Podophyllotoxin purification production optimization management system based on artificial intelligence

Abstract

The invention discloses an artificial intelligence-based podophyllotoxin purification production optimization management system, and particularly relates to the technical field of podophyllotoxin purification production management. A high-speed camera is combined with a deep learning model to carry out visual sorting on raw materials to remove waste materials in the raw materials, and unqualified raw materials detected by near infrared spectroscopy are subjected to trigger pretreatment; the waste removal accuracy, the pretreatment triggering accuracy and the sorting precision are improved; dynamic counter-current extraction is adopted, so that the extraction efficiency and the purity of podophyllotoxin in a light phase are improved; the gradient slope or flow velocity is adjusted based on the real-time UV absorption peak shape AI, the leading edge and trailing of an elution peak are identified through the AI, only the middle section of the main peak is collected, the low-purity edge part is abandoned, when the main peak of podophyllotoxin is detected, a fraction collector is triggered to be started to collect high-purity eluent, and when impurity peaks are detected, the fraction collector is automatically switched to a waste liquid pipeline, so that the recovery rate of target components is increased; the purity and the yield of the podophyllotoxin are favorably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of podophyllotoxin purification production management, and more specifically, to an artificial intelligence-based podophyllotoxin purification production optimization management system. Background Art

[0002] Podophyllotoxin is a natural lignan product with significant anti-tumor and antiviral activity. It is widely used in the preparation of anticancer and dermatological drugs. However, its complex molecular structure, poor thermal stability, and low content in plant raw materials require precise purification processes for industrial production. Traditional extraction and purification technologies suffer from low efficiency, high cost, and large quality fluctuations, making it difficult to meet the stringent requirements of the modern pharmaceutical industry for high purity and high yield. Further optimization of the podophyllotoxin purification and extraction process is needed.

[0003] An existing podophyllotoxin purification production management system uses a mechanical vibrating screen combined with manual sampling to sort podophyllotoxin raw materials, uses a fixed-wavelength spectrometer, and preset thresholds to determine whether the raw material grade is qualified. Qualified raw materials are dried at low temperature and then processed through a fixed-speed crusher, effectively reducing waste rejection errors. The flow rate and pH range are fixed, and extraction is controlled by PLC timing. After extraction, the extraction is allowed to stand, and the phase interface is manually observed before pumping and separation, avoiding complete reliance on manual adjustment. The repeatability between extraction batches is effectively improved. A three-stage linear gradient is used, and high-purity eluent is collected based on the fixed wavelength absorbance threshold to trigger the collection of high-purity eluent, realize automatic segmented collection of eluent, reduce the frequency of manual intervention, and help improve the purity and yield of podophyllotoxin.

[0004] However, the existing management system still has some problems: it relies on physical property differences when sorting raw materials, making it difficult to identify complex defects such as mold, insufficient sorting accuracy, and insufficient pretreatment triggering accuracy; using a fixed flow rate and pH range, it is difficult to cope with seasonal fluctuations in raw material composition, and the extraction efficiency fluctuates greatly. It relies on static sedimentation separation, and the phase interface identification accuracy is insufficient, resulting in a high loss rate of target components in the crude extract; using a preset linear gradient, it cannot be dynamically adjusted according to the elution curve, and the impurity residue is high. The existence of the above defects makes it difficult to meet the requirements of high purity and high yield for the purification production of podophyllotoxin, and further optimization of the podophyllotoxin purification production management system is still needed. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides an artificial intelligence-based podophyllotoxin purification production optimization management system to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an artificial intelligence-based podophyllotoxin purification production optimization management system, comprising:

[0007] Raw material processing module: Visually sort the raw materials to remove waste materials, perform near-infrared spectroscopy detection after the waste materials are removed, and execute different processing processes based on the detection results;

[0008] Dynamic countercurrent extraction module: After pre-equilibration, countercurrent extraction is started. During the countercurrent extraction, the flow rate and pH are dynamically regulated. After the countercurrent extraction is completed, continuous phase separation is performed to obtain a crude podophyllotoxin extract.

[0009] Intelligent chromatography purification module: After the column is balanced, the crude podophyllotoxin extract is pretreated, loaded, and gradient eluted in sequence. The high-purity podophyllotoxin eluate is identified and collected, and the chromatography column is regenerated.

[0010] AI-driven crystallization module: Evaporates and concentrates the eluent to a preset supersaturation range, monitors the supersaturation of the eluent, and triggers seed addition when the supersaturation exceeds the limit. The crystallization process is programmed to cool and dynamically controlled by AI, and solid-liquid separation is performed when the crystallization endpoint is reached.

[0011] Nanomembrane filtration module: Pre-treats podophyllotoxin crystallization mother liquor. AI predicts the optimal operating parameters of the nanomembrane and performs cross-flow filtration based on the predicted results. The high-purity permeate after filtration is collected and returned to the crystallization process.

[0012] Vacuum drying module: Determines key parameters of the vacuum drying equipment based on the moisture content of the podophyllotoxin crystals and dynamically controls the drying process. When the determination conditions are met, a shutdown command is triggered, switching to cooling mode.

[0013] Purification production data collection module: collects key process data and production quality data of each batch of podophyllotoxin purification production process in the production task and transmits the collection results to the purification production effect evaluation module;

[0014] Purification production effect evaluation module: evaluates the podophyllotoxin purification production effect based on the key process data and production quality data of each batch of podophyllotoxin purification production process collected in the production task, and executes different instructions based on the evaluation results.

[0015] Technical effects and advantages of the present invention:

[0016] 1. A high-speed camera is used to take high-speed photos of the raw materials transported to the shooting area via a conveyor belt. The captured images are denoised, enhanced, and ROI segmented, and then input into a deep learning model for inference to output the defect type and location in the captured images. The image coordinate system is mapped to the mechanical coordinate system, and compressed air is precisely sprayed according to the defect location to remove waste, thereby improving the accuracy of waste removal. A near-infrared spectrometer is used to scan the sorted raw materials at a preset frequency and output the raw material purity value, cellulose signal intensity, and lipid signal intensity. Raw material purity values ​​greater than or equal to the preset standard are marked as qualified, otherwise they are marked as unqualified, triggering preprocessing, thereby improving the accuracy of preprocessing triggering, improving the sorting accuracy, and further improving the purity and yield of podophyllotoxin.

[0017] 2. Dynamic countercurrent extraction is adopted, and the flow direction, flow rate of the raw material slurry and the flow direction and flow rate of the solvent are dynamically controlled when the countercurrent contact is started. The centrifugal force of each stage of the extraction tank is dynamically controlled to accelerate the separation of the light and heavy phases. The light phase containing the target component enters the next stage, and the heavy phase containing the residue is removed. The pH of the mixed liquid at the outlet of the extraction tank is dynamically regulated to improve the extraction efficiency. The light phase and the heavy phase of the extract are separated by the first stage centrifugation, and the suspended particles in the light phase are removed by the second stage centrifugation, thereby improving the purity of podophyllotoxin in the light phase, reducing the impurity content, and being conducive to improving the podophyllotoxin yield.

[0018] 3. The gradient slope or flow rate is adjusted based on the real-time UV absorption peak shape AI. When impurity interference appears at the front of the main peak, the gradient segmentation optimization is triggered. The front and tail of the elution peak are identified by AI, and only the middle section of the main peak is collected, and the low-purity edge part is discarded. When the main peak of podophyllotoxin is detected, the fraction collector is triggered to start collecting high-purity eluent. When the impurity peak is detected, it automatically switches to the waste liquid pipeline, which improves the recovery rate of the target component, improves the separation rate, reduces the impurity content of the target component, and improves the purity of the target component, which is beneficial to improve the yield of podophyllotoxin. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a system structure diagram of the present invention.

[0020] Figure 2 A diagram showing the steps of the method of the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] like Figure 1The present embodiment shown provides an artificial intelligence-based podophyllotoxin purification production optimization management system, including a raw material processing module, a dynamic countercurrent extraction module, an intelligent chromatography purification module, an AI-driven crystallization module, a nanomembrane filtration module, a vacuum drying module, a purification production data acquisition module, a purification production effect evaluation module and a database. The raw material processing module, the dynamic countercurrent extraction module, the intelligent chromatography purification module, and the AI-driven crystallization module are connected in sequence, the AI-driven crystallization module is connected to the nanomembrane filtration module, the vacuum drying module, and the purification production data acquisition module, the nanomembrane filtration module is connected to the AI-driven crystallization module, the material processing module, the dynamic countercurrent extraction module, and the intelligent chromatography purification module are connected to the purification production data acquisition module, the purification production data acquisition module is connected to the purification production effect evaluation module, and all modules in the system are connected to the database.

[0023] The raw material processing module visually sorts the raw materials to remove waste materials, performs near-infrared spectroscopy detection after removing the waste materials, and executes different processing processes based on the detection results.

[0024] Furthermore, the raw material processing module includes an AI visual sorting unit, a near-infrared spectrum detection unit, a preprocessing unit, a low-temperature drying unit and an intelligent crushing unit. The AI ​​visual sorting unit uses a high-speed camera to take high-speed photos of the raw materials transmitted to the shooting area via a conveyor belt, and performs denoising, enhancement and ROI segmentation on the captured images, and then inputs them into the deep learning model for inference to output the defect type and position in the captured images, and maps the image coordinate system to the mechanical coordinate system. According to the defect position, compressed air is accurately sprayed to remove waste materials; the near-infrared spectrum detection unit uses a near-infrared spectrometer to scan the sorted raw materials at a preset frequency and output the raw material purity value, cellulose signal strength and lipid signal strength. The raw material purity If the value is greater than or equal to the preset standard, it is marked as qualified, otherwise it is marked as unqualified and pretreatment is triggered; the pretreatment unit triggers enzymatic hydrolysis assistance when the cellulose signal intensity is greater than the lipid signal intensity, and triggers solvent pre-infiltration when the cellulose signal intensity is less than the lipid signal intensity. After pretreatment, the raw material is subjected to rapid HPLC sampling. If the purity is still lower than the threshold, secondary processing is triggered, otherwise secondary sorting is triggered; the low-temperature drying unit performs temperature PID control in the main control loop, and adjusts the temperature setting value according to the moisture data in the auxiliary control loop, and finally obtains dry raw materials whose moisture content and drying temperature meet the preset standards; the intelligent crushing unit adopts dual-loop PID control combining the particle size control loop and the pressure control loop to obtain homogenized raw material powder.

[0025] It should be specifically explained in this embodiment that the raw material processing module further includes a finished raw material powder output unit, which is used to output the homogenized finished raw material powder to the dynamic countercurrent extraction module.

[0026] What needs to be specifically explained in this embodiment is that the AI ​​visual sorting unit adopts Gaussian filtering denoising, CLAHE algorithm enhancement and ROI segmentation based on HSV color space when denoising, enhancing and segmenting the captured images. The deep learning model adopts YOLOv5 architecture, and the training data set contains 5 types of defects such as mildew, insect damage, and foreign matter. The labeled samples are ≥10,000 and the confidence threshold is ≥90%. There are differences in the applicable scenarios of the enzymatic hydrolysis assisted method and the solvent pre-infiltration method provided by the pretreatment unit. The former is suitable for raw materials with high cellulose or lignin content, and the latter is suitable for raw materials with significant cell membrane lipid barriers. Triggering enzymolysis automatically starts the enzymatic hydrolysis tank stirring, temperature control equipment and injects enzyme solution. Triggering solvent pre-infiltration controls the metering pump to spray ethanol and adjust the immersion tank level. When using an enzymatic hydrolysis-assisted solution, the corresponding enzymatic hydrolysis parameters are: cellulase concentration of 0.5% to 1.0% (w / w, mass fraction), temperature of 50°C ± 2°C, and pH 4.8-5.2. When using a solvent pre-infiltration solution, the corresponding solvent pre-infiltration parameters are: ethanol concentration of 20% to 40% (v / v, volume fraction), material-to-liquid ratio of 1:5-1:8, and soaking at room temperature for 1-2 hours. When the raw material contains both high cellulose and high lipid content (such as old rhizomes), a step-by-step "solvent pre-infiltration → enzymatic hydrolysis" treatment is used: 20% ethanol pre-infiltration (1:6, 30 minutes) to disrupt the lipid layer; then enzymatic hydrolysis of the cellulase (0.8%, 50°C, pH 5.0, 40 minutes) to release intracellular toxins.

[0027] In this embodiment, it should be specifically explained that the specific scheme for the low-temperature drying unit to perform temperature PID control in the main control loop is as follows: the real-time drying temperature PV of the raw material is collected. temp and target drying temperature SP temp , then the PID calculation formula for the heating power output percentage u(t) is: , K p1 , K i1 , K d1 is the temperature error e calculated based on temp and temperature error change rate de temp / dt PID parameters are updated in real time after fuzzy inference and defuzzification, e temp =SP temp -PV temp The specific scheme for the low-temperature drying unit to adjust the temperature setting value according to the moisture data in the auxiliary control loop is as follows: Collect the real-time moisture content PV of the raw material moisture and target moisture content SP moisture When the moisture drop rate is greater than the preset value (such as 0.1% / min), maintain the current target drying temperature. When the moisture drop rate is insufficient, increase the SP proportionally. temp , the specific calculation formula is: An example of the linkage control process of the main control loop and the auxiliary control loop of a low-temperature drying unit is given: the heating power is started at 50% in the startup phase to prevent temperature overshoot. temp When the temperature reaches 40℃, it switches to PID automatic control; during the operation phase, PV is collected every 10 seconds. temp and PV moisture , dynamically adjust SP according to moisture data temp , and update the PID output; terminate the stage PV moisture When the temperature is ≤5% and lasts for 5 minutes, stop heating and start the cooling fan; temp When the temperature is ≥55℃, the heating will be shut off immediately, an alarm will be given and the fault will be recorded.

[0028] In this embodiment, it should be specifically explained that the specific scheme for the intelligent crushing unit to perform particle size PID control in the particle size control loop is as follows: the real-time particle size PV of the raw material is collected. size and target granularity SP size , then the target speed output value SP spee The PID calculation formula for d is: ,K p2 , K i2 , K d2 is the particle size error e calculated based on size and particle size error change rate de size / dt PID parameters are updated in real time after fuzzy inference and defuzzification, e size =SP size -PV size The specific scheme of the intelligent crushing unit to perform particle size PID control in the pressure control loop is as follows: Collect the real-time pressure PV of the air flow pressure and target airflow pressure SP pressure , then the proportional valve opening output percentage Valve openin The PID calculation formula for g is: ,K p3 , K i3 , K d3 is the pressure error e calculated based on pressure and particle size error change rate de pressure / dt PID parameters are updated in real time after fuzzy inference and defuzzification, e pressure =SP pressure -PV pressure When SP speed Predict airflow requirements and adjust SP in advance when changes occur pressure , the specific formula is: , ΔSPs peedThe target speed change value is given as an example of a double-loop PID control process: in the initialization phase, the crusher is started to the basic speed (1000rpm) and the air flow pressure is 0.3MPa; in the operation phase, the particle size data is updated every 5 seconds and the SP is adjusted. speed , adjust the proportional valve opening every 1 second to maintain SP pressure ; The particle size was tested three times in a row at the end stage. D50 was between 145-155μ m If the particle size exceeds the standard for 10 minutes, the machine will alarm and stop to check the hardness of the raw materials.

[0029] What needs to be specifically explained in this embodiment is that the low-temperature drying control realizes an efficient and energy-saving drying process through the temperature main PID loop and moisture feedforward compensation, dynamically adjusts the temperature setting value, ensures that the moisture meets the standard and avoids thermal damage; the intelligent crushing control dual-loop PID decoupling design controls the particle size and airflow pressure respectively, combines fuzzy logic adaptive parameters, and responds to changes in raw material properties. These two links can improve system stability through multivariable collaborative control; parameter optimization based on real-time data drive can reduce manual intervention. Through the above scheme, the precise control of the raw material pre-treatment link in the podophyllotoxin production process can be ensured to meet the high standards of pharmaceutical production.

[0030] The dynamic countercurrent extraction module starts countercurrent extraction after pre-equilibration treatment, dynamically regulates the flow rate and pH during the countercurrent extraction, and performs continuous phase separation after the countercurrent extraction to obtain a crude podophyllotoxin extract.

[0031] Furthermore, the dynamic countercurrent extraction module includes a raw material powder receiving unit, a pre-equilibrium processing unit, a countercurrent extraction operation control unit, a continuous phase separation unit, a solvent recovery unit and a podophyllotoxin crude extract output unit. The raw material powder receiving unit is used to receive the finished raw material powder output by the raw material processing module; the pre-equilibrium processing unit preheats the solvent after preparing the solvent, mixes the received finished raw material powder with the preheated solvent according to a preset solid-liquid ratio and stirs it into a uniform slurry, and then rinses the extraction unit pipeline with pure ethanol to remove air and wet the filler; the countercurrent extraction operation control unit includes countercurrent contact start-up control, multi-stage gradient extraction control and pH dynamic regulation, countercurrent contact start-up control The system is used to control the flow direction, flow rate of the raw material slurry and the flow direction and flow rate of the solvent when the countercurrent contact is started. The multi-stage gradient extraction control is used to control the centrifugal force of each stage of the extraction tank to accelerate the separation of the light and heavy phases. The light phase containing the target component enters the next stage, and the heavy phase containing the residue is discharged. The pH dynamic regulation is used to regulate the pH of the mixed liquid at the outlet of the extraction tank; the continuous phase separation unit separates the light phase and the heavy phase of the extract through the first stage centrifugation, and removes the suspended particles in the light phase through the second stage centrifugation; the solvent recovery unit transfers the separated heavy phase to the molecular sieve distillation tower to recover ethanol, and the waste residue is dried and incinerated; the crude extract output unit is used to transfer the final podophyllotoxin crude extract to the intelligent chromatography purification module.

[0032] What needs to be specifically explained in this embodiment is that the solvent prepared in the pre-equilibrium treatment unit is an ethanol-water solvent. Through the solubility curve experiment, it is determined that podophyllotoxin has the highest solubility in 70% ethanol (v / v), and this concentration can effectively reduce the dissolution of polar impurities such as lignin and polysaccharides. Therefore, 70% ethanol (v / v) is mixed with deionized water and 0.1 mol / L phosphate buffer is added to obtain the target solvent. In the process of mixing the finished raw material powder and the preheated solvent according to the preset solid-liquid ratio in the pre-equilibrium treatment unit, the solid-liquid ratio is locked at 1:8 (w / v) based on the pilot test data. The dynamic control of pH is specifically to set a pH probe at the outlet of each stage, and automatically inject a buffer solution for correction when the pH deviates from the preset range.

[0033] The intelligent chromatography purification module controls the column equilibrium and sequentially performs pretreatment, sample loading and gradient elution on the podophyllotoxin crude extract, identifies and collects the high-purity podophyllotoxin eluate, and then regenerates the chromatography column.

[0034] Furthermore, the intelligent chromatography purification module includes a crude extract receiving unit, a chromatography column balance control unit, a crude extract pretreatment unit, a crude extract loading control unit, a gradient elution unit and an eluent collection unit. The crude extract receiving unit is used to receive the podophyllotoxin crude extract transmitted by the dynamic countercurrent extraction module; the chromatography column balance control unit flushes the chromatography column with an initial methanol-water solvent until the UV baseline is stable; the crude extract pretreatment unit is used to defatting and pH adjusting the crude extract; the crude extract loading control unit pumps the pretreated crude extract into the chromatographic column at a preset flow rate, and AI monitors the UV absorption curve in real time, and stops loading when the column load reaches saturation; the gradient elution unit adjusts the gradient slope or flow rate based on the real-time UV absorption peak shape AI, and triggers gradient segmentation optimization when impurity interference appears at the front of the main peak; the eluent collection unit uses AI to identify the front and tail of the elution peak, only collects the middle section of the main peak, and discards the low-purity edge part. When the main peak of podophyllotoxin is detected, the fraction collector is triggered to start collecting high-purity eluent, and automatically switches to the waste liquid pipeline when an impurity peak is detected.

[0035] What needs to be specifically explained in this embodiment is that the intelligent chromatography purification module also includes a chromatography column regeneration unit, a waste liquid treatment unit and an eluent output unit. The chromatography column regeneration unit uses a strong elution solvent to flush the chromatography column to remove residual impurities. If the column efficiency verification is passed, the chromatography column is regenerated successfully, otherwise the chromatography column continues to be flushed; the waste liquid treatment unit transfers the impurity-containing eluent into the activated carbon adsorption tank, and discharges it after meeting the standards, and the waste methanol is recovered and recycled by distillation; the eluent output unit is used to transmit the high-purity eluent to the AI-driven crystallization module.

[0036] The AI-driven crystallization module evaporates and concentrates the eluent to a preset supersaturation range, monitors the supersaturation of the eluent, and triggers intelligent seed addition when the supersaturation exceeds the limit. The crystallization process executes programmed cooling and AI dynamic regulation, and solid-liquid separation is performed when the crystallization endpoint is reached.

[0037] Furthermore, the AI-driven crystallization module includes an eluent receiving unit, an evaporation and concentration unit, a supersaturation monitoring unit, a crystal seed intelligent addition unit, a program cooling control unit, a crystal growth optimization unit, a crystal endpoint determination unit, a solid-liquid separation unit, an abnormality handling unit, and a crystal transmission unit. The eluent receiving unit is used to receive the high-purity eluent transmitted by the intelligent chromatography purification module; the evaporation and concentration unit is used to evaporate and concentrate the high-purity eluent; the supersaturation monitoring unit predicts the critical supersaturation threshold based on the Apelblat equation and the real-time solubility curve, and then monitors the supersaturation of the solution in real time through the Raman spectrometer; the crystal seed intelligent addition unit automatically adds crystal seeds when the Raman spectrum detects that the supersaturation is close to the critical supersaturation threshold; the program cooling control unit automatically adds crystal seeds according to the FBRM The cooling rate is adjusted dynamically based on the particle size distribution; the crystal growth optimization unit calculates the crystal growth rate based on real-time CLD data. If the crystal growth rate is less than the preset rate, the stirring rate is reduced. If the CLD shows a bimodal distribution, PVP is added and the stirring rate is increased; the crystallization endpoint determination unit determines the supersaturation threshold of the crystallization endpoint based on the thermodynamic equilibrium method. When the supersaturation in the solution drops below the supersaturation threshold of the crystallization endpoint and the FBRM shows that the particle size distribution is stable, the crystallization is determined to be complete; the solid-liquid separation unit separates the crystals and the mother liquor through a centrifuge; the abnormality handling unit triggers heating and suspends cooling when explosive nucleation is detected, rebalances the supersaturation, and increases the stirring rate or adds a surfactant when crystal agglomeration is detected; the crystal transfer unit transfers the obtained podophyllotoxin crystals to the vacuum drying module.

[0038] The nano-membrane filtration module pre-treats the podophyllotoxin crystallization mother liquor, AI predicts the optimal operating parameters of the nano-membrane and performs cross-flow filtration based on the predicted results, and collects the filtered high-purity permeate and returns it to the crystallization process.

[0039] The vacuum drying module determines key parameters of the vacuum drying equipment based on the moisture content of the podophyllotoxin crystals and dynamically controls the drying process. When the determination conditions are met, a shutdown instruction is triggered and the system switches to a cooling mode.

[0040] Furthermore, the vacuum drying module includes an online detection unit, a key parameter control unit, a loading unit, a dynamic drying control unit, an endpoint determination unit, a discharging control unit and a quality verification unit. The online detection unit is used to detect the moisture content and solvent residue of the podophyllotoxin crystals; the key parameter control unit determines the control values ​​of the drying temperature, vacuum degree and stirring speed based on the moisture content of the podophyllotoxin crystals; the loading unit starts the vacuum drying equipment after evenly spreading the wet crystals on the drying plate and predicts the drying time based on the initial moisture content and temperature setting; the dynamic drying control unit controls the moisture content based on the real-time moisture content detection results of the vacuum drying process, i.e. adjusts the temperature or vacuum degree, and controls the solvent residue based on the real-time solvent residue detection results of the vacuum drying process, i.e. starts nitrogen purge when the standard is exceeded; the endpoint determination unit triggers a shutdown command when the moisture content and solvent residue of the podophyllotoxin crystals meet the standards and switches to the cooling mode; the discharging control unit automatically discharges the dried crystals through a screw conveyor, sieves to remove lumps, and screens out unqualified products to trigger the dissolution process for reprocessing; the quality verification unit is used to verify whether the purity and microbial limits of the podophyllotoxin crystal powder meet the standards.

[0041] Specifically, it should be noted that the water content standard of podophyllotoxin crystals is the pharmacopoeia standard, the solvent residue standard is ICHQ3C, the purity standard of the podophyllotoxin crystal powder is the mode of the purity control standard in the existing disclosed podophyllotoxin purification production, and if there are multiple modes, the one with a higher control standard is used as the purity standard of the podophyllotoxin crystal powder, and the yield standard of the podophyllotoxin crystal powder is the mode of the yield control standard in the existing disclosed podophyllotoxin purification production, and if there are multiple modes, the one with a higher control standard is used as the yield standard of the podophyllotoxin crystal powder.

[0042] The purification production data acquisition module collects key process data and production quality data of each batch of podophyllotoxin purification production process in the production task and transmits the collection results to the purification production effect evaluation module.

[0043] Furthermore, the key process data of the podophyllotoxin purification production process for each batch in the production task collected by the purification production data collection module include the average sorting accuracy, the average crushing particle size, the average pretreatment trigger accuracy, the actual average mass ratio of podophyllotoxin to impurities in the crude extract, the average solvent consumption ratio, the average ethanol recovery rate, the average separation degree, the average cubic crystal ratio of podophyllotoxin crystals, the average median particle size and the average particle size distribution width; the production quality data of each batch of podophyllotoxin purification production collected in the production task include the average yield and average purity of podophyllotoxin.

[0044] Specifically, it should be noted that the yield is the ratio of the mass of podophyllotoxin in the batch of final finished podophyllotoxin crystalline powder to the mass of podophyllotoxin in the batch of raw materials, the average yield is the average value of the cumulative yields of all batches of podophyllotoxin in the production task, the purity is the ratio of the mass of podophyllotoxin in the batch of final finished podophyllotoxin crystalline powder to the mass of the batch of final finished podophyllotoxin crystalline powder, and the average purity is the average value of the cumulative purity of all batches of podophyllotoxin in the production task.

[0045] The purification production effect evaluation module evaluates the podophyllotoxin purification production effect based on the key process data and production quality data of each batch of podophyllotoxin purification production process collected in the production task, and executes different instructions based on the evaluation results.

[0046] Furthermore, the purification production effect evaluation module includes a data receiving unit, a key process optimization coefficient calculation unit, a production quality optimization coefficient calculation unit, a purification production effect judgment unit, an instruction generation unit, a purification production comprehensive effect improvement analysis unit and a data output unit. The data receiving unit is used to receive the key process data and production quality data of each batch of podophyllotoxin purification production process collected in the production task; the key process optimization coefficient calculation unit calculates the key process optimization coefficient based on the key process data; the production quality optimization coefficient calculation unit calculates the production quality optimization coefficient based on the production quality data; the purification production effect judgment unit calculates the key process optimization coefficient and the production quality optimization coefficient when the key process optimization coefficient and the production quality optimization coefficient are both greater than or equal to At 0 o'clock, it is determined that the purification production effect meets expectations, otherwise it is determined that the purification production effect does not meet expectations; the instruction generation unit generates an instruction for calculating the purification production comprehensive effect improvement index when the purification production effect meets expectations, and generates self-inspection and optimization instructions when the purification production effect does not meet expectations; the purification production comprehensive effect improvement analysis unit adds the key process optimization coefficient and the production quality optimization coefficient to obtain the purification production comprehensive effect improvement index; the data output unit transmits the calculated purification production comprehensive effect improvement index to the podophyllotoxin purification production management center when the purification production effect meets expectations, and transmits the self-inspection and optimization instructions to the podophyllotoxin purification production management center when the purification production effect does not meet expectations.

[0047] What needs to be specifically explained in this embodiment is that the key process optimization coefficient calculation unit calculates the raw material processing effect improvement coefficient based on the average sorting accuracy, the average crushing particle size, the average pretreatment trigger accuracy and the corresponding standard value; calculates the dynamic countercurrent extraction compliance coefficient based on the actual average mass ratio of crude extract podophyllotoxin and impurities, the average solvent consumption ratio, the average ethanol recovery rate and the corresponding standard value; calculates the intelligent chromatography purification compliance coefficient based on the average separation degree and the corresponding standard value; calculates the crystal quality compliance coefficient based on the average cubic crystal ratio, the average median particle size and the average particle size distribution width of podophyllotoxin crystals and the corresponding standard value; the raw material processing effect improvement coefficient, the dynamic countercurrent extraction compliance coefficient, the intelligent chromatography purification compliance coefficient and the crystal quality compliance coefficient are added together to obtain the key process optimization coefficient; the production quality optimization coefficient calculation unit calculates the production quality optimization coefficient based on the average yield, average purity and corresponding standard value of podophyllotoxin.

[0048] Specifically, in this embodiment, a method for evaluating the execution effect of a raw material processing module is provided, and the specific steps are as follows:

[0049] A1. Collect the sorting accuracy, crushing particle size qualification rate and pre-treatment trigger accuracy of each batch of raw materials and calculate the average value of each index respectively, which are recorded as A au 、A bu 、A cu ;

[0050] A2. Retrieve the publicly available sorting accuracy control standards, crushing particle size qualification control standards, and pre-treatment trigger accuracy control standards for podophyllotoxin purification production raw material processing as of the start date of the production task, and extract the mode of each indicator control standard. If there are multiple modes, use the one with the higher control standard as the standard value for each indicator, and record them as A in sequence. ag 、A bg 、A cg ;

[0051] A3. Calculate the raw material processing effect improvement coefficient X AR , the specific formula is: , X AR >0, the raw material processing effect is positively improved, XA R =0, the raw material processing effect has no improvement or degradation, X AR <0, the raw material processing effect degenerates in the opposite direction.

[0052] Specifically, it should be noted that a method for evaluating the performance of dynamic countercurrent extraction is provided in this embodiment, and the specific steps are as follows:

[0053] B1. Collect the actual average mass ratio of podophyllotoxin to impurities, average solvent consumption ratio and average ethanol recovery rate of crude extracts from each batch of dynamic countercurrent extraction process, and record them as Bas 、B bs and B cs ;

[0054] B2. Obtain the standard mass ratio of podophyllotoxin to impurities, standard solvent consumption ratio, and standard ethanol recovery rate of the crude extract from the dynamic countercurrent extraction process. Specifically, the mode of the control standard for each indicator extracted from the dynamic countercurrent extraction of podophyllotoxin purification production currently available as of the start date of the production task is used. If there are multiple modes, the one with the higher control standard is used as the standard value for each indicator, and they are recorded as B in sequence. ac 、B bc and B cc ;

[0055] B3. Calculate the dynamic countercurrent extraction compliance coefficient X BR , the specific formula is: , X BR >0 means the dynamic countercurrent extraction effect is positively improved, X BR =0, the dynamic countercurrent extraction effect has no improvement or degradation, X BR <0, the dynamic countercurrent extraction effect degenerates in the opposite direction.

[0056] Specifically, it should be noted that a method for evaluating the performance of intelligent chromatography purification is provided in this embodiment, and the specific steps are as follows:

[0057] C1. Calculate the separation R of intelligent chromatography purification process s , the specific formula is: , T R1 、T R2 W1 and W2 are the time from injection to appearance of the main peak of podophyllotoxin and the nearest impurity peak, respectively;

[0058] C2. Collect the separation R of each batch of intelligent chromatography purification process si Calculate the average separation R e , the specific formula is: , N a The number of podophyllotoxin purification production batches in the current production task;

[0059] C3, retrieve the standard separation R of the intelligent chromatography purification process a Specifically, the mode of the separation degree extracted from the currently disclosed intelligent chromatography purification process for podophyllotoxin purification production as of the start date of the production task. If there are multiple modes, the one with the higher control standard shall be used as the standard value of the separation degree;

[0060] C4. Calculate the intelligent chromatography purification compliance coefficient X RR , the specific formula is: , X RR>0 means the purification effect of intelligent chromatography is positively improved, X RR =0, the intelligent chromatography purification effect has no improvement or degradation, X RR <0, the intelligent chromatography purification effect degenerates in the opposite direction.

[0061] Specifically, in this embodiment, a method for evaluating the execution effect of AI-driven crystallization is provided, and the specific steps are as follows:

[0062] D1. Collect the average cubic crystal ratio of each batch of podophyllotoxin crystals D au , average median particle size D bu And the average particle size distribution width D cu ;

[0063] D2. Get the standard cubic crystal ratio of podophyllotoxin crystals D ag , standard median particle size range [D bg min ,D bg max ] and standard particle size distribution width D cg The standard cubic crystal ratio, standard median particle size interval and particle size distribution width are the modes of the cubic crystal ratio, median particle size interval and particle size distribution width extracted from the existing disclosed podophyllotoxin purification production crystallization process as of the start date of the production task. If there are multiple modes, the one with a higher control standard is used as the standard value of the cubic crystal ratio, median particle size interval and particle size distribution width. The particle size distribution width is the Span value, and the specific calculation formula is Span=(D90-D10) / D50. In the present invention, D cu Instead of expressing the average Span value, D90, D50, and D10 are the particle size values ​​corresponding to when the cumulative distribution of particles reaches 90%, the median particle size, and the particle size values ​​corresponding to when the cumulative distribution of particles reaches 10%, respectively. In the present invention, D bu Instead, it indicates the average D50;

[0064] D3. Calculate the crystal quality compliance coefficient X DR , the specific formula is: , X DR >0 means AI-driven crystallization execution effect is improving positively, X DR =0, the AI-driven crystal execution effect will not be improved or degraded, X DR <0, the AI-driven crystallization execution effect will degenerate in the opposite direction.

[0065] In this embodiment, it should be specifically explained that the key process optimization coefficient Y A The specific calculation formula is: ;Production quality optimization coefficient Y B The specific calculation formula is: , where Fas 、F ac 、F bs 、F bc They are the average yield of podophyllotoxin, standard yield, average purity of podophyllotoxin, and standard purity, respectively. The specific calculation formula for the comprehensive improvement index of purification production effect ZT is: .

[0066] It should be specifically noted in this embodiment that the remaining preset values ​​and set values ​​that are not explained are selected based on actual needs and are not limited to specific values ​​here.

[0067] The database is used to store data information of all modules in the system.

[0068] like Figure 2 This embodiment provides an artificial intelligence-based method for optimizing the management of podophyllotoxin purification production, comprising the following steps:

[0069] S1. Raw material processing: High-speed cameras combined with deep learning models are used to visually sort raw materials to remove waste materials, near-infrared spectroscopy is performed on clean raw materials, qualified raw materials are sequentially subjected to low-temperature drying and intelligent crushing, and pre-processing is triggered for unqualified raw materials;

[0070] S2, dynamic countercurrent extraction: after pre-equilibration treatment, countercurrent extraction is started, the flow rate and pH are dynamically controlled during the countercurrent extraction operation, continuous phase separation and solvent recovery are performed after the countercurrent extraction is completed, and the obtained podophyllotoxin crude extract is transferred to the intelligent chromatography purification process;

[0071] S3, intelligent chromatography purification: filling the chromatography column and controlling the column balance, pre-treating the crude podophyllotoxin extract before sample loading and gradient elution, identifying and collecting the high-purity podophyllotoxin eluate, and then regenerating the chromatography column;

[0072] S4, AI-driven crystallization: The eluent is evaporated and concentrated to a preset supersaturation range. The supersaturation of the eluent is monitored and intelligent seed addition is triggered when the supersaturation exceeds the limit. The crystallization process is programmed to cool down and dynamically controlled by AI. Solid-liquid separation is performed when the crystallization endpoint is reached.

[0073] S5. Nanomembrane Filtration: Pre-treat the podophyllotoxin crystallization mother liquor, configure the nanomembrane equipment and control the parameters. Based on the concentration and flux attenuation rate of the pre-treated mother liquor, AI predicts the optimal operating parameters and performs cross-flow filtration. The high-purity permeate after filtration is collected and returned to the crystallization process, and the retained liquid is harmlessly treated.

[0074] S6. Vacuum drying: Determine the key parameters of the vacuum drying equipment based on the moisture content of the podophyllotoxin crystals, and dynamically control the drying process. When the judgment conditions are met, a shutdown command is triggered and the machine switches to cooling mode.

[0075] S7. Purification production data collection: Collect key process data and production quality data of each batch of podophyllotoxin purification production process in the production task;

[0076] S8. Purification production effect evaluation: Evaluate the purification production effect of podophyllotoxin based on the key process data and production quality data of each batch of podophyllotoxin purification production process collected in the production task, and execute different instructions based on the evaluation results.

[0077] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. Artificial intelligence-based podophyllotoxin purification production optimization management system, characterized by: include: Raw material processing module: Visually sort the raw materials to remove waste materials, perform near-infrared spectroscopy detection after the waste materials are removed, and execute different processing processes based on the detection results; Dynamic countercurrent extraction module: After pre-equilibration, countercurrent extraction is started. During the countercurrent extraction, the flow rate and pH are dynamically regulated. After the countercurrent extraction is completed, continuous phase separation is performed to obtain a crude podophyllotoxin extract. Intelligent chromatography purification module: After the column is balanced, the crude podophyllotoxin extract is pretreated, loaded, and gradient eluted in sequence. The high-purity podophyllotoxin eluate is identified and collected, and the chromatography column is regenerated. AI-driven crystallization module: Evaporates and concentrates the eluent to a preset supersaturation range, monitors the supersaturation of the eluent, and triggers seed addition when the supersaturation exceeds the limit. The crystallization process is programmed to cool and dynamically controlled by AI, and solid-liquid separation is performed when the crystallization endpoint is reached. Nanomembrane filtration module: Pre-treats podophyllotoxin crystallization mother liquor. AI predicts the optimal operating parameters of the nanomembrane and performs cross-flow filtration based on the predicted results. The high-purity permeate after filtration is collected and returned to the crystallization process. Vacuum drying module: Determines key parameters of the vacuum drying equipment based on the moisture content of the podophyllotoxin crystals and dynamically controls the drying process. When the determination conditions are met, a shutdown command is triggered, switching to cooling mode. Purification production data collection module: collects key process data and production quality data of each batch of podophyllotoxin purification production process in the production task and transmits the collection results to the purification production effect evaluation module; Purification production effect evaluation module: evaluates the podophyllotoxin purification production effect based on the key process data and production quality data of each batch of podophyllotoxin purification production process collected in the production task, and executes different instructions based on the evaluation results.

2. The artificial intelligence-based podophyllotoxin purification production optimization management system according to claim 1, characterized in that: The raw material processing module includes an AI visual sorting unit, a near-infrared spectrum detection unit, a preprocessing unit, a low-temperature drying unit, and an intelligent crushing unit. The AI ​​visual sorting unit uses a high-speed camera to take high-speed photos of the raw materials transmitted to the shooting area via a conveyor belt, performs denoising, enhancement, and ROI segmentation on the captured images, and then inputs them into a deep learning model for inference to output the defect type and location in the captured images, maps the image coordinate system to the mechanical coordinate system, and accurately sprays compressed air according to the defect location to remove waste materials; the near-infrared spectrum detection unit uses a near-infrared spectrometer to scan the sorted raw materials at a preset frequency and output the raw material purity value, cellulose signal intensity, and lipid signal intensity. If the raw material purity value is greater than or equal to the preset standard, it is marked as qualified; otherwise, it is marked as unqualified, triggering preprocessing; The pretreatment unit triggers enzymatic hydrolysis assistance when the cellulose signal intensity is greater than the lipid signal intensity, and triggers solvent pre-infiltration when the cellulose signal intensity is less than the lipid signal intensity. After pretreatment, the raw materials are subjected to rapid HPLC sampling inspection. If the purity is still lower than the threshold, secondary processing is triggered, otherwise secondary sorting is triggered; the low-temperature drying unit performs temperature PID control in the main control loop, and adjusts the temperature setting value according to the moisture data in the auxiliary control loop, and finally obtains dry raw materials whose moisture content and drying temperature meet the preset standards; the intelligent crushing unit adopts dual-loop PID control combining the particle size control loop and the pressure control loop to obtain homogenized raw material powder.

3. The artificial intelligence-based podophyllotoxin purification production optimization management system according to claim 1, characterized in that: The dynamic countercurrent extraction module includes a raw material powder receiving unit, a pre-equilibrium processing unit, a countercurrent extraction operation control unit, a continuous phase separation unit, a solvent recovery unit and a podophyllotoxin crude extract output unit. The raw material powder receiving unit is used to receive the finished raw material powder output by the raw material processing module; the pre-equilibrium processing unit preheats the solvent after preparing the solvent, mixes the received finished raw material powder with the preheated solvent according to a preset solid-liquid ratio and stirs it into a uniform slurry, and then rinses the extraction unit pipeline with pure ethanol to remove air and wet the filler; the countercurrent extraction operation control unit includes a countercurrent contact start-up control, a multi-stage gradient extraction control and a pH dynamic regulation, the countercurrent contact start-up control is used to control the raw material slurry flow direction, flow rate and solvent flow direction and flow rate when the countercurrent contact is started, the multi-stage gradient extraction control is used to control the centrifugal force of each stage extraction tank to accelerate the separation of light and heavy phases, the light phase containing the target component enters the next stage, and the heavy phase containing the residue is excluded, and the pH dynamic regulation is used to regulate the pH of the mixed liquid at the outlet of the extraction tank; The continuous phase separation unit separates the light phase and heavy phase of the extract through the first stage centrifugation and removes suspended particles in the light phase through the second stage centrifugation; The solvent recovery unit transfers the separated heavy phase to the molecular sieve distillation tower to recover ethanol, and the waste residue is dried and incinerated; the crude extract output unit is used to transfer the final podophyllotoxin crude extract to the intelligent chromatography purification module.

4. The artificial intelligence-based podophyllotoxin purification production optimization management system according to claim 1, characterized in that: The intelligent chromatography purification module includes a crude extract receiving unit, a chromatography column balance control unit, a crude extract pretreatment unit, a crude extract loading control unit, a gradient elution unit and an eluent collection unit. The crude extract receiving unit is used to receive the podophyllotoxin crude extract transmitted by the dynamic countercurrent extraction module; The chromatography column balance control unit flushes the chromatography column with the initial methanol-water solvent until the UV baseline is stable; the crude extract pretreatment unit is used to defatting and pH adjusting the crude extract; the crude extract loading control unit pumps the pretreated crude extract into the chromatographic column at a preset flow rate, and AI monitors the UV absorption curve in real time, and stops loading when the column load reaches saturation; the gradient elution unit adjusts the gradient slope or flow rate based on the real-time UV absorption peak shape AI, and triggers gradient segmentation optimization when impurity interference appears at the front of the main peak; the eluent collection unit uses AI to identify the front and tail of the elution peak, only collects the middle section of the main peak, and discards the low-purity edge part. When the main peak of podophyllotoxin is detected, the fraction collector is triggered to start collecting high-purity eluent, and automatically switches to the waste liquid pipeline when an impurity peak is detected.

5. The artificial intelligence-based podophyllotoxin purification production optimization management system according to claim 1, characterized in that: The AI-driven crystallization module includes an eluent receiving unit, an evaporation and concentration unit, a supersaturation monitoring unit, an intelligent seed addition unit, a program cooling control unit, a crystal growth optimization unit, a crystallization endpoint determination unit, a solid-liquid separation unit, an abnormality handling unit, and a crystal transmission unit. The eluent receiving unit is used to receive the high-purity eluent transmitted by the intelligent chromatography purification module; The evaporation and concentration unit is used to evaporate and concentrate high-purity eluent; the supersaturation monitoring unit predicts the critical supersaturation threshold based on the Apelblat equation and real-time solubility curve, and then monitors the supersaturation of the solution in real time using a Raman spectrometer; The intelligent seed adding unit automatically adds seed crystals when the Raman spectrum detects that the supersaturation is close to the critical supersaturation threshold; The programmed cooling control unit dynamically adjusts the cooling rate according to the FBRM particle size distribution; The crystal growth optimization unit calculates the crystal growth rate based on real-time CLD data. If the crystal growth rate is less than the preset rate, the stirring rate is reduced. If the CLD shows a bimodal distribution, PVP is added and the stirring rate is increased. The crystallization endpoint determination unit determines the crystallization endpoint supersaturation threshold based on the thermodynamic equilibrium method. When the supersaturation in the solution drops below the crystallization endpoint supersaturation threshold and the FBRM shows a stable particle size distribution, the crystallization is determined to be complete. The solid-liquid separation unit separates the crystals and the mother liquor through a centrifuge. The abnormality handling unit triggers heating and pauses cooling when nucleation explosion is detected to rebalance the supersaturation. When crystal agglomeration is detected, the stirring rate is increased or a surfactant is added. The crystal transfer unit transfers the obtained podophyllotoxin crystals to the vacuum drying module.

6. The artificial intelligence-based podophyllotoxin purification production optimization management system according to claim 1, characterized in that: The vacuum drying module includes an online detection unit, a key parameter control unit, a charging unit, a dynamic drying control unit, an endpoint determination unit, a discharging control unit and a quality verification unit. The online detection unit is used to detect the moisture content and solvent residue of the podophyllotoxin crystals; the key parameter control unit determines the control values ​​of the drying temperature, vacuum degree and stirring speed based on the moisture content of the podophyllotoxin crystals; the charging unit evenly spreads the wet crystals on the drying plate and starts the vacuum drying equipment and predicts the drying time based on the initial moisture content and temperature setting; the dynamic drying control unit controls the moisture content based on the real-time moisture content detection result of the vacuum drying process, i.e. adjusts the temperature or vacuum degree, and controls the solvent residue based on the real-time solvent residue detection result of the vacuum drying process, i.e. starts nitrogen purge when the standard is exceeded; the endpoint determination unit triggers a shutdown command when the moisture content and solvent residue of the podophyllotoxin crystals meet the standards and switches to the cooling mode; the discharging control unit automatically discharges the dried crystals through a screw conveyor, sieves to remove lumps, and screens out unqualified products to trigger the dissolution process for reprocessing; the quality verification unit is used to verify whether the purity and microbial limit of the podophyllotoxin crystal powder meet the standards.

7. The artificial intelligence-based podophyllotoxin purification production optimization management system according to claim 1, characterized in that: The key process data of the podophyllotoxin purification production process for each batch in the production task collected by the purification production data collection module include the average sorting accuracy, the average crushing particle size, the average pretreatment trigger accuracy, the actual average mass ratio of podophyllotoxin to impurities in the crude extract, the average solvent consumption ratio, the average ethanol recovery rate, the average separation degree, the average cubic crystal ratio of podophyllotoxin crystals, the average median particle size and the average particle size distribution width; the production quality data of each batch of podophyllotoxin purification production in the production task collected include the average yield and average purity of podophyllotoxin.

8. The artificial intelligence-based podophyllotoxin purification production optimization management system according to claim 1, characterized in that: The purification production effect evaluation module includes a data receiving unit, a key process optimization coefficient calculation unit, a production quality optimization coefficient calculation unit, a purification production effect determination unit, an instruction generation unit, a purification production comprehensive effect improvement analysis unit and a data output unit. The data receiving unit is used to receive the key process data and production quality data of each batch of podophyllotoxin purification production process collected in the production task; The key process optimization coefficient calculation unit calculates the key process optimization coefficient based on the key process data; the production quality optimization coefficient calculation unit calculates the production quality optimization coefficient based on the production quality data; The purification production effect determination unit determines that the purification production effect meets expectations when the key process optimization coefficient and the production quality optimization coefficient are both greater than or equal to 0, and otherwise determines that the purification production effect does not meet expectations; the instruction generation unit generates an instruction for calculating the purification production comprehensive effect improvement index when the purification production effect meets expectations, and generates self-inspection and optimization instructions when the purification production effect does not meet expectations; the purification production comprehensive effect improvement analysis unit adds the key process optimization coefficient and the production quality optimization coefficient to obtain the purification production comprehensive effect improvement index; the data output unit transmits the calculated purification production comprehensive effect improvement index to the podophyllotoxin purification production management center when the purification production effect meets expectations, and transmits the self-inspection and optimization instructions to the podophyllotoxin purification production management center when the purification production effect does not meet expectations.

Images