discourse for the lower limit threshold of the output variable dissolution rate is set to 0.2 kg / min to 1.0 kg / min, and divided into five fuzzy sets: very low, low, medium, high, and very high.

[0044] The fuzzy rule base was established through correlation analysis of dissolution rate and various influencing factors in historical extraction batches. Fifty batches of historical extraction data were analyzed to extract the relationship between liquid-to-solid ratio, stirring current, temperature, and dissolution rate during each batch's extraction process. Fuzzy rules were summarized based on the analysis results. For example, if the liquid-to-solid ratio is high, the stirring current is high, and the temperature is high, then the lower limit threshold for dissolution rate is very high; if the liquid-to-solid ratio is medium, the stirring current is medium, and the temperature is medium, then the lower limit threshold for dissolution rate is medium; if the liquid-to-solid ratio is low, the stirring current is low, and the temperature is low, then the lower limit threshold for dissolution rate is very low. A total of 27 fuzzy rules were established to cover all input combinations.

[0045] The three input variable values ​​at the current moment are substituted into the fuzzy inference system for fuzzification, and the membership degree of each input value to each fuzzy set is calculated. Fuzzy inference is performed according to fuzzy rules to obtain the output fuzzy set corresponding to each rule. The centroid method is used for defuzzification, converting the output fuzzy set into a precise numerical value. For example, the lower limit threshold for dissolution rate output by the fuzzy inference system is 0.48 kg / min. This value is used as the lower limit threshold for dissolution rate at the current moment for judgment in S5.

[0046] In another preferred embodiment of the present invention, the specific method for simultaneously adjusting the circulation flow rate during the solvent discharge or replenishment process in step S6 is as follows: When the controller performs the solvent holding adjustment operation in step S6, it simultaneously acquires the absolute value of the change in solvent holding, the current real-time dry weight, and the current real-time solvent holding. For example, if the current real-time solvent holding is 800 kg and the previous real-time solvent holding was 790 kg, then the change in solvent holding is +10 kg, and the absolute value of the change in solvent holding is 10 kg. The current real-time dry weight is 200 kg, and the current real-time solvent holding is 800 kg.

[0047] When the solvent holding change is positive, the controller discharges the solvent from the extraction tank to an external storage tank. Simultaneously with opening the valve connecting the bottom of the extraction tank and the external storage tank, the controller calculates a first adjustment factor based on the ratio of the absolute value of the solvent holding change to the current real-time dry weight. The ratio is calculated by dividing the absolute value of the solvent holding change by the current real-time dry weight; for example, 10 kg divided by 200 kg equals 0.05. The first adjustment factor is calculated by decreasing it by 0.02 for every 0.01 increase in the ratio, with a minimum value of 0.5. A ratio of 0.05 corresponds to a first adjustment factor of 0.9. The controller multiplies the current operating frequency of the circulation pump by the first adjustment factor of 0.9 to obtain the adjusted circulation pump frequency. The original frequency of the circulation pump was 50 Hz, and the adjusted frequency is 45 Hz. The controller outputs a 45 Hz frequency command signal to the circulation pump inverter, causing the circulation pump speed to decrease. During the solvent discharge process, the controller continuously monitors the real-time solvent holding amount. When the real-time solvent holding amount recovers to the previous value of 790 kg, the controller closes the connecting valve and restores the circulation pump frequency to the original frequency of 50 Hz.

[0048] When the solvent holding change is negative, the controller replenishes the extraction tank with solvent from the external storage tank. Simultaneously with opening the valve connecting the external storage tank and the top of the extraction tank, the controller calculates a second adjustment coefficient based on the ratio of the absolute value of the solvent holding change to the current real-time solvent holding. For example, if the solvent holding change is -10 kg, the absolute value is 10 kg, and the current real-time solvent holding is 790 kg, the ratio 10 kg divided by 790 kg is approximately 0.0127. The calculation rule for the second adjustment coefficient is set so that for every 0.01 increase in the ratio, the second adjustment coefficient increases by 0.03, and the maximum value of the second adjustment coefficient is limited to 1.5. The second adjustment coefficient corresponding to a ratio of 0.0127 is 1.04. The controller multiplies the original circulation pump frequency of 50 Hz by the second adjustment coefficient 1.04 to obtain an adjusted circulation pump frequency of 52 Hz. The controller outputs a 52 Hz frequency command signal to the circulation pump inverter, increasing the circulation pump speed. During solvent replenishment, the controller continuously monitors the real-time solvent holding level. When the real-time solvent holding level recovers to the previous value of 790 kg, the controller closes the connecting valve and restores the circulation pump frequency to the original frequency of 50 Hz.

[0049] The setting of the first adjustment coefficient decreasing as the ratio increases is based on the following considerations: a larger ratio indicates that the amount of solvent to be discharged is greater relative to the dry weight of the raw material. In this case, reducing the circulation flow rate can reduce the scouring of the raw material bed during solvent discharge and prevent excessive carry-over of fine particles. The setting of the second adjustment coefficient increasing as the ratio increases is based on the following considerations: a larger ratio indicates that the amount of solvent to be replenished is greater relative to the current solvent holding. In this case, increasing the circulation flow rate can accelerate the mixing of the newly replenished solvent with the existing solvent in the tank, thereby rapidly homogenizing the solvent concentration and temperature.

[0050] In another preferred embodiment of the present invention, the specific handling process when the pressure exceeds the limit during the real-time solvent holding capacity recovery process in step S6 is as follows: During the solvent holding adjustment operation in step S6, the controller continuously monitors the pressure inside the extraction tank via a pressure transmitter installed on top of the tank. The pressure transmitter has a range of 0 kPa to 200 kPa and outputs a standard signal of 4 mA to 20 mA. The controller reads the pressure value once per second. Simultaneously, the controller calculates the rate of change of the pressure value over time, obtained by subtracting the pressure value from the previous second's pressure value, in kPa / s. For example, if the current pressure is 85 kPa and the previous second's pressure was 82 kPa, the rate of change is +3 kPa / s.

[0051] The controller pre-stores a pressure control decision table, which is built based on historical operating data and equipment characteristics. Rows in the decision table correspond to different pressure value ranges, and columns correspond to different pressure change rate ranges. The table content includes the pressure relief valve opening and the pressure relief duration. The pressure value ranges are divided into five intervals: less than 50 kPa, 50 kPa to 70 kPa, 70 kPa to 90 kPa, 90 kPa to 110 kPa, and greater than 110 kPa. The pressure change rate ranges are divided into four intervals: less than -5 kPa / s, -5 kPa / s to 0 kPa / s, 0 kPa / s to 5 kPa / s, and greater than 5 kPa / s. For example, a pressure value of 85 kPa belongs to the 70 kPa to 90 kPa range, and a pressure change rate of 3 kPa / s belongs to the 0 kPa / s to 5 kPa / s range. The corresponding pressure relief valve opening read from the decision table is 30%, and the pressure relief duration is 15 seconds.

[0052] When the pressure exceeds the preset upper pressure threshold (set at 100 kPa), the controller pauses the solvent holding adjustment operation, temporarily closing any valves currently being used for solvent discharge or replenishment. The controller then sends a control signal to the vent valve at the top of the extraction tank, based on the 30% opening of the pressure relief valve read from the decision table, causing the vent valve to open to 30% for pressure relief. The controller starts timing; after 15 seconds of pressure relief, it sends a close signal to the vent valve, completely closing it. After pressure relief, the controller recalculates the pressure change rate. For example, if the pressure drops to 78 kPa after pressure relief, the pressure change rate is -2 kPa per second. The controller compares the new pressure value of 78 kPa and the pressure change rate of -2 kPa per second with the upper pressure threshold of 100 kPa. Since 78 kPa is lower than 100 kPa, the pressure over-limit condition is resolved, and the controller resumes the paused solvent holding adjustment operation. If the pressure value still exceeds 100 kPa after depressurization, for example, if the pressure value is 102 kPa, the controller repeats the above depressurization operation, reads the opening degree of the depressurization valve and the depressurization duration from the decision table again according to the current pressure value and pressure change rate, and depressurizes again until the pressure value is lower than 100 kPa, then continues the solvent holding amount adjustment operation.

[0053] In another preferred embodiment of the present invention, the specific process of judging the adjustment effect after recalculating the predicted dissolution rate in step S6 is as follows: After adjusting the solvent holding capacity and restoring the real-time solvent holding capacity to the value of the previous moment in step S6, the controller re-executes S3 and S4, collects absorbance, and calculates the new predicted dissolution rate. The new predicted dissolution rate is denoted as Vnew, and the predicted dissolution rate at the time of the previous trigger in S5 is denoted as Vold. For example, if Vold is 0.4 kg / min and Vnew is 0.55 kg / min, then the restored dissolution rate value is Vnew minus Vold, which equals 0.15 kg / min. The absolute value of the change in solvent holding capacity is 10 kg. Dividing the restored dissolution rate value of 0.15 kg / min by the absolute value of the change in solvent holding capacity (10 kg) yields a dissolution rate response of 0.015 kg / min / kg per unit of solvent adjustment.

[0054] The controller stores historical response values ​​calculated after each S5 trigger and adjustment within the previous five sampling periods. For example, the historical response values ​​for the first five periods were 0.018, 0.022, 0.016, 0.020, and 0.019 kg / min / kg. The controller performs a moving average calculation on these five values ​​to obtain the dynamic response baseline. The moving average calculation method is to add the five values ​​and divide by 5, which equals 0.019 kg / min / kg. The current dissolution rate response value of 0.015 is compared with the dynamic response baseline of 0.019. The ratio 0.015 divided by 0.019 is approximately 0.79. This ratio is greater than 30%, therefore the current solvent holding adjustment operation is considered to be responding normally, and no further adjustments will be triggered.

[0055] If, in another adjustment, the calculated dissolution rate response is 0.005 and the dynamic response baseline is 0.018, the ratio 0.005 divided by 0.018 is approximately 0.28, which is less than 30%. In this case, the controller determines that the current solvent holding adjustment operation is insufficiently responsive. The controller records this determination and modifies the calculation method and actual adjustment amount for the change in solvent holding during subsequent repeated executions of S5. The modified calculation method no longer uses the difference between the current real-time solvent holding and the previous real-time solvent holding, but instead uses the difference between the current real-time solvent holding and the real-time solvent holding at the time before that. For example, if the current real-time solvent holding is 800 kg, the previous time it was 790 kg, and the time before that it was 785 kg, then the difference obtained by the original calculation method is 10 kg, and the difference obtained by the new calculation method is 800 kg minus 785 kg, which equals 15 kg. Simultaneously, the controller multiplies the difference of 15 kg obtained by the new calculation method by the gain coefficient as the actual adjustment amount. With the gain factor set to 1.2, the actual adjustment amount is 15 kg multiplied by 1.2, which equals 18 kg. When the controller subsequently executes S6, it performs solvent discharge or replenishment based on this 18 kg adjustment. The introduction of the gain factor expands the adjustment range to address situations of insufficient response.

[0056] In another preferred embodiment of the present invention, in step S7, after the slag discharge valve is opened, the residue is discharged through the slag discharge pipe. A sampling branch pipe is installed on the slag discharge pipe, which diverts a portion of the residue from the slag discharge pipe. The diverted residue is filtered through a filter screen to remove large particles before entering the detection chamber. A near-infrared spectral sensor and a microwave resonance sensor are installed in the detection chamber. The probe of the near-infrared spectral sensor is in contact with the residue in the detection chamber, and the diffuse reflectance spectrum of the residue is collected every 5 seconds, with a spectral acquisition range of wavenumbers from 4000 cm⁻¹ to 10000 cm⁻¹. The probe of the microwave resonance sensor is also in contact with the residue, emitting a microwave signal at a frequency of 2.45 GHz and receiving the reflected signal. The dielectric constant is calculated based on the frequency shift of the reflected signal.

[0057] The extraction process of moisture content spectral characteristics is as follows: The diffuse reflectance spectrum acquired by the near-infrared spectral sensor is preprocessed by the controller, including noise removal and baseline correction. Absorbance values ​​at characteristic wavelengths related to moisture absorption are extracted from the preprocessed spectrum. These characteristic wavelengths are wavenumbers (WF) 5200 cm⁻¹ and 7000 cm⁻¹. The absorbance values ​​at the two characteristic wavelengths are weighted and combined, with the weights determined based on the absorption coefficients of moisture at the two wavelengths. For example, the weight at WF 5200 cm⁻¹ is 0.6, and the weight at WF 7000 cm⁻¹ is 0.4. The weighted combined value is used as the moisture content spectral characteristic value. A standard curve between the moisture content spectral characteristic value and the actual moisture content is pre-stored in the controller. This standard curve is established by acquiring the spectra of residue samples with different moisture contents and measuring their actual moisture contents. The current moisture content spectral characteristic value is substituted into the standard curve to calculate the actual moisture content of the residue. For example, the calculated actual moisture content is 65%.

[0058] The dielectric constant output by the microwave resonant sensor directly reflects the residual coix seed oil content in the residue. The principle behind the dielectric constant measurement is that the dielectric constant of oil differs from that of water and other components; when the oil content in the residue changes, the overall dielectric constant changes accordingly. A standard curve between the dielectric constant and the residual oil content is pre-stored in the controller. This standard curve is established by preparing residue samples with different oil contents and measuring their dielectric constants. Substituting the currently measured dielectric constant into the standard curve, the residual coix seed oil content in the residue is calculated. For example, the calculated residual oil content is 3.8%.

[0059] The controller has preset emission qualification thresholds: a moisture content qualification threshold of 70% and a residual oil content qualification threshold of 5%. Comparing the calculated actual moisture content of 65% to 70%, the actual moisture content is below 70%, falling within the qualification range. Comparing the calculated residual oil content of 3.8% to 5%, the residual oil content is below 5%, also falling within the qualification range. Therefore, the quality of the residue discharged in this operation is qualified, and no abnormal batch data recording is triggered.

[0060] In another slag discharge process, the calculated actual moisture content was 75%, exceeding the 70% acceptable threshold, or the residual oil content was 6.2%, exceeding the 5% acceptable threshold. When either the moisture content or the residual oil content exceeds the acceptable discharge threshold, the controller determines the current batch as an abnormal batch. The controller retrieves the recorded data from memory each time S5 is triggered during the current batch's extraction process. Each time S5 is triggered, the controller records the real-time dry basis weight, predicted dissolution rate, and solvent holding change. For example, if S5 is triggered eight times for the current batch, eight sets of data are recorded. Each set of data includes the real-time dry basis weight value, predicted dissolution rate value, and absolute value of the solvent holding change at the trigger time. The eight sets of data are then compiled into abnormal batch characteristic data.

[0061] Abnormal batch feature data is input into the online learning module of the dissolution quality prediction model. The online learning module uses recursive least squares (RLS) to incrementally update the coefficients of the regression equation of the dissolution quality prediction model. The RLS calculation process is as follows: the abnormal batch feature data is used as a new sample; the prediction error of the new sample under the current model coefficients is calculated; the coefficient correction is calculated based on the prediction error and the gain matrix; and the correction is added to the current coefficients to obtain the updated coefficients. The gain matrix is ​​calculated based on the covariance matrix and the input vector of the new sample, and the covariance matrix is ​​adjusted synchronously after each update. After the update is completed, the new regression equation coefficients are fixed in the controller, replacing the original model coefficients, for use in subsequent batch dissolution quality predictions. For example, the first regression coefficient before the update is 0.85, and after the update it is adjusted to 0.87; the second regression coefficient before the update is -0.02, and after the update it is adjusted to -0.019.

[0062] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A multi-parameter linkage control method for the extraction stage of coix seed oil, characterized in that, Includes the following steps: S1. Continuously collect the cumulative flow of solvent entering the extraction tank and the cumulative flow of solvent leaving the extraction tank, and subtract the cumulative flow from the cumulative flow to obtain the real-time solvent holding amount in the extraction tank; S2. Obtain the total weight of the extraction tank, subtract the empty weight of the extraction tank from the total weight to obtain the total weight of the material, and subtract the real-time solvent holding amount from the total weight of the material to obtain the real-time dry weight of the coix seed raw material. S3. Collect the absorbance of the extract at the characteristic absorption wavelength, and input the product of the absorbance and the real-time solvent holding amount into the dissolution quality prediction model to obtain the predicted value of the dissolved coix seed oil quality at the current moment. S4. The difference between the predicted mass of dissolved coix seed oil at the current time and the previous time is taken as the predicted increment of dissolution mass. The predicted dissolution mass increment is divided by the sampling period duration to obtain the predicted dissolution rate. S5. When the real-time dry basis weight is greater than the lower limit threshold of the dry basis weight and the predicted dissolution rate is less than the lower limit threshold of the dissolution rate, calculate the difference between the current real-time solvent holding amount and the real-time solvent holding amount at the previous moment as the change in solvent holding amount. S6. If the change in solvent holding is positive, the solvent in the extraction tank will be discharged to the external temporary storage tank. If it is negative, the solvent will be replenished from the external temporary storage tank to the extraction tank until the real-time solvent holding is restored to the value of the previous moment, and then the predicted dissolution rate will be recalculated. S7. Repeat S5 and S6 until the real-time dry basis weight is less than or equal to the lower limit threshold of the dry basis weight, stop the solvent holding amount adjustment operation and open the slag discharge valve at the bottom of the extraction tank.

2. The multi-parameter linkage control method for the extraction stage of coix seed oil according to claim 1, characterized in that, In S3, the specific process of establishing the dissolution quality prediction model is as follows: Samples were taken from the raw coix seed to be extracted, and the initial moisture content and breakage index of the samples were measured. The samples were placed in a micro-extraction column, and dissolution tests were conducted at multiple preset temperature gradients. At each temperature gradient, solvent was continuously introduced into the micro-extraction column, and the absorbance of the eluent was continuously collected. The overall weight change of the micro-extraction column was weighed and converted into the real-time dry basis weight of the sample. The product of absorbance and real-time solvent holding amount was used as the model input value, and the mass of dissolved coix seed oil in the micro-extraction column was used as the model output value. The initial moisture content, breakage index, and temperature value were used as model covariates. The particle swarm optimization algorithm was used to optimize the parameters of the multivariate nonlinear regression equation containing covariates. The coefficients of the optimized regression equation were solidified into the controller as internal parameters of the dissolution quality prediction model.

3. The multi-parameter linkage control method for the extraction stage of coix seed oil according to claim 1, characterized in that, In step S5, the specific process for determining the lower limit threshold of dry basis weight is as follows: Acoustic emission sensors are installed on the outer wall of the extraction tank. During the extraction process, broadband acoustic emission signals generated by the collision of coix seed raw material with the inner wall of the extraction tank are continuously collected. After wavelet packet decomposition of the broadband acoustic emission signals, the energy values ​​of multiple characteristic frequency bands are extracted. The energy values ​​of multiple characteristic frequency bands are weighted and fused to obtain the comprehensive crushing energy value. The second derivative of the curve of the comprehensive crushing energy value over time is used to obtain the energy change acceleration. When the energy change acceleration changes from negative to positive at the zero-crossing point, the real-time dry basis weight at the zero-crossing point is recorded as the critical weight for cell crushing. The critical weight for cell crushing is multiplied by the hardness coefficient of the current batch of raw material to obtain the lower limit threshold of dry basis weight. The hardness coefficient is determined in advance by crushing test of the raw material.

4. The multi-parameter linkage control method for the extraction stage of coix seed oil according to claim 3, characterized in that, In step S5, the specific calculation process for the lower limit threshold of dissolution rate is as follows: The instantaneous liquid-to-solid ratio is obtained by dividing the real-time solvent holding amount by the real-time dry basis weight. The instantaneous liquid-to-solid ratio is then subjected to an exponentially weighted moving average with the average liquid-to-solid ratio of the previous sampling period to obtain a smoothed liquid-to-solid ratio value. Simultaneously, the stirring motor current value and circulating solvent temperature value in the current extraction tank are acquired. The smoothed liquid-to-solid ratio value, stirring motor current value, and circulating solvent temperature value are input into a pre-established fuzzy inference system. The fuzzy inference system outputs a lower limit threshold for dissolution rate based on a preset membership function and a fuzzy rule base. The fuzzy rule base is established through correlation analysis of dissolution rate and various influencing factors in historical extraction batches.

5. The multi-parameter linkage control method for the extraction stage of coix seed oil according to claim 1, characterized in that, In step S6, the specific method for synchronously adjusting the circulation flow rate during solvent discharge or replenishment is as follows: When the solvent in the extraction tank is discharged to the external storage tank, a first adjustment factor is calculated based on the ratio of the absolute value of the change in solvent holding to the current real-time dry weight, and the circulation pump frequency is reduced to the value obtained by multiplying the original frequency by the first adjustment factor. When the solvent is replenished from the external storage tank to the extraction tank, a second adjustment factor is calculated based on the ratio of the absolute value of the change in solvent holding to the current real-time solvent holding, and the circulation pump frequency is increased to the value obtained by multiplying the original frequency by the second adjustment factor. After the real-time solvent holding is restored to the value of the previous moment, the circulation pump frequency is restored to the original frequency. The first adjustment factor is less than 1 and decreases as the ratio increases, while the second adjustment factor is greater than 1 and increases as the ratio increases.

6. The multi-parameter linkage control method for the extraction stage of coix seed oil according to claim 5, characterized in that, In step S6, the specific handling process for pressure exceeding the limit during the real-time solvent holding capacity recovery process is as follows: During the process of restoring the real-time solvent holding to the value of the previous moment, the pressure value inside the extraction tank is monitored simultaneously, the rate of change of the pressure value over time is calculated, and the pressure value and the rate of change of pressure are input into the pre-stored pressure control decision table. The corresponding pressure relief valve opening degree and pressure relief duration are read from the decision table, and the vent valve at the top of the extraction tank is opened according to the pressure relief valve opening degree to relieve pressure. After the pressure relief duration ends, the vent valve is closed, and the pressure change rate is recalculated. If the pressure value still exceeds the pressure upper limit threshold, the pressure relief operation is repeated until the pressure value is lower than the pressure upper limit threshold, and then the solvent holding adjustment operation continues.

7. The multi-parameter linkage control method for the extraction stage of coix seed oil according to claim 6, characterized in that, In step S6, the specific process for judging the adjustment effect after recalculating the predicted dissolution rate is as follows: The recalculated predicted dissolution rate is compared with the predicted dissolution rate at the time of the last S5 trigger to obtain the dissolution rate recovery value. The dissolution rate recovery value is divided by the absolute value of the change in solvent holding amount to obtain the dissolution rate responsiveness per unit solvent adjustment amount. The dissolution rate responsiveness is compared with the dynamic responsiveness benchmark obtained based on the moving average of historical responsiveness over the previous five sampling periods. When the dissolution rate responsiveness is less than 30% of the dynamic responsiveness benchmark, it is determined that the current solvent holding amount adjustment operation is insufficient. When S5 is executed again in subsequent iterations, the calculation method of the change in solvent holding amount is adjusted from the difference between the current real-time solvent holding amount and the real-time solvent holding amount at the previous moment to the difference between the current real-time solvent holding amount and the real-time solvent holding amount at the moment before that. At the same time, the change in solvent holding amount at the current moment is multiplied by the gain coefficient and used as the actual adjustment amount.

8. The multi-parameter linkage control method for the extraction stage of coix seed oil according to claim 1, characterized in that, In step S7, after opening the slag discharge valve at the bottom of the extraction tank, the moisture content spectral characteristics of the discharged residue are continuously collected by a near-infrared spectral sensor, and the dielectric constant of the residual coix seed oil in the residue is continuously collected by a microwave resonance sensor. The moisture content spectral characteristics and dielectric constant are compared with preset discharge qualification thresholds. When the moisture content spectral characteristics or dielectric constant exceeds the discharge qualification threshold, the real-time dry basis weight, predicted dissolution rate, and solvent holding change at each trigger of step S5 during the current batch extraction process are extracted as abnormal batch characteristic data. The abnormal batch characteristic data are input into the online learning module of the dissolution quality prediction model. The online learning module uses the recursive least squares method to incrementally update the regression equation coefficients of the dissolution quality prediction model.

Images