Method for preparing high-purity steviol glycosides by enzyme catalysis combined with membrane separation

Abstract

This invention discloses a method for preparing high-purity steviol glycosides using enzyme-catalyzed membrane separation, specifically relating to the field of process control technology for enzyme-membrane co-separation of functional sweeteners. It addresses the problems of existing steviol glycoside enzyme-catalyzed membrane separation processes being highly dependent on manual experience, exhibiting large fluctuations in product composition due to variations in raw materials and membrane conditions, and difficulty in consistently obtaining high-purity, high-order steviol glycoside concentrates. By constructing an enzyme-membrane co-separation operating condition fingerprint for each batch, encompassing raw materials, enzymes, membranes, and the target product grade, and generating target glycoside composition trajectories, membrane operation constraints, and phased operation strategies based on this fingerprint, and combining online status acquisition and real-time deviation judgment for closed-loop linkage control, this method achieves stable convergence to the predetermined high-purity, high-order steviol glycoside product composition even with fluctuations in raw material quality and membrane conditions. This solves the technical problems of large product composition fluctuations and reliance on repeated trial and error in existing enzyme-catalyzed membrane separation processes.

Description

Technical Field

[0001] This invention relates to the field of process control technology for enzyme-membrane coupling of functional sweeteners, specifically a method for preparing high-purity steviol glycosides using enzyme-catalyzed membrane separation. Background Technology

[0002] Current steviol glycoside production generally employs a process route of "raw material extraction - enzyme-catalyzed modification - membrane separation purification". The extraction stage primarily involves water or dilute alcohol extraction, coarse filtration, and clarification; the enzyme catalysis stage increases the proportion of higher-order steviol glycosides by adding donor sugars and specific enzyme preparations; and the membrane separation stage removes impurities and fractionates glycosides through microfiltration, ultrafiltration, and nanofiltration. Current production line control mainly revolves around individual equipment or processes, relying heavily on operators manually setting parameters such as reaction temperature, enzyme dosage, donor sugar dosage, and membrane flux, pressure differential, and cross-flow velocity based on limited online measurements and laboratory test results.

[0003] In this model, on the one hand, the composition and impurities of stevia raw materials vary significantly between batches, enzyme activity changes with batch and storage conditions, and the water flux and retention performance of membrane modules drift with operating time and cleaning conditions. On the other hand, existing technologies typically only record raw material composition, enzyme activity, and membrane performance in a scattered manner, lacking a unified operating condition description for each batch that simultaneously covers "raw material-enzyme-membrane-target product grade." Furthermore, they lack the ability to construct a trajectory of the target glycoside composition as it evolves with the reaction, as well as corresponding phased operating strategies and coordinated adjustment rules. In actual production, when the raw material, enzyme activity, or membrane condition deviates from the design operating conditions, operators often have to adjust the feed and start-up load based on experience. The reaction stage and membrane grading stage operate independently, making it difficult to correct the operating conditions in a timely and coordinated manner.

[0004] Therefore, existing steviol glycoside enzyme-catalyzed membrane separation processes struggle to quantify the comprehensive changes in reaction solution composition and membrane operating status in a timely manner when faced with factors such as batch fluctuations in raw materials, enzyme activity decay, and membrane fouling. Furthermore, they cannot achieve orderly and coordinated control of substrate and donor sugar addition, retentate reflux ratio, and membrane flux based on unified operating condition information. This results in significant batch-to-batch fluctuations in the proportion of high-grade steviol glycosides and the quality of the final product. How to construct traceable enzyme-membrane co-processing operating condition information on the same production line, based on raw material composition, enzyme activity, membrane baseline performance, and target product grade, and use this information to constrain and guide the operation of each stage of enzyme catalysis and membrane fractionation, thereby ensuring the stable preparation of high-purity, high-grade steviol glycoside concentrates, has become a pressing technical problem in this field. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for preparing high-purity steviol glycosides using enzyme-catalyzed membrane separation, thereby solving the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for preparing high-purity steviol glycosides using enzyme-catalyzed membrane separation, comprising: S1. Collect the composition of raw material extract, impurities, enzyme activity and membrane baseline performance to generate an enzyme-membrane coupling condition fingerprint that includes the grades of raw materials, enzymes, membranes and target products. S2. Based on the operating condition fingerprint, set the target glycoside composition and concentration constraints, and determine the safe operating zone for membrane flux, transmembrane pressure difference and crossflow velocity, and construct the target composition trajectory and membrane operation constraints; S3. Based on the target composition trajectory and membrane operation constraints, the enzyme catalysis process is divided into an early stage, a middle stage and a late stage, and a phased operation strategy is formed by giving reaction parameters and membrane parameter control ranges for each stage. S4. Construct an enzyme reaction vessel, an enzyme retention membrane circuit, and a glycoside fractionation membrane circuit system. Perform enzyme catalysis in the system. Adjust the flux of the enzyme retention membrane and the start and stop of the glycoside fractionation membrane according to the operating strategy. Reflux part of the glycoside fractionation membrane retentate back into the enzyme reaction vessel and divide the glycoside fractionation membrane permeate into a discharge stream and a return stream. S5. Collect reaction solution and membrane flux according to the control cycle to update the operating condition fingerprint. When the operating condition fingerprint does not meet the above constraints, first adjust the substrate and donor sugar addition, and then adjust the reflux ratio of the retentate and the membrane flux to complete the linkage correction. S6. When the operating condition fingerprint meets the above constraints, stop the addition of enzyme and donor sugar, keep the glycoside fractionation membrane circuit running, and collect the glycoside fractionation membrane retentate as a high-purity steviol glycoside concentrate.

[0007] Furthermore, S1 includes: Register the batch number, raw material grade, and dry leaf moisture content upon receiving raw materials; At the outlet of the extraction tank, the flow rate, conductivity, pH and temperature of the extractant were collected, and abnormalities were removed and the stability of the extraction conditions was determined. After the extraction conditions are determined to be stable, the composition of the extract, the results of impurity detection, the results of enzyme activity test, and the baseline performance of the membrane are written into the enzyme membrane coupling condition fingerprint record according to the batch number.

[0008] Furthermore, the enzyme membrane coupled operation condition fingerprint record is stored in the shared database of the production control system and the manufacturing execution system using the batch number and operation condition fingerprint version number as indexes. When generating the enzyme membrane coupled operation condition fingerprint, the generation time, operator identity and detection equipment code are recorded. When an enzyme-membrane co-processing fingerprint record is detected for the same batch under the same operating condition fingerprint version number, the first record is retained and duplicate generation is prevented. If an enzyme-membrane co-processing fingerprint is not generated after a preset delay, the batch is marked as having expired operating condition information and enzyme catalysis is prevented from being automatically started according to the operating condition of that batch.

[0009] Furthermore, S2 includes: Based on the target product grade, raw material extract composition, enzyme activity range and membrane baseline performance in the enzyme-membrane coupling condition fingerprint, records matching the batch are extracted from the product specification library and the membrane baseline performance test record library, respectively. The target glycoside composition range, membrane flux operation safety zone, transmembrane pressure difference operation safety zone, and crossflow velocity operation safety zone are set according to the stages. The target glycoside composition range and the corresponding operation safety zone of each stage are combined to generate target composition trajectory configuration record and membrane operation constraint record, and written into the configuration database with batch number and trajectory version number as index.

[0010] Furthermore, S3 includes: The production control system divides the enzyme catalysis process into early, middle and late stages based on the target composition trajectory configuration record and membrane operation constraint record, and determines the identifier of each stage according to the conversion indication quantity. For each stage, the control range of reaction parameters including substrate feed flow rate, donor sugar addition flow rate, enzyme dosage, reaction temperature, pH, and stirring intensity is determined, as well as the control range of membrane parameters including membrane flux, transmembrane pressure difference, and crossflow velocity. A staged operation strategy record with batch number and trajectory version number as keys is generated and stored in the configuration database.

[0011] Furthermore, S4 includes: A system that includes an enzyme reaction vessel, an enzyme retention membrane circuit, and a glycoside fractionation membrane circuit on the same production line; The glycoside fractionation membrane circuit sends the retentate from the glycoside fractionation membrane back to the enzyme reactor through the retentate reflux branch, and divides the permeate from the glycoside fractionation membrane into an outflow and a return flow. The production control system collects signals of enzyme reactor temperature, pH, liquid level, and flow and pressure of each loop according to a phased operation strategy. It generates pump speed adjustment commands and valve opening adjustment commands, which are executed by the field control unit according to batch number and command sequence number. The system also records phased operation logs based on the returned status number.

[0012] Furthermore, S5 includes: The production control system sets a control cycle, and within each control cycle, it collects the temperature, pH, estimated total steviol glycoside mass fraction, estimated proportion of higher-order steviol glycosides, enzyme retention membrane flux, and glycoside fractionation membrane flux of the enzyme reactor. The above representative values, along with the retentate reflux ratio, permeate discharge ratio, permeate remixing ratio, and current stage identifier, form a dynamic operating condition segment. This segment is then aligned with the static operating condition information to generate an enzyme-membrane coupling operating condition fingerprint snapshot. This snapshot is written into the operating condition fingerprint database according to the batch number and timestamp, recording the target composition trajectory version number, membrane operation constraint version number, and phased operation strategy version number.

[0013] Furthermore, the production control system compares the estimated total steviol glycoside mass fraction, the estimated proportion of higher-order steviol glycosides, the enzyme retention membrane flux, the glycoside fractionation membrane flux, and the retentate reflux ratio with the target composition trajectory, membrane operation constraints, and phased operation strategies in the enzyme-membrane co-operation condition fingerprint snapshot according to the control cycle. When the constraints are not met, the feed pump speeds for substrate addition and donor sugar addition are adjusted according to the preset step size. If the constraints are still not met and the feed pump speed reaches the limit range of the process specification, the retentate reflux ratio and membrane flux are adjusted according to the preset sequence. The judgment result, adjustment amount and status number are written into the batch operation log library.

[0014] Furthermore, S6 includes: When the enzyme-membrane co-processing fingerprint updated within a preset number of continuous control cycles meets the endpoint stage constraint, the production control system stops enzyme addition and donor sugar addition and switches the batch status to endpoint enrichment. The glycoside fractionation membrane loop is kept running under the enrichment state at the endpoint, and the flux of the enzyme retention membrane and the flux of the glycoside fractionation membrane are controlled according to the target flux range of the later stage. The reflux ratio of the retentate is controlled according to the ratio range of the endpoint stage. The retentate of the glycoside fractionation membrane is collected through the inlet branch of the product collection tank as a high-purity steviol glycoside concentrate.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. By constructing an enzyme-membrane co-processing condition fingerprint for each batch, including raw materials, enzymes, membranes, and target product grades, and generating target glycoside composition trajectories, membrane operation constraints, and phased operation strategies based on this fingerprint, and combining online status acquisition and real-time deviation judgment for closed-loop linkage control, it is possible to achieve stable convergence to the predetermined high-purity, high-order steviol glycoside product composition even under fluctuations in raw material quality and membrane status. This solves the technical problems of large product composition fluctuations and reliance on repeated trial and error based on manual experience in existing enzyme-catalyzed membrane separation processes.

[0016] 2. By prioritizing the addition of substrate and donor sugars within the control cycle and making minor adjustments, and by adjusting the reflux ratio of the retentate and the flux of each membrane when necessary, and by setting idempotent rules and evidence-keeping mechanisms for the safe operating zone, stage switching conditions and endpoint determination, the system can improve the enzyme recycling efficiency and membrane separation efficiency, reduce membrane fouling and cleaning downtime, and reduce the cost of process trial and error between batches without exceeding the safety boundaries of the equipment and membrane. Attached Figure Description

[0017] Figure 1 This is a schematic flowchart of a method for preparing high-purity steviol glycosides using enzyme-catalyzed membrane separation according to 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] Example: Figure 1 A flowchart illustrating a method for preparing high-purity steviol glycosides using enzyme-catalyzed membrane separation according to the present invention is provided. The method includes: S1. Collect the composition of the raw material extract, impurities, enzyme activity, and membrane baseline performance to generate an enzyme-membrane coupling condition fingerprint that includes the grades of raw materials, enzymes, membranes, and target products. The specific implementation is as follows: Enzyme-membrane co-processing condition fingerprint refers to the condition description record established around the raw material extract, enzyme preparation and membrane module before enzyme-catalyzed membrane separation of the same batch of stevia raw material on an actual production line. It includes key parameters of raw material, enzyme, membrane and target product grade, and can be called up by batch for subsequent linkage control.

[0020] On this production line, the raw material extract refers to the aqueous solution obtained at the outlet of the extraction tank after the dried stevia leaves have undergone crushing, extraction, and coarse filtration processes. Its composition includes the total steviol glycoside mass fraction, the relative content of each major steviol glycoside, the mass fraction of small molecule sugars, and also conductivity, pH, and temperature. Impurities refer to the total amount and distribution of proteins, polysaccharides, suspended solids, ash, and pigments. Turbidity can be described using turbidity units, and turbidity levels are divided into three grades: zero, first, and second. Grade zero corresponds to a turbidity of less than five turbidity units, first grade corresponds to five to thirty turbidity units, and second grade corresponds to a turbidity of more than thirty turbidity units. Enzyme activity refers to the activity of a unit mass of enzyme preparation in a standard steviol glycoside substrate solution under predetermined temperature and pH conditions. The ability to reduce substrate concentration within a given time period can be expressed as the reduction in substrate mass fraction per unit mass per hour. Membrane baseline performance refers to the water flux and retention behavior of the membrane module for steviol glycosides and macromolecular impurities measured under standard clean water and standard simulated feed solutions. The water flux baseline level can be divided into three grades: low, medium, and high. Low corresponds to less than 20 liters per square meter per hour under standard transmembrane pressure difference and crossflow velocity, medium corresponds to 20 to 50 liters, and high corresponds to more than 50 liters. The retention level can be divided into three grades: low, medium, and high. Low corresponds to a mass retention rate of less than 60% for steviol glycosides or impurities, medium corresponds to 60% to 85%, and high corresponds to more than 85%.

[0021] When the production line is started, the batch number, grade and moisture content of the raw materials are registered at the raw material receiving stage. The batch number serves as the index key for all subsequent records. The moisture content of the dried leaves is obtained by on-site sampling, weighing and constant temperature drying. The moisture content can be set to a common range of 5% to 20%. The measurement and batch information record are completed within one to two hours.

[0022] Subsequently, an online flow meter, conductivity meter, pH meter, and thermometer were installed at the extraction tank outlet. The flow rate, conductivity, pH, and temperature of the extract were continuously collected. Instantaneous readings were acquired at a fixed rhythm, e.g., once per minute. The midpoint of multiple readings within a minute was selected as the representative value for that minute, and readings significantly exceeding the upper or lower limits of the instruments were discarded. This method effectively eliminated individual abnormal peaks and smoothed noise. After the liquid level in the extraction tank stabilized, a 20-30 minute observation window was used to calculate the representative values ​​of conductivity, pH, and temperature within that time period. The difference between the maximum and minimum values ​​is compared with the average value of that range to obtain the relative fluctuation range. The tolerance zone can be set to a range of ±5% to ±15% of the average value, preferably ±10%. When the relative fluctuation range of each parameter falls within the tolerance zone in two to three consecutive observation windows, the extraction condition of the batch is judged to be stable and recorded in the control interface as ready to enter the enzyme catalysis step. If the relative fluctuation of any parameter in any observation window exceeds the tolerance zone more than twice, the extraction condition of the batch is marked as unstable, triggering a process inspection prompt, requiring the inspection of extraction temperature, raw material feeding or filtration status.

[0023] The total steviol glycoside mass fraction and the relative content of major steviol glycosides in the extract composition can be determined by sampling samples in batches during the later stages of extraction at online sampling points next to the extraction tank. The samples are then sent to the laboratory for determination using liquid chromatography. The sampling time, extraction tank level, and flow rate are recorded during sample collection. The analysis should be completed within two hours of sample collection and the results should be entered into the system. The mass fraction of small molecule sugars can be determined by using different chromatographic conditions on the same sample. Protein, polysaccharide, and ash can be determined separately using colorimetric, precipitation, and ignition methods. Turbidity can be measured in the laboratory using a turbidimeter, and colorimetry can be measured using a spectrometer. All the above items are recorded in numerical form such as mass fraction, turbidity unit, and absorbance. At least two to three samples should be taken from the same batch for determination. When the difference between the results of repeated determinations of the same index exceeds 10% of the average value of the index, it is considered inconsistent and resampling and determination are required until the difference between repeated determinations is controlled within 10%.

[0024] In this implementation scheme, higher steviol glycosides refer to a collective term for several types of steviol glycoside components whose retention time falls within a preset window during liquid chromatography determination. This window can be listed in the process specification in the form of specific peak numbers or specific component names. For example, components with glycosidic bonds of riboboroside A and its homologues that have a glycosidic bond number of not less than a preset threshold are considered as a set of higher steviol glycosides. The percentage of higher steviol glycosides refers to the percentage of the sum of the mass fractions of each component in this set to the total mass fraction of steviol glycosides. Bitter glycosides refer to a collective term for several types of steviol glycoside components that have been proven to contribute significantly to bitterness in sensory evaluation and physicochemical analysis. This set can be locked in the process specification in the form of a list of chromatographic peak numbers or component names. The percentage of bitter glycosides refers to the percentage of the sum of the mass fractions of each component in this set to the total mass fraction of steviol glycosides. The aforementioned sets of higher-order steviol glycosides and bitter glycosides can be recorded in the specification list maintained by the quality management department in the form of version numbers. When the scope of the set changes due to the inclusion or exclusion of new ingredients, the version number of the specification list is updated and the corresponding version number is referenced in the enzyme membrane coupling condition fingerprint and product specification library to ensure that the calculation of all subsequent "proportion of higher-order steviol glycosides" and "proportion of bitter glycosides" is consistent with the set definition that was in effect at that time.

[0025] Enzyme activity is determined when enzyme preparations are received into the warehouse or during batch changes. A dedicated enzyme activity test tank is set up in the reaction workshop. A small amount of enzyme preparation is dissolved in a buffer solution containing a standard concentration of steviol glycoside substrate. The pH of the buffer solution can be set to a value in the range of 5 to 7, and the temperature can be set to a value in the range of 40 to 70 degrees Celsius. Stirring is maintained, and samples are taken at fixed time intervals, such as every ten minutes, to measure the change in substrate concentration. The effective activity is calculated based on the ratio between the average decrease in substrate mass fraction per unit time and the amount of enzyme used. The enzyme activity, test pH, test temperature, and test end time are recorded. Enzyme activity determination is performed at least once for each enzyme batch. If the conversion rate of the corresponding reaction of the same enzyme batch is found to be more than 10% lower than the historical average in multiple consecutive production processes, the frequency of enzyme activity retesting can be increased.

[0026] Membrane baseline performance is determined after membrane assembly or replacement. Standard clean water is introduced into the membrane feed side and operated under a fixed transmembrane pressure differential and cross-flow velocity until the flux stabilizes. The water flux per square meter per hour during the stable phase is recorded as the water flux baseline. This flux is compared with the aforementioned low, medium, and high-grade classifications to determine the water flux baseline level. Subsequently, a standard simulated feed solution, such as a test solution containing representative steviol glycosides and high molecular weight impurities, is used. The system is operated under the same pressure differential and cross-flow conditions, and liquids from both the permeate and retrieval sides are collected. The concentrations of steviol glycosides and impurities are determined using chromatographic and colorimetric methods. The rejection rate is obtained by comparing the mass fraction on the retrieval side with the mass fraction in the feed. The steviol glycoside rejection level and impurity rejection level are then classified according to the aforementioned rejection level rules. This test is performed at least once after each new membrane batch or each deep cleaning. For each test, the membrane module number, test pressure differential, test temperature, and running time are recorded.

[0027] The above-mentioned online measurement, laboratory determination, enzyme activity test, and membrane baseline performance test results of the extract are aligned by batch number under a unified time reference through the control system of the production site and the laboratory information system. A 30-minute time window is set between the online records and offline records that need to be matched. Within this window, the online representative value with the closest time is selected to correspond with the offline measurement result. For time periods with missing online data, the representative values ​​of adjacent time periods can be used to fill in the gaps without affecting safety. For items with missing offline data, the typical value of the same grade of raw material in the previous batch can be temporarily used as the estimated value, and an identification field is added to the record to indicate that it is an estimated value.

[0028] For the same batch, continuous online data acquisition records throughout the extraction stage are statistically analyzed in rolling statistical windows of half an hour. Within each rolling window, the fluctuation range of conductivity, pH, and temperature is calculated and it is determined whether the tolerance zone conditions are met. When the conditions are met for two to three consecutive windows, the batch is marked as having stable extraction conditions, and a condition status record is generated in the control system. During a trial operation period, the relationship between the condition status records of several batches and the subsequent enzyme catalytic conversion rate and membrane flux performance can be statistically analyzed. When it is found that the proportion of batches with significant deviations due to delays or missing condition fingerprint information exceeds 10%, the acquisition rhythm, tolerance settings, or time window length can be adjusted.

[0029] Based on the above measurement results, when generating the fingerprint of the enzyme-membrane coupling operation, a record is constructed for each batch, including batch number, raw material grade, dry leaf moisture range, total steviol glycoside mass fraction range of extract, relative content vector of major steviol glycosides, mass fraction of small molecule sugars, mass fraction of protein and polysaccharides, turbidity grade, representative conductivity value, conductivity unit, representative pH value, representative temperature value, enzyme preparation name, enzyme effective activity range, microfiltration membrane module model and its water flux baseline grade and retention grade, nanofiltration membrane module model and its water flux baseline and retention grade, and target product grade identifier. A condition fingerprint version number is assigned to this record, which is bound and archived with the parameter set and rule set on which the generation is based. At the same time, the generation time, operator identity, and detection equipment code used are registered in the control system.

[0030] The enzyme-membrane linkage operation condition fingerprint is stored in a database shared by the production control system and the manufacturing execution system. The combination of batch number and version number is used as the key, and the above set of fields is used as the value. The subsequent enzyme catalysis and membrane linkage control link reads the corresponding operation condition fingerprint through industrial Ethernet according to the current batch number and target version number. If the read fails, an error identifier number is returned. The number zero indicates that the database connection failed, the number one indicates that the specified batch does not exist, and the number two indicates that the version number does not match. The control logic selects to suspend the automatic start of the batch or use the operation condition fingerprint of the most recent batch of raw materials of the same grade as the downgrade basis according to the error identifier number. At the same time, the time, batch number, error identifier number and manual intervention information are recorded in the log.

[0031] To ensure idempotency, only one operating condition fingerprint is generated for the same batch under the same rule version. If the same version record for the same batch already exists when the generation command is triggered, the first generated record is retained and not generated again, and the operation interface will indicate that the current operating condition fingerprint already exists. If modification is required, a new record is generated with the new version number, without overwriting the original record.

[0032] To ensure time and resource constraints, when developing process specifications, the maximum allowable delay from the first sampling of the extract to the enzyme-membrane coupling condition fingerprint being invoked by downstream control links can be set to two to eight hours, preferably four hours. When the actual delay exceeds this upper limit, the batch will be automatically marked as having expired condition information, and automatic enzyme catalysis will be prohibited according to the condition of that batch. It will only be allowed to run according to the downgrade strategy after manual confirmation.

[0033] When verifying the effectiveness of this step on-site, more than 20 batches of raw materials can be sampled within a trial operation period of about one month. The correspondence between the extract composition, enzyme activity and membrane baseline performance recorded in the operating condition fingerprint and the subsequent reaction conversion rate and membrane flux performance can be compared. When the number of batches with significant deviations due to information errors or omissions is less than 10%, the rules disclosed in this step are considered to be directly applicable to guide actual production.

[0034] In a preferred embodiment, for a batch of raw materials numbered a certain, the raw material grade is medium, the dry leaf moisture content is between 8% and 12%, the total steviol glycoside content of the extracted solution is between 7% and 9%, the relative content of rebaudioside A is between 55% and 65%, the small molecule sugar content is between 2% and 4%, the protein content is between 0.3% and 0.7%, the turbidity grade is level 1, the conductivity is between 5 and 15 millisiemens per centimeter, the pH is between 6 and 7, the temperature is between 45 and 55 degrees Celsius, the effective activity of the same batch of enzyme preparation is between 1% and 3% decrease in the corresponding substrate mass fraction per gram per hour, the baseline water flux of the microfiltration membrane is medium, the baseline water flux of the nanofiltration membrane is medium, the nanofiltration membrane has a high retention rate for steviol glycosides and a low retention rate for small molecule sugars, and the target product grade is high-purity, high-grade steviol glycosides. The resulting enzyme-membrane coupling condition fingerprint includes the above-mentioned numerical range. In one embodiment, the batch number and version number are used to lock the data, providing a basis for subsequent linkage control. In another embodiment, the composition and impurity information of the extract can be obtained by deploying a near-infrared spectroscopy online analysis device at the outlet of the extraction tank. Near-infrared spectra are collected from samples of no less than ten batches at no less than five time points per batch, and laboratory concentration is measured simultaneously. By constructing a function relating the spectral intensity of each band to the laboratory concentration, the function coefficients are gradually adjusted so that the average absolute deviation between the predicted concentration and the measured concentration of all samples does not exceed 10% of the average measured concentration. When this condition is met, the current set of function coefficients is used as the calibrated relationship. During the stable production stage, the near-infrared device is used to collect spectra at a fixed rhythm, and the total steviol glycoside mass fraction and main impurity indicators are calculated using this relationship. This method reduces the frequency of laboratory analysis and ensures that the concentration prediction error is within an acceptable range. The enzyme membrane coupling working condition fingerprint can still be constructed by referring to the above steps, which is an equivalent replacement of the above-mentioned implementation method in terms of measurement mode.

[0035] S2. Based on the operating condition fingerprint, set the target glycoside composition and concentration constraints, and determine the safe operating zone for membrane flux, transmembrane pressure difference, and crossflow velocity. Construct the target composition trajectory and membrane operating constraints. The specific implementation is as follows: Setting the target glycoside composition and concentration constraints and determining the safe operating range for membrane flux, transmembrane pressure difference, and crossflow velocity based on the enzyme-membrane coupling operating condition fingerprint is a configuration action performed on the same production line, after the enzyme-membrane coupling operating condition fingerprint has been generated and stored in the database, and before the enzyme-catalyzed membrane separation of this batch has been started.

[0036] The target glycoside composition refers to the expected range of total steviol glycoside mass fraction, relative content of each major steviol glycoside, and proportion of higher-order steviol glycosides for the steviol glycoside product planned for this batch. Concentration constraints refer to the limits that the total steviol glycoside mass fraction and solid content on the reaction liquid and membrane feed sides must not exceed the boundaries of equipment capacity and product requirements. Membrane flux refers to the permeate volume per unit membrane area per unit time under specified transmembrane pressure differential and crossflow velocity. Transmembrane pressure differential refers to the pressure difference between the membrane feed side and the permeate side. Crossflow velocity refers to the linear velocity of the feed liquid flowing along the membrane surface. The above three quantities together describe the membrane operating intensity and load. The safe operating zone refers to the allowable combination range of membrane flux, transmembrane pressure differential, and crossflow velocity without causing excessive membrane fouling, equipment overpressure, or excessive damage to the product composition. The target composition trajectory refers to the interval sequence of target glycoside composition changes over time at each stage when the enzyme catalysis process is divided into multiple stages according to the degree of conversion. Membrane operating constraints refer to the allowable range and variation rules of membrane flux and related process parameters at each stage.

[0037] In this implementation, the production control system first reads the target product grade identifier, raw material extract composition, enzyme activity range, and membrane baseline performance grade of the current batch from the enzyme-membrane coupling operation condition fingerprint record. Simultaneously, it reads the product technical standards corresponding to the target product grade from the product specification library. This library, which can be pre-maintained by quality management personnel, records the range of total steviol glycoside mass fraction, the relative content range of major steviol glycosides, and the percentage range of higher-order steviol glycosides for different product grades, as well as the upper limits for allowable bitterness-related steviol glycosides and small molecule sugar content. These ranges are stored in mass fraction or relative percentage form, along with a specification version number and effective date. The control system searches the product specification library for a unique matching record based on the target product grade and specification version number in the enzyme-membrane coupling operation condition fingerprint. If no match is found, a specification missing error number is returned, and a prompt is displayed on the interface, requiring process personnel to supplement or confirm the product specification, preventing the generation of a target composition trajectory when the target is unclear.

[0038] After locating the target record, the control system calculates the expected compositional difference that needs to be traversed from the current raw material state to the target state based on the range of total steviol glycoside mass fraction and the relative content vector of major steviol glycosides in the raw material extract. Combined with the reaction rate capability given by the enzyme activity range, the enzyme catalytic process is divided into at least three stages, such as the initial conversion stage, the intermediate enrichment stage, and the high-purity convergence stage. Each stage corresponds to a conversion degree range and an expected duration range. This division can be achieved by statistically analyzing the conversion time distribution and compositional changes in several recent batches of production. The time points when each batch reaches the median value of the target total steviol glycoside mass fraction and the median value of higher-order steviol glycosides are used as the dividing points, and the median value is taken as the default duration. Then, an allowable deviation of ±10% to 20% is set for subsequent adjustments.

[0039] Based on the phase division, the control system sets a target glycoside composition range for each phase, specifically the lower and upper limits of the total steviol glycoside mass fraction, the lower and upper limits of the target higher-order steviol glycoside proportion, the maximum allowable proportion of bitter glycosides, and the maximum mass fraction of small molecule sugars. By extrapolating the final requirements of the product specification forward to the intermediate targets of each phase, the calculation takes into account the enzyme activity range and the donor sugar addition capacity. If the enzyme activity is near the low value given in the operating condition fingerprint or the initial proportion of higher-order glycosides in the raw materials is at the lower limit of the range, the target range of the intermediate phase is appropriately widened so that the intermediate phase target does not exceed 80% of the final specification, in order to prevent excessive pursuit of higher-order glycosides in the intermediate phase, which may lead to the reaction being difficult to converge. The above phase division and target range form a target composition trajectory configuration record. This record stores the start and end conversion degree, expected time range, and composition range of each phase in phase order, and is accompanied by a trajectory version number. This version number is bound to the current enzyme membrane coupling operating condition fingerprint version and product specification version and is stored in the process configuration library.

[0040] To determine the safe operating range for membrane flux, transmembrane pressure differential, and crossflow velocity, the control system retrieves historical operating records from the membrane baseline performance test record library that match the membrane module model, water flux baseline level, and rejection level of this batch. These historical records include stable flux, flux decay rate, and cleaning cycle under different transmembrane pressure differential and crossflow velocity conditions, as well as the corresponding reaction solution viscosity and solids content. After aligning the historical records by timestamp and removing obviously abnormal flux drop data, the flux and flux decay rate are statistically analyzed. Preferably, the set of points where the flux decay rate is lower than a set threshold can be used as a safe sample. The flux decay rate threshold can be set to a flux decrease rate of no more than 2% to 15% of the initial flux per hour. The highest and lowest flux during stable operation are found in this safe sample, and the upper and lower limits of membrane flux are set accordingly for this batch. The upper limit of membrane flux can be set as the historical safe limit. The high flux threshold is set at 50% to 90%, preferably 60% to 80%. The lower limit of membrane flux can be set at 60% to 90% of the historical safe minimum flux to balance equipment capacity and economy. The upper limit of transmembrane pressure difference can be determined based on the maximum allowable pressure difference provided by the membrane supplier and historical cleaning frequency records, preferably 50% to 90% of the supplier's allowable value, combined with the pressure difference range in historical records where no mechanical damage occurred as the final upper limit. The lower limit of crossflow velocity can be determined based on empirical values ​​to prevent local deposition on the membrane surface and the lower limit of velocity in the flux-stable region in historical records, and can be set at 50% to 80% of the velocity lower limit to the velocity upper limit in the historical flux-stable sample. The upper limit of crossflow velocity, under the premise of preventing pump and pipeline overload, can refer to the maximum continuous operating speed recommended by the pump manufacturer and multiply it by a coefficient of 50% to 90% as the upper limit. In one embodiment, the specific values ​​of the above proportions and thresholds can be given by the plant's technical management department in the process specifications, recorded in the form of rule version numbers, and updated synchronously with changes in the process specifications.

[0041] After obtaining the upper and lower limits of the above-mentioned flux, pressure difference, and velocity, the control system refines the safe operating range of each stage according to the expected range of feed viscosity and solids content in different reaction stages. For example, in the high viscosity and high solids stage, the flux target range is set to 40% to 70% of the global flux allowable range, and the pressure difference upper limit is reduced accordingly. In the low viscosity stage, the flux target range is set to 60% to 90% of the global flux allowable range. In each stage, the crossflow velocity target can be maintained in the middle to high range between the above-mentioned velocity upper and lower limits to ensure sufficient membrane shear force. The above approach is to store a set of triplets for each stage in the process configuration library, namely the flux target range, pressure difference target range, and crossflow velocity target range. These ranges, together with the target composition range of the corresponding stage, constitute the membrane operation constraints.

[0042] The above-mentioned target composition trajectory configuration records and membrane operation constraint records are stored in the configuration database with the current batch number and trajectory version number as keys. The key values ​​include the stage number, stage start and end identifiers, the range of total steviol glycoside mass fraction of each stage, the range of the proportion of higher-order steviol glycosides of the target, the upper limit of bitter glycosides, the upper limit of small molecule sugars, and the upper and lower limits of flux, pressure difference and crossflow velocity.

[0043] To ensure version locking and evidence retention, the database is configured to record the generation time, operator identity, and the version number of the enzyme membrane coupling condition fingerprint and the product specification version number. The version identifier of the currently effective trajectory is also written into the additional field of the enzyme membrane coupling condition fingerprint for that batch. This allows for checking whether the versions are consistent when calling the database in subsequent reaction control steps.

[0044] The production control system and the process configuration system communicate via an industrial network. When calling, the system requests the target composition trajectory and membrane operation constraints according to the batch number. If the configuration database cannot return data within a limited time, a timeout limit of five to thirty seconds can be set, preferably four to ten seconds. After the timeout, the control system can send another request to the configuration database at fixed intervals. The number of retries can be set to two to five, preferably three. If the system still fails after exceeding the number of retries, a configuration read failure event is recorded and automatic startup of the enzyme-catalyzed membrane operation for that batch is prohibited. It can only be started manually under safe parameters after the configuration status is manually confirmed.

[0045] To ensure idempotency, only one target composition trajectory record is allowed to be generated in the same batch under the same rule version. When the configuration system receives the generation instruction, it first searches the configuration database for the record corresponding to the batch and the current rule version. If it already exists, it directly returns the existing record identifier, which is considered as successful generation, and no new record is created. Only when the rule version is updated will a new trajectory record be generated and assigned a new version number, while the old version is retained for traceability.

[0046] To ensure the safety boundaries of this step, the process specifications may stipulate that any batch that is not associated with a valid product specification version or has not passed the membrane operation safety zone calculation is prohibited from entering the automatic operation state, and any manual modification of trajectory parameters must be confirmed by authorized technical personnel on the control interface, and modification records should be automatically generated and archived.

[0047] When verifying the effectiveness of this step on-site, several batches can be selected within a continuous trial operation period. The results can be statistically analyzed to determine whether the membrane flux decay rate remains within the aforementioned flux decay threshold range and whether the proportion of the target higher-order glycosides gradually approaches the final target specification range at each stage. If the number of batches that deviate from the flux decay threshold or the target composition trajectory exceeds the set proportion of the total number of batches, the target composition trajectory can be regenerated by adjusting the intermediate target range in the product specification library or adjusting the proportion of flux and differential pressure safety zones. The set proportion can be set to 5% to 20%, preferably 10%. After adjustment, a new version is formed and verified in subsequent batches.

[0048] In a preferred embodiment, for a batch of high-purity, high-order steviol glycosides as the target product grade, the product specifications can be set as follows: total steviol glycoside mass fraction not less than a certain value, high-order steviol glycoside percentage not less than a certain value, bitter glycoside percentage not more than a certain value, and small molecule sugar mass fraction not more than a certain value. Under this specification, based on the enzyme-membrane coupling operating condition fingerprint, where the total steviol glycoside mass fraction of the raw material is approximately within a certain range, the initial percentage of high-order glycosides is approximately within a certain range, the enzyme activity is within a certain range, and the membrane water flux baseline level is medium, the reaction process is divided into three stages. In the first stage, the target total steviol glycoside mass fraction range is set slightly higher than the raw material level, and the target high-order glycoside percentage range is set slightly higher than the initial level. In the second stage, the above target ranges are moved closer to the endpoint specification. In the third stage, the ranges are tightened to be close to the endpoint specification, while the flux upper limit of the high viscosity stage is set to the historically safe highest level. The median value of flux within a certain percentage range is used to set the upper limit of the pressure difference to the median value within the percentage range allowed by the supplier, and the lower limit of the crossflow velocity to the lower-middle segment of the velocity range in the historical flux stability sample. The flux target is appropriately increased in the low viscosity stage. The above values ​​are stored in the target composition trajectory configuration record in the form of specific numerical ranges. In another embodiment, the target composition trajectory can also be given by dividing the reaction process time axis into several time periods according to the actual total reaction time, and giving the time series of the target total steviol glycoside mass fraction and the proportion of higher-order steviol glycosides by uniformly increasing the time or smoothly changing according to a preset piecewise linear relationship within each time period. That is, a monotonically changing target sequence is constructed by setting the start value and the end value between adjacent time periods, and then the corresponding membrane operation safety zone is matched according to the target time point, thereby forming an equivalent trajectory and constraint configuration. This method is regarded as an equivalent replacement in the trajectory construction method in this invention.

[0049] S3. Based on the target composition trajectory and membrane operation constraints, the enzyme catalysis process is divided into an early stage, a middle stage, and a late stage. Reaction parameters and membrane parameter control ranges are provided for each stage to form a phased operation strategy. The specific implementation is as follows: The enzyme catalysis process is divided into early, middle and late stages based on the target composition trajectory and membrane operation constraints. A phased operation strategy is formed by giving reaction parameters and membrane parameter control ranges for each stage. This process is completed by the production control system in process configuration mode after the target composition trajectory configuration record and membrane operation constraint record have been written into the configuration database in batches and before the enzyme catalysis combined with membrane separation of this batch officially starts continuous operation.

[0050] The early, middle, and late stages refer to three consecutive operating segments in the process of enzyme-catalyzed reaction evolving from the raw material state to the target product state: low conversion degree, medium conversion degree, and near-end conversion degree, respectively. The conversion degree is a dimensionless indicator obtained by comparing the total steviol glycoside mass fraction and the proportion of higher-order steviol glycosides in the reaction solution measured online or near-line with the corresponding interval in the target composition trajectory. The production control system sets online sampling points and online analysis devices at the reactor outlet or circulation pipeline. Through near-infrared spectroscopy and the aforementioned calibration relationship, or through simplified sampling and testing, the estimated values ​​of the total steviol glycoside mass fraction and the proportion of higher-order steviol glycosides in the reaction solution are obtained at fixed time intervals. The time interval can be set to two to fifteen minutes. In each acquisition cycle, multiple instantaneous acquisition values ​​are smoothed by taking the median value and discarding readings that significantly exceed the instrument range to obtain the representative value of that cycle. This representative value is compared with the target interval of the corresponding time point or corresponding stage in the target composition trajectory configuration record of the current batch to obtain the conversion indicator interval of that cycle.

[0051] In this embodiment, the production control system reads the conversion criteria for the start and end of each stage in the target composition trajectory configuration record. For example, the stage boundary is defined by the total steviol glycoside mass fraction and the proportion of higher steviol glycosides reaching a certain intermediate value or higher. The system also reads the corresponding flux target range, pressure difference target range, and crossflow velocity target range for each stage. Simultaneously, it combines the raw material composition, enzyme activity range, and equipment capacity recorded in the enzyme membrane coupling operating condition fingerprint to derive the reaction parameter control range for each stage. The reaction parameter control range includes the substrate feed flow rate range, the donor sugar addition flow rate range, the enzyme initial dosage and replenishment amount range, the reaction temperature range, the pH range, and the stirring intensity range. The substrate feed flow rate and the donor sugar addition flow rate are recorded as volumetric flow rate or mass flow rate, the reaction temperature is recorded as Celsius temperature, the pH is recorded as pH value, and the stirring intensity can be expressed by the stirrer speed or power per unit volume.

[0052] When determining the control range of reaction parameters at each stage, the production control system can refer to the temperature, pH, and donor sugar content distribution of historical batches when they reached similar target composition ranges. By statistically analyzing the distribution range of these historical data in stable batches, this range can be used as the default range. Then, based on the enzyme activity level of this batch and whether the initial composition of the raw materials is biased towards high impurities, high viscosity, or low activity, the range can be widened or tightened. When widening, the upper and lower limits of the range can be expanded by a certain proportion; when tightening, the upper and lower limits of the range can be contracted towards the middle value by a certain proportion. This adjustment ratio can be given in the process specification as a percentage range.

[0053] When generating a phased operation strategy, the production control system links the phase identifier, phase judgment condition, reaction parameter control range, and membrane parameter control range of each phase to form a phased strategy record. The phase judgment condition can be set to the conversion indicator falling into the target composition range corresponding to a certain phase within a certain number of consecutive acquisition cycles. The number of consecutive cycles can be set to two to five times to avoid frequent switching caused by short-term fluctuations. During operation, the production control system decides whether to switch from the early stage to the middle stage or from the middle stage to the late stage based on the current phase identifier and phase judgment condition. When switching, the target values ​​of reaction parameters and membrane parameters are smoothly adjusted from the target range of the previous stage to the target range of the next stage. For example, within a preset buffer time, the target values ​​of substrate feed flow rate, donor sugar addition flow rate, reaction temperature, and membrane flux are pushed to the middle value of the next stage in a fixed step size to prevent sudden changes from causing equipment shock or drastic fluctuations in composition.

[0054] After generating the phased operation strategy record, the production control system stores the record in the phased strategy table of the configuration database using the batch number and trajectory version number as keys. It also writes the generation time, operator identity, and the trajectory version number and membrane operation constraint version number of the target components to provide a traceable chain of evidence. Before startup, the downstream enzyme catalysis and membrane linkage control links request the corresponding batch's phased operation strategy from the configuration database via the industrial network. Upon receiving the request, the configuration database searches for the strategy record by batch number and version number and returns it within a limited time, which can be set to five to thirty seconds. If no return is received within the limited time, the production control system can request again at fixed time intervals, with the number of retries set to two to five. If the retries still fail after exceeding the set number, a strategy missing error number is returned, and the batch is automatically prohibited from entering automatic control mode. Only after process personnel confirm the strategy status can the batch be manually set to fixed parameters in safe mode for operation.

[0055] To ensure idempotency, only one phased operation strategy record is allowed to be generated for the same batch under the same rule version. When the production control system receives a new strategy generation instruction, it first checks the configuration database to see if a strategy record with the corresponding batch and version number already exists. If it already exists, it directly returns the record identifier and prompts that it does not need to be regenerated and will not be written repeatedly. If it is necessary to modify the strategy record due to rule adjustment, the technical personnel will upgrade the rule version number and regenerate the strategy record while retaining the old version for traceability.

[0056] In one embodiment, to ensure the safety boundaries of the phased operation strategy, the process specification can stipulate that the initial stage should focus on a higher substrate feed rate, a lower donor sugar addition rate, a moderate enzyme dosage, and a lower membrane flux target; the intermediate stage should gradually increase the donor sugar addition rate and membrane flux target; and the later stage should reduce the substrate feed rate, increase the membrane flux target, and slightly increase the crossflow velocity within the allowable range, so as to concentrate on increasing the total steviol glycoside mass fraction and the proportion of higher-order steviol glycosides. At the same time, a safe range for reaction temperature and pH should be set in each stage to meet the enzyme activity requirements without exceeding the equipment tolerance.

[0057] When verifying the effectiveness of the phased operation strategy on-site, several batches can be selected within a trial operation period. The duration of each stage, the trajectory of the change in conversion indicator, the rate of membrane flux decay, and the composition of the final product are statistically analyzed to determine whether they conform to the target composition trajectory and the range of membrane operation constraints. The percentage of deviations can be set to 5% to 20% as an evaluation indicator. When the deviation ratio exceeds the set range, the control range or stage switching criteria in the phased operation strategy need to be adjusted, and the adjusted strategy should be recorded with a new version number.

[0058] In a preferred embodiment, for batches targeting high-purity, high-order steviol glycosides, in the initial stage, the target range for total steviol glycoside mass fraction can be set to a relatively low range, and the target range for the proportion of high-order steviol glycosides can be set to a range slightly higher than the initial level of the raw materials. During this stage, the substrate feed flow rate is controlled within the upper-middle range of the equipment capacity, the donor sugar addition flow rate is controlled within a lower range, the enzyme dosage is taken near the midpoint of the enzyme activity range, the reaction temperature is controlled within the midpoint of the suitable enzyme temperature range, and the membrane flux target is taken in the lower half of the global flux allowable range. In the intermediate stage, the target range for total steviol glycoside mass fraction is increased to a middle level, and the target range for high-order steviol glycosides is set to a range slightly higher than the initial level of the raw materials. The target range for the proportion of steviosides is increased to a portion close to the target specification, the substrate feed rate is slightly reduced, the donor sugar addition rate is increased, the enzyme replenishment frequency is increased, and the membrane flux target is raised to the middle of the global flux allowable range. In the later stage, the target range for the total steviol glycoside mass fraction and the target range for the proportion of higher steviol glycosides are tightened to a range close to the endpoint specification, the substrate feed rate is reduced to a lower level, the addition of donor sugar is stopped or significantly reduced, enzyme replenishment is stopped, and the membrane flux target is raised to the upper half of the global flux allowable range. At the same time, the crossflow velocity is controlled to a portion close to the aforementioned velocity upper limit, so that the product converges towards high purity in the later stage.

[0059] In another embodiment, the enzyme catalysis process can first provide a more detailed target range in the target composition trajectory configuration record according to more subdivision stages, and then merge adjacent subdivision stages into three major stages: early stage, middle stage and late stage when generating the phased operation strategy. Within each major stage, secondary control ranges can be set according to time or degree of conversion. This "subdivision and merging" method is equivalent to the method of directly dividing into three stages in terms of strategy generation logic, and can also form a clear phased operation strategy. It is an equivalent replacement of the present invention in terms of the granularity of stage division.

[0060] S4. Construct an enzyme-containing reaction vessel, enzyme retention membrane circuit, and glycoside fractionation membrane circuit system. Perform enzyme catalysis within the system. Adjust the flux of the enzyme retention membrane and the start / stop of the glycoside fractionation membrane according to the operating strategy. Part of the glycoside fractionation membrane retentate is returned to the enzyme reaction vessel. The glycoside fractionation membrane permeate is divided into a discharge stream and a return stream. The specific implementation is as follows: The process involves constructing an enzyme reaction vessel, an enzyme retention membrane loop, and a glycoside fractionation membrane loop system, and performing enzyme catalysis within the system. The flux of the enzyme retention membrane and the start-up and shutdown of the glycoside fractionation membrane are adjusted according to a phased operation strategy. The glycoside fractionation membrane retentate is partially refluxed back into the enzyme reaction vessel, and the glycoside fractionation membrane permeate is divided into a discharge stream and a return stream. This is a continuous operation process carried out on the same production line for specific batches after the target composition trajectory, membrane operation constraints, and phased operation strategy records have been written into the configuration database.

[0061] In this embodiment, the enzyme reactor is a jacketed stirred tank equipped with a raw material extract inlet, an enzyme preparation inlet, a substrate replenishment inlet, an enzyme retention membrane circuit connection port, and a glycoside fractionation membrane circuit connection port. The reactor is equipped with a temperature probe, a pH probe, and a level gauge. The temperature is expressed in Celsius, and the pH is expressed in pH values. The level signal is used to prevent idling or overflow. The enzyme retention membrane circuit consists of a circulating pump, an enzyme retention membrane module, a reflux pipeline, and corresponding valves. The circulating pump draws the reaction solution from the side or bottom of the enzyme reactor, filters it through the enzyme retention membrane module, and returns the enzyme-rich retentate to the enzyme reactor. The permeate is sent to the glycoside fractionation membrane circuit or a temporary storage tank. Pressure gauges, flow meters, and thermometers are installed at the circulating pump outlet, the membrane module inlet and outlet, and the reflux pipeline. The system is used to obtain signals such as transmembrane pressure difference, loop flow rate, and temperature. The transmembrane pressure difference can be calculated from the difference between the inlet and outlet pressures in each acquisition cycle. The glycoside fractionation membrane loop refers to a loop consisting of an outlet pump, a glycoside fractionation membrane module, a retentate return branch, a permeate discharge branch, and a permeate refill branch. The outlet pump draws the feed liquid from the enzyme reactor or the permeate storage tank of the enzyme retention membrane. After filtration by the glycoside fractionation membrane module, the retentate is divided by an adjustable diverter valve into a retentate branch that returns to the enzyme reactor and an optional external concentration branch. The permeate is divided by another set of adjustable diverter valves into a discharge flow that serves as product pre-concentration or waste liquid and a refill flow that returns to the enzyme reactor or the fore-stage storage tank. Each of the above branches is equipped with a flow meter and a switching valve to control the flow ratio according to the staged operation strategy.

[0062] During continuous operation, the production control system collects signals such as enzyme reactor temperature, pH, liquid level, pump speed, pipeline flow, membrane module inlet and outlet pressure and temperature from the field control unit according to a preset rhythm. The acquisition rhythm can be set from ten seconds to sixty seconds. Within each acquisition cycle, the instantaneous signal sequence is smoothed by removing out-of-range values ​​and taking the median value to obtain the measurement value representing that cycle. Then, it is compared with the reaction parameter control range and membrane parameter control range of the current stage in the phased operation strategy record to determine whether each parameter is within the target range or the degree of deviation from the target range. Based on this, the pump speed adjustment amount, valve opening target value, and glycoside fractionation membrane circuit start and stop status for the next cycle are generated.

[0063] In one embodiment, during the enzyme catalytic start-up phase, the production control system, based on the phased operation strategy for the current batch, first stabilizes the enzyme reactor within the target temperature and pH range for a period of time. Stability can be determined by setting the representative values ​​of temperature and pH to be within their respective safe ranges for three to five consecutive sampling cycles. Then, according to the reaction parameters recorded in the previous phase, the raw material extract and enzyme preparation are added to the reactor in batches or half-batch, and the enzyme retention membrane loop circulation pump is activated, allowing some of the reaction solution to pass through the enzyme retention membrane assembly, achieving enzyme molecule retention and small molecule glycoside permeation. At this time, the glycoside fractionation membrane loop can remain closed or only maintain a low-load pre-circulation state, specifically determined by the previous phase settings in the phased operation strategy. During the previous phase operation... During this period, the production control system continuously collects flow and pressure data of the enzyme retention membrane loop according to the above rhythm, calculates the transmembrane pressure difference and representative values ​​of membrane flux, and compares the representative values ​​of membrane flux with the current stage's flux target range. When the representative values ​​of membrane flux for two consecutive collection cycles are lower than the lower limit of the target range, the production control system increases the circulation pump speed or adjusts the valve opening appropriately without exceeding the safe operating range, so that the membrane flux approaches the midpoint of the target range. When the representative values ​​of membrane flux for two consecutive collection cycles are higher than the upper limit of the target range, the circulation pump speed is reduced or the throttling is appropriately increased to avoid overload. If the transmembrane pressure difference is close to the upper limit of the safe operating range, even if the target flux is not reached, the pump speed is reduced first and the risk event of pressure difference exceeding the limit is recorded.

[0064] Once the conversion reaches the mid-stage criteria specified in the phased operation strategy, such as the total steviol glycoside mass fraction and the percentage of higher-order steviol glycosides both falling within the mid-stage target range over several consecutive data collection cycles, the production control system switches to the mid-stage parameter set. During the mid-stage, the glycoside fractionation membrane loop is switched from closed to open. The opening operation can gradually increase the outlet pump speed and membrane flux target values ​​according to the buffer time to avoid instantaneous shocks. Simultaneously, according to the target range for the retentate recirculation ratio given in the phased operation strategy record, the representative value of the glycoside fractionation membrane retentate flow rate is compared with the representative value of the enzyme reactor recirculation branch flow rate. By adjusting the opening of the retentate diversion valve, the proportion of retentate recirculating to the enzyme reactor relative to the total glycoside fractionation membrane retentate flow rate is maintained at the median of the target range. Near the target range, the representative value of this ratio can be calculated by the ratio of the return branch flow rate to the total retentate flow rate over several consecutive sampling cycles. If the ratio deviates from the target range for two to three consecutive sampling cycles, the production control system issues an adjustment command. The adjustment range can be limited by the proportional step size parameter in the rule version. At the same time, based on the permeate discharge ratio and refill ratio in the phased operation strategy, the system divides the total flow rate of the glycoside fractionation membrane permeate into discharge flow and refill flow according to the target ratio. The representative value of the discharge flow rate is used to determine whether the predetermined discharge volume has been reached or whether the discharge needs to be reduced to ensure stable refill. Within a certain observation window, such as five to thirty minutes, if it is found that the refill flow rate is continuously low, causing the enzyme reactor liquid level or total glycoside concentration to be lower than the target lower limit, the discharge ratio is appropriately reduced and the reason for the adjustment is recorded.

[0065] In the later stage, when the conversion indicator continuously reaches the high-purity convergence range of the target composition trajectory, the production control system further reduces the addition of raw materials and donor sugars according to the phased operation strategy, keeping the enzyme retention membrane flux in the lower range to stabilize the enzyme enrichment state. At the same time, the target value of the glycoside fractionation membrane flux is increased and the cross-flow velocity is appropriately increased to allow more glycosides to permeate through the fractionation membrane. The reflux ratio of the retentate in the later stage can be appropriately reduced so that the high-purity retentate can be collected as the final concentrate. In this stage, the permeate return flow rate is mainly used to maintain the reactor liquid level and dilution, while the discharge flow rate is directly related to the final product yield.

[0066] During operation, the production control system compiles each stage switch time, enzyme retention membrane flux adjustment record, retentate reflux ratio adjustment record, permeate discharge and remixing ratio adjustment record, and the reasons for deviations that trigger these adjustments into a phased operation log entry. The entry records the time, batch number, current stage identifier, key parameter representative value, target range, adjustment range, operator confirmation information, and the version number of the phased operation strategy on which it is based. These log entries are stored in the batch operation log library for post-event traceability and rule optimization.

[0067] Between the field control unit and the upper-level production control system, instruction transmission can be organized by batch number and stage identifier. Each adjustment instruction includes the target pump speed, target valve opening, stage identifier, and instruction sequence number. After the field control unit completes the execution, it returns a status number. Status number zero indicates successful execution, number one indicates that the parameter exceeds the local hardware's allowable range, number two indicates that the equipment is not ready, and number three indicates that the critical sensor is faulty. When the production control system receives a non-zero status number, it records the instruction as a failure and executes the corresponding robust handling strategy based on the status number. For example, when the equipment is not ready or the sensor is faulty, it stops the relevant pump and closes the relevant valve, switches the system to a safe waiting state, and issues an audible and visual alarm and prompts for manual intervention on the interface.

[0068] To ensure idempotency, execution results received repeatedly under the same instruction sequence number in the same batch will not be repeatedly recorded in the operation log. The field control unit can directly return the previous result and ignore the actual action for new instructions with instruction sequence numbers that have been executed, thus avoiding duplicate actions caused by network jitter.

[0069] During a trial operation period, process engineers can select more than 20 batches of operating records to statistically analyze whether the enzyme retention membrane flux remains within the target range for each stage under the system configuration, whether the glycoside fractionation membrane reflux ratio and permeate discharge and remixing ratio are stably close to the strategic target, and whether the total steviol glycoside mass fraction and the proportion of higher-order steviol glycosides in the final product meet the endpoint requirements of the target composition trajectory. When it is found that the batch proportion deviates from the predetermined proportion, it can be determined that the segmented operation strategy and loop control combination can be stably implemented under the current equipment and raw material conditions.

[0070] In a preferred embodiment, the effective volume of the enzyme reactor can be set within a range of several cubic meters. The target flux range for the enzyme retention membrane in the early stage can be set to a range of several to several per square meter per hour. The target flux range for the glycoside fractionation membrane in the intermediate stage can be set to a range of several to several per square meter per hour. The target range for the reflux ratio of the glycoside fractionation membrane retentate can be set to a percentage to several percent of the total retentate flow rate. The permeate discharge ratio can be set to a percentage to several percent of the total permeate flow rate, with the remainder returned to the enzyme reactor as a reflux. In the later stage, the reflux ratio can be reduced to a lower range. By increasing the permeate discharge ratio to a higher range, a high-purity high-order steviol glycoside concentrate is obtained. In another embodiment, the retentate from the glycoside fractionation membrane circuit may not be directly refluxed back to the enzyme reactor, but instead refluxed to an intermediate buffer tank. The buffer tank is then distributed to the enzyme reactor and subsequent purification units by a separately regulated reflux pump according to the proportion set in the phased operation strategy. This "intermediate buffer tank + secondary reflux pump" structure is equivalent to the structure of direct reflux through a diversion valve in achieving the technical effects of enzyme enrichment and glycoside fractionation. It only changes the liquid storage location and flow rate regulation method, and is still an equivalent replacement of the device form in this invention.

[0071] S5. Update the operating condition fingerprint by collecting the reaction solution and the flux of each membrane according to the control cycle. When the operating condition fingerprint does not meet the above constraints, first adjust the substrate and donor sugar addition, then adjust the reflux ratio of the retentate and the membrane flux to complete the linkage correction. The specific implementation is as follows: The reaction solution and the flux of each membrane are collected according to the control cycle, and the fingerprint of the enzyme-membrane coupling operation is updated. When the updated fingerprint of the enzyme-membrane coupling operation no longer meets the constraints given in the target composition trajectory, membrane operation constraints and phased operation strategy, the production control system completes the linkage correction in the priority order of "substrate and donor sugar first, then retentate reflux ratio and membrane flux".

[0072] The control cycle refers to the time period during which the production control system periodically completes one on-site signal acquisition, feature processing, constraint comparison, and adjustment determination. The control cycle can be set to 30 to 300 seconds, preferably 60 to 120 seconds, and should be an integer multiple of the underlying acquisition rhythm (e.g., 10 to 60 seconds) so as to accumulate a number of acquisition points for smoothing within one control cycle.

[0073] The reaction liquid state parameters include temperature, pH, liquid level, estimated total steviol glycoside mass fraction, and estimated percentage of higher-order steviol glycosides within the enzyme reactor. Temperature is obtained using a temperature probe installed in the reactor body, pH is obtained using a pH probe, and liquid level is obtained using a level gauge. The total steviol glycoside mass fraction and percentage of higher-order steviol glycosides can be estimated using the aforementioned online near-infrared device and calibration relationship. The online device collects spectral signals within a short rhythm. At the end of the control cycle, the multiple estimation results within that cycle are denoised and the median value is taken to obtain the composition estimation value representing this control cycle. The flux of each membrane includes the flux of the enzyme retention membrane module and the flux of the glycoside fractionation membrane module. The flux is indirectly obtained by the ratio of the flow meter reading on the permeate side of the corresponding membrane module to the membrane area. Within the control cycle, the representative flux value is also obtained by removing out-of-range values ​​and taking the median value.

[0074] At the end of each control cycle, the production control system encapsulates the representative values ​​of the reaction liquid state quantity, the representative values ​​of each membrane flux, the representative values ​​of the current retentate reflux ratio, the representative values ​​of the permeate discharge ratio and the remix ratio, as well as the current early, middle and late stage identifiers, into a dynamic operating condition segment for this cycle. This segment is then aligned with the aforementioned static operating condition information (raw material extract composition, enzyme activity range, membrane baseline performance level, and target product level) to form a snapshot of the enzyme-membrane coupling operating condition at the current moment. The operating condition fingerprint snapshot is stored in the operating condition fingerprint database by batch number and timestamp. When storing the snapshot, the version number of the currently effective target composition trajectory, the version number of the membrane operation constraint, and the version number of the phased operation strategy are written to maintain the consistency of the evidence chain.

[0075] To determine whether the current operating condition meets the constraints, the production control system compares the estimated total steviol glycoside mass fraction and the estimated proportion of higher-order steviol glycosides in the operating condition fingerprint snapshot with the target interval of the corresponding stage in the target composition trajectory within each control cycle. It also compares the representative values ​​of each membrane flux with the current stage's flux target interval and operating safety zone, and compares the representative values ​​of the retentate recirculation ratio and the permeate discharge / recombination ratio with the target ratio interval in the phased operation strategy. If any critical quantity deviates from the target interval for several consecutive control cycles, and the deviation exceeds the set tolerance, the operating condition fingerprint is determined to not meet the above constraints. The number of consecutive control cycles can be set to two to five, and the tolerance can be given as a percentage range in the process specification. For example, the total steviol glycoside mass fraction and the proportion of higher-order steviol glycosides are allowed to deviate from the target interval boundary by a certain percentage, and each membrane flux is allowed to deviate from the target interval median by a certain percentage.

[0076] After determining that the constraints are not met, the production control system first makes corrections in the dimensions of substrate and donor sugar addition. In this embodiment, substrate addition refers to the flow rate of raw material extract and necessary feed, and donor sugar addition refers to the flow rate of saccharifying substance solution as enzyme catalytic donor. The operation of both is indirectly reflected by the rotation speed of the corresponding feed pump and the valve opening.

[0077] Based on the direction and degree of the current deviation, the production control system determines which type of feed to adjust first and the adjustment range according to the pre-set adjustment matrix in the rule version. For example, when the total steviol glycoside mass fraction is lower than the lower limit of the target range while the proportion of higher-order steviol glycosides is still within or slightly below the range, the combination of substrate feed flow rate and donor sugar addition flow rate is appropriately increased to increase the number of glycoside precursors participating in the reaction per unit time. When the proportion of higher-order steviol glycosides is significantly lower than the target range while the total steviol glycoside mass fraction is close to or has reached the upper limit, the substrate addition is maintained or even slightly reduced, and only the donor sugar addition combination is adjusted and the temperature or pH is fine-tuned if necessary to prevent the total steviol glycoside mass fraction from rising further.

[0078] The adjustment range of substrate and donor sugar addition can be limited by the step size range given in the process specification. For example, the change in the speed of each feed pump in a single control cycle should not exceed a certain percentage of the current value to avoid impacting the equipment and causing drastic fluctuations in the system. After each adjustment target is calculated, the production control system compares the new target value with the minimum and maximum values ​​in the rules. If it exceeds the boundary, it is truncated at the boundary value.

[0079] When the underlying execution unit receives the bottom material and donor sugar addition adjustment instructions, it identifies them by batch number and instruction sequence number. If the sequence number that has been executed appears again, it directly returns to the previous execution result and does not repeat the action. The production control system determines whether the adjustment is successful based on the status number. Status number one indicates that the set value exceeds the local hardware's allowable range, number two indicates that the pump is not ready, and number three indicates that the critical measurement signal has failed. When a non-zero status number is received, the adjustment record is marked as failed and the working condition fingerprint snapshot of this cycle is marked as "addition correction failed". At the same time, it enters the robust handling logic, such as suspending the addition adjustment of this batch and maintaining the safety parameter value of the previous cycle, waiting for manual intervention.

[0080] After the substrate and donor sugar adjustments are completed, the production control system observes whether the operating condition fingerprint falls back into the target range in the subsequent control cycles. If the key quantities such as total steviol glycoside mass fraction, higher steviol glycoside ratio, and membrane flux meet the constraints again within an observation window (e.g., five to thirty minutes) and do not trigger deviation judgment for several consecutive control cycles, then there is no need to further adjust the retentate reflux ratio and membrane flux. If there is still a continuous deviation after the observation window ends, or if the feed is close to the boundary value allowed by the process specification, the production control system enters the second layer of correction, that is, to perform linkage correction in the dimensions of retentate reflux ratio and membrane flux.

[0081] The retentate reflux ratio is reflected by the ratio of the retentate reflux branch flow rate of the glycoside fractionation membrane to the total retentate flow rate. It is calculated using the representative value of the flow meter in each control cycle. The membrane flux is calculated using the representative value of the permeate flow rate and the membrane area, as previously described, and has been compared with the safe operating range. The production control system adjusts the opening of the retentate reflux diversion valve and the speed of the fractionation membrane outlet pump according to the priority order and step size rules set in the rule version. For example, when the proportion of higher-order steviol glycosides is significantly low and the retentate reflux ratio is lower than the strategic target range, the retentate reflux ratio is increased first to send more of the enriched higher-order glycosides back to the enzyme reactor, while the flux of the glycoside fractionation membrane is controlled at a lower value to prolong the fractionation time. When the total steviol glycoside mass fraction is higher than the target range and the proportion of higher-order steviol glycosides has reached the target, the retentate reflux ratio can be appropriately reduced and the fractionation membrane flux increased to increase the discharge flow rate and accelerate the output of high-purity products.

[0082] The above adjustments are also iterated within each control cycle and are subject to operating safety zone constraints. When the transmembrane pressure difference approaches the upper limit or the flux decay rate exceeds the aforementioned threshold, the production control system shall not further increase the membrane flux. Otherwise, it shall be recorded as a request that violates the safety constraints and rejected. At the same time, a "membrane flux request rejected by safety rules" mark shall be written into the operating condition fingerprint snapshot.

[0083] To clearly retain the above dynamic updates and linkage correction records, the production control system writes the judgment results, adjustment measures and execution results of the current cycle into the batch operation log library at the end of each control cycle. The log record fields include batch number, timestamp, current stage identifier, representative value of main operating condition quantity, whether it deviates from the constraint flag, feed adjustment amount, retentate reflux ratio adjustment amount, membrane flux adjustment amount, status number and the rule version number on which it is based.

[0084] In a preferred embodiment, the control cycle can be set to 60 seconds, the number of consecutive cycles for deviation determination can be set to three, the observation window can be set to 15 minutes, the allowable deviation of the total steviol glycoside mass fraction and the proportion of higher-order steviol glycosides from the target range boundary can be set to 10% to 20% of the target range width, the adjustment range of the feed pump speed within a single cycle can be limited to 5% to 15% of the current value, the adjustment range of the retentate reflux ratio and the permeate discharge ratio within a single cycle can be limited to within 10% of the target range width, and the flux adjustment range can be limited to within 10% of the allowable flux range width. Through these settings, engineers can adjust the process according to the process specifications and equipment capabilities on the actual production line. Specific values ​​are selected to ensure that the operating condition fingerprint can stably follow the target trajectory during multiple batches of operation, and can also pull the system back to the target range when deviations occur by adjusting the substrate and donor sugar addition, as well as by linking the retentate reflux ratio and membrane flux. In another embodiment, the operating condition fingerprint update and linkage correction logic can also be deployed on the edge control unit. The edge unit directly performs deviation judgment and adjustment calculations in the local storage window according to the same rules, and then synchronizes the operating condition fingerprint snapshot and operation log to the upper system via the fieldbus. At this time, the overall control idea and parameter constraints remain unchanged, only the calculation position is moved from the upper system to the side closer to the device. In this invention, it is regarded as an equivalent replacement of the calculation deployment form.

[0085] S6. When the operating condition fingerprint meets the above constraints, stop the addition of enzyme and donor sugar, maintain the operation of the glycoside fractionation membrane circuit, and collect the glycoside fractionation membrane retentate as a high-purity steviol glycoside concentrate. The specific implementation is as follows: When the enzyme-membrane co-operation condition fingerprint obtained in several consecutive control cycles simultaneously satisfies all constraints regarding the endpoint stage in the target composition trajectory, membrane operation constraints, and phased operation strategy, the production control system executes the sequence of stopping feeding and endpoint enrichment collection, stopping the addition of enzyme and donor sugar, while keeping the glycoside fractionation membrane loop running under controlled conditions and collecting the glycoside fractionation membrane retentate as a high-purity steviol glycoside concentrate.

[0086] The "endpoint stage constraint" in this step refers to the fact that in the later stage, the estimated values ​​of total steviol glycoside mass fraction and the estimated values ​​of higher-order steviol glycoside proportion both fall within the endpoint target range and do not trigger deviation judgment for several consecutive control cycles. The representative values ​​of each membrane flux are within the corresponding stage flux target range and do not trigger the alarm flag of the safe operating zone. At the same time, the retentate return ratio and the permeate discharge and remixing ratio remain within the narrowing ratio range given by the endpoint stage strategy. The aforementioned judgment quantities are all derived from the control cycle acquisition signals and representative value generation methods defined in the previous embodiment.

[0087] In this embodiment, at the end of each control cycle, the production control system executes endpoint determination logic on the fingerprint of the current enzyme-membrane coupling operation. The endpoint determination logic includes: determining whether the estimated value of total steviol glycoside mass fraction and the estimated value of higher steviol glycoside proportion fall within the endpoint target range, and statistically analyzing whether these two estimated values ​​have continuously fallen within the target range and the deviation does not exceed the preset tolerance, such as 10% to 20% of the target range width in the most recent control cycles; determining whether the representative values ​​of enzyme retention membrane flux and glycoside fractionation membrane flux fall in the middle of the corresponding flux target range, and whether the transmembrane pressure difference is lower than the warning ratio of the upper limit of the operating safety zone, such as within 80% of the upper limit of the operating safety zone; determining whether the retentate reflux ratio has been adjusted to the lower range set in the endpoint stage, such as a ratio range that is more inclined to send out concentrate, whether the permeate discharge ratio has reached the higher range required in the endpoint stage, and whether the remixing ratio is in the lower range required in the endpoint stage. When the above conditions are met in several consecutive control cycles, such as three to five consecutive control cycles, and no new deviation events or safety rule rejections are recorded in the observation window, the endpoint determination logic gives the result of "endpoint conditions met".

[0088] When the production control system determines for the first time within a certain control cycle that the endpoint conditions are met, it does not immediately stop feeding. Instead, it first generates an "endpoint candidate" record, writes it to the batch operation log library, and records the current batch number, timestamp, key operating condition representative value, and the rule version number on which it is based. At the same time, it starts the endpoint confirmation observation window. Within this observation window, it continues to collect and update the operating condition fingerprint according to the original control cycle and keeps the aforementioned substrate and donor sugar linkage correction and retentate reflux ratio and membrane flux linkage correction in a frozen state. Safety rules are only executed when there is a significant safety risk. When the observation window ends, if no operating condition fingerprint triggers deviation judgment again or membrane operation safety zone alarm occurs during the period, the production control system generates an "endpoint confirmation passed" record and pushes the batch status to switch from "running" to "endpoint stop feeding preparation".

[0089] When the batch status switches to "endpoint stop feeding preparation", the production control system sends a stop feeding command to the field feeding unit through the industrial network. The stop feeding command includes the batch number, operation type field (stop enzyme feeding, stop donor sugar feeding), command sequence number and expected effective time. After receiving the command, the field feeding unit sets the target speed of the enzyme feeding pump and the donor sugar feeding pump to zero or to the safe stop speed, and closes the relevant valves. After completing the valve closing and pump shutdown actions, the field feeding unit returns a status number. A status number of zero indicates that the execution is complete. A non-zero status number indicates that the pump or valve did not respond as expected or that the liquid level or pressure detected on-site is not suitable for immediate shutdown. The production control system does not count repeated stop feeding commands of the same batch and the same command sequence number, but only retains the first execution result, thereby ensuring idempotency.

[0090] If the stop feeding status number is zero, the production control system will update the batch status to "endpoint enrichment" and write the status along with the stop timestamp into the operating condition fingerprint database and the operation log database for subsequent traceability and endpoint quality assessment. If the stop feeding status number is not zero, the production control system will record the fault and maintain the current process parameters within the safe range, waiting for manual handling. The endpoint enrichment step will not be entered until manual confirmation is obtained.

[0091] After the batch enters the "endpoint enrichment" stage, the feed to the enzyme reactor mainly comes from the reflux and internal reflux of the system. No new substrates or donor sugars are added. The production control system keeps the glycoside fractionation membrane loop running continuously. In one embodiment, the flux of the enzyme retention membrane can be maintained in the low to medium range of the flux target range in the later stage, and the flux of the glycoside fractionation membrane can be maintained in the high to medium range of the flux target range in the later stage. The retentate reflux ratio is kept in the low range set in the end-point stage so that the fractionation membrane retentate can be preferentially sent out as a high-purity steviol glycoside concentrate.

[0092] To achieve the collection of high-purity steviol glycoside concentrate, a dedicated product collection tank inlet branch is set up on the main line of the glycoside fractionation membrane retentate. The branch is equipped with an electric switching valve and a flow meter. After the production control system enters the "endpoint enrichment" state, there is a buffer period. The buffer period can be set to the length of several control cycles, such as three to ten control cycles, to allow the liquid composition in the system to smoothly transition from the "reaction running state" to the "endpoint enrichment state". During the buffer period, the retentate is still returned to the enzyme reactor or intermediate buffer tank through the reflux branch. Only when the buffer period ends and the endpoint conditions are still met will the production control system issue a "start collecting concentrate" command to introduce part or all of the retentate into the product collection tank.

[0093] During the product collection phase, the production control system collects the inlet flow rate of the product collection tank, the liquid level of the collection tank, and necessary online composition estimation signals according to the control cycle. It compares the estimated total steviol glycoside mass fraction and the estimated proportion of higher steviol glycosides at the inlet of the product collection tank with the target range at the endpoint. It sets quality monitoring thresholds for the collection phase, such as requiring the composition at the inlet of the product collection tank to remain within the target range and not lower than the median value of the target range for several consecutive control cycles. If it is found that the estimated composition value falls below the lower limit of the target or is significantly lower than the median value for several consecutive control cycles, the collection quality deviation event is recorded and collection can be paused. The retentate is then completely returned, and collection is restarted after adjustment.

[0094] During the collection process, the production control system records the product collection volume and corresponding composition estimate for each control cycle as a product batch sub-record and writes it into the product batch database. The product batch sub-record contains the batch number, sampling time, collection flow rate, cumulative collection volume, and online composition estimate. After the final collection is completed, quality management personnel can take samples at certain representative time points for retesting to verify the validity of the online estimation relationship.

[0095] In a preferred embodiment, the endpoint determination can be set as follows: the estimated total steviol glycoside mass fraction is not lower than a certain higher value; the estimated proportion of higher-order steviol glycosides is not lower than a certain higher value; the proportion of bitter glycosides is not higher than a certain lower value; and the mass fraction of small molecule sugars is not higher than a certain lower value. These indicators are required to be stable for at least three to five control cycles. Subsequently, a buffer time of six to twelve minutes is set, controlling the enzyme retention membrane flux to be in the lower-middle range of the global flux allowable range, for example, 40% to 60% of the global flux allowable range; controlling the glycoside fractionation membrane flux to be in the upper-middle range of the global flux allowable range, for example, 60% to 80% of the global flux allowable range; reducing the retentate reflux ratio to 10% to 30% of the total retentate flow rate; and increasing the permeate discharge ratio to 70% to 90% of the total permeate flow rate. In this way, within a collection period of one to several hours, a concentrate with a total steviol glycoside mass fraction and a higher-order steviol glycoside proportion close to the upper limit of the endpoint target can be obtained.

[0096] In another embodiment, the high-purity steviol glycoside concentrate may not enter the final product collection tank directly from the main line of the glycoside fractionation membrane retentate. Instead, it may be first introduced into an intermediate stabilizing tank. In the stabilizing tank, the composition fluctuations are further smoothed through small-scale reflux and mixing. Then, the stabilizing tank is sent to the subsequent crystallization or drying section at a constant flow rate. This "intermediate stabilizing tank + subsequent delivery" organization method is technically equivalent to entering the product collection tank directly from the fractionation membrane retentate branch. It is an equivalent replacement of the process layout in the final collection stage of this invention.

[0097] In the operating scenario shown in this embodiment: On a continuous production line equipped with an enzyme reaction vessel, an enzyme retention membrane circuit, and a glycoside fractionation membrane circuit, for a batch of stevia raw material numbered "B20260301-01", the planned product grade is high-purity high-grade steviol glycosides. The production line completes the automated operation of the entire batch according to the predetermined S1 to S6 steps.

[0098] Before production begins, the raw material receiving process registers batch number B20260301-01 in the Manufacturing Execution System. The raw material grade is medium, and on-site sampling tests show that the moisture content of the dried leaves is 9.8%, which is within the range of 8% to 12% allowed by the process specifications. The system uses the batch number as the index key for all subsequent online and laboratory records.

[0099] After being crushed, extracted, and coarsely filtered, the dried leaves from this batch entered the extraction tank. Flow meters, conductivity meters, pH meters, and thermometers installed online at the tank outlet collected the flow rate, conductivity, pH, and temperature signals of the extract at a 1-minute rhythm. The on-site control system discarded readings exceeding the range within each minute and used the median value as the representative value. After the liquid level in the extraction tank stabilized, the maximum, minimum, and average values ​​of each parameter were calculated using a 30-minute observation window. The results showed that the relative fluctuation amplitude was less than ±10% of the average value. Since this tolerance was met for three consecutive observation windows, the system marked the extraction condition of batch B20260301-01 as "stable."

[0100] During the later stages of extraction, three consecutive samples of the batch were taken from the online sampling point next to the extraction tank and sent to the laboratory for liquid chromatography analysis and physicochemical determination. The analysis results showed that the total steviol glycoside content of the extract ranged from 7.6% to 8.4%, the relative content of riboboroside A was approximately 60%, the total proportion corresponding to higher-order steviol glycosides was approximately 62%, the total proportion of bitter glycosides was less than 8%, the mass fraction of small molecule sugars ranged from 2.5% to 3.5%, the protein mass fraction was approximately 0.5%, the polysaccharide mass fraction was approximately 0.8%, the turbidity was grade I, the conductivity was 8–12 mS / cm, the pH was 6.3–6.7, and the temperature was 48–52℃. The difference between multiple measurements of the same indicator did not exceed 10% of the average value, and the analysis system automatically recorded the composition data of this batch as "qualified".

[0101] Meanwhile, in the enzyme activity test tank of the reaction workshop, the enzyme activity of a certain batch of enzyme preparation to be used was determined. Under the conditions of pH=6.0 and temperature of 50℃, the standard steviol glycoside substrate solution was continuously tested for 60 minutes. The effective activity of the enzyme batch was calculated to be that the substrate mass fraction decreased by about 2% per gram of enzyme per hour, which is within the range of 1% to 3% specified in the process specification.

[0102] The newly replaced microfiltration and nanofiltration membrane modules underwent baseline water flux testing under standard clean water conditions, yielding a baseline water flux of approximately 30 L / (m²). 2 The baseline water flux of the nanofiltration membrane is approximately 35 L / (m³). 2·h), according to the preset level classification, all belong to "medium level"; under the test of standard simulated material conditions, the nanofiltration membrane has a total steviol glycoside rejection rate of more than 90% and a small molecule sugar rejection rate of less than 30%, corresponding to a steviol glycoside rejection level of "high" and a small molecule sugar rejection level of "low".

[0103] The above online measurements, laboratory determinations, enzyme activity tests, and membrane baseline performance tests were aligned with batch number B20260301-01 under a unified time reference. Under the operating condition fingerprint generation rule version v1.3, the production control system generated the enzyme-membrane coupling operating condition fingerprint for this batch. The record includes fields such as raw material grade, medium dry leaf moisture range, total steviol glycoside mass fraction range of 7%–9% in the extract, relative content of rebaudioside A of 55%–65%, proportion of high-grade steviol glycosides of approximately 60%–65%, mass fraction of small molecule sugars of 2%–4%, mass fraction of protein and polysaccharides, turbidity grade, representative conductivity value, representative pH value, representative temperature value, enzyme effective activity range of 1%–3%, baseline water flux grade and retention grade of microfiltration and nanofiltration membranes, and target product grade "high purity high-grade steviol glycosides". The operating condition fingerprint version number FP-B20260301-01-v1.3 was assigned. The record is stored in the operating condition fingerprint database, using "batch number + version number" as the key.

[0104] Before the enzyme-catalyzed membrane separation of batch B20260301-01 is started, the process configuration system reads all fields of FP-B20260301-01-v1.3 from the operating condition fingerprint database, and at the same time reads the product technical standards that match the "high purity high-grade steviol glycosides" grade and the current specification version number PS-v2.0 from the product specification database: the endpoint requirement for total steviol glycoside mass fraction is ≥95%, the endpoint requirement for high-grade steviol glycoside ratio is ≥85%, the endpoint requirement for bitter glycoside ratio is ≤5%, and the endpoint requirement for small molecule sugar mass fraction is ≤3%.

[0105] The control system, based on the raw material extract's total steviol glycoside mass fraction of approximately 8% and the initial proportion of higher-order steviol glycosides of approximately 62%, combined with the enzyme's effective activity range of 1%–3%, and by statistically analyzing the historical conversion times of recent similar batches, divides the reaction process into three stages: the initial stage is expected to take 1.5–2.0 hours, the intermediate stage 1.5–2.5 hours, and the final stage 1.0–1.5 hours. The target total steviol glycoside mass fraction range for the initial stage is set at 8%–40%, and the target proportion of higher-order steviol glycosides is set at 62%–70%; the intermediate stage aims to increase the target total steviol glycoside mass fraction range to 40%–75%, and the proportion of higher-order steviol glycosides to 70%–80%; the final stage aims to set the target total steviol glycoside mass fraction range at 75%–95%, the proportion of higher-order steviol glycosides at 80%–85% or higher, with an upper limit of 5% for bitter glycosides and 3% for small molecule sugars. The aforementioned target intervals form a target composition trajectory TC-B20260301-01-v1.1 that progresses according to time and degree of transformation.

[0106] To determine the safe operating zone of the membrane, the control system retrieved data samples matching the current membrane module model, flux baseline level, and retention level from the membrane baseline performance test records and historical operating records. Data points with flux decay rates below 10% of the initial hourly flux were selected as safe samples. Statistical analysis showed that the stable flux of the microfiltration membrane in the historical safe samples was concentrated between 20 and 45 L / (m²). 2 The stable flux of nanofiltration membranes is concentrated in the range of 22–40 L / (m³). 2 Based on this, the allowable flux range for the microfiltration membrane in this batch is set at 18–40 L / (m). 2 The allowable flux range for nanofiltration membranes is 20–38 L / (m³). 2 Based on the maximum permissible transmembrane pressure difference from the membrane supplier and historical mechanical integrity records, the upper limit of the transmembrane pressure difference is set to 70% of the supplier's permissible value. The target crossflow velocity range is set to 70%–90% of the historical stable region, based on the principles of preventing deposition and avoiding overload. The above-mentioned permissible ranges for flux, transmembrane pressure difference, and crossflow velocity constitute the membrane operation safety zone MR-B20260301-01-v1.0, and are written into the configuration database along with the target composition trajectory, serving as the basis for subsequent phased operation strategy generation and linkage control.

[0107] After the target composition trajectory TC-B20260301-01-v1.1 and the membrane operation safety zone MR-B20260301-01-v1.0 have been written into the configuration database, the production control system enters the process configuration mode. Based on the target composition intervals and expected durations of each stage in the trajectory record, the enzyme catalysis process is defined as three stages: early stage, middle stage, and late stage.

[0108] The control system, based on the raw material composition and enzyme activity range in the operating condition fingerprint FP-B20260301-01-v1.3 and referring to the operating data of recent stable batches, sets the substrate feed flow rate control range for the early stage to 70%–90% of the reactor's rated feed capacity, the donor sugar addition flow rate control range to 20%–40% of the donor sugar metering pump's rated flow rate, the reaction temperature control range to 48–52℃, and the pH control range to pH 6.0–6.5. The initial enzyme dosage is calculated based on the median enzyme activity, and small-dose supplementary additions are allowed in the middle and later stages depending on the conversion indicator. In the intermediate stage, the substrate feed flow rate is appropriately reduced to 50%–70% of the rated capacity, and the donor sugar addition flow rate is increased to 40%–60%. The membrane flux target range is adjusted to the middle of their respective allowable ranges, and the cross-flow velocity target is kept near the middle of the allowable range. In the later stage, the substrate feed flow rate is further reduced to 20%–40%, the donor sugar addition is stopped or only the maintenance flow rate is maintained, and the flux targets of the microfiltration membrane and glycoside fractionation membrane are increased to the middle to high range of their respective allowable ranges. The cross-flow velocity target is increased to the upper half of the allowable range in order to accelerate the separation and concentration of high-purity glycosides.

[0109] Regarding the stage determination criteria, the system uses "the representative values ​​of total steviol glycoside mass fraction and the proportion of higher-order steviol glycosides falling within the target composition range of the stage for three consecutive collection cycles (5 minutes per cycle)" as the stage switching trigger condition. During stage switching, the adjustment of substrate feed flow rate, donor sugar addition flow rate, reaction temperature, and membrane flux target values ​​is set to a buffer gradient of 10–20 minutes, gradually and smoothly transitioning from the median of the target range of the previous stage to the median of the target range of the next stage. The aforementioned stage identifiers, determination criteria, and corresponding reaction parameter control ranges and membrane parameter control ranges are packaged into a staged operation strategy record RS-B20260301-01-v1.0, written to the configuration database using the batch number and trajectory version number as keys.

[0110] After configuration was completed, batch B20260301-01 entered the actual operation stage. The production control system first added an appropriate amount of process water to the enzyme reactor, heated the jacket to about 50°C and adjusted the pH to 6.2. After the representative values ​​of temperature and pH stabilized within the target range for 5 consecutive sampling cycles, the system was marked as "feedable".

[0111] Subsequently, following the parameter settings of the initial stage in the phased operation strategy, the raw material extract and enzyme preparation were added to the reactor in two steps. The initial substrate dosage corresponded to approximately 6%–8% of the total steviol glycosides in the reactor, and the enzyme dosage corresponded to the median of the expected conversion rate. The enzyme retention membrane loop circulation pump was started, and the inlet and outlet pressures were gradually increased. The microfiltration membrane flux stabilized in the lower half of the allowable range within 10–15 minutes, for example, 20–25 L / (m²).2 ·h); At this time, the glycoside fractionation membrane circuit remains closed and operates only in low-flux pre-circulation mode to ensure that the system is filled and the air is purged.

[0112] As the initial reaction proceeds, an online near-infrared spectral analyzer acquires spectra at 5-minute intervals along the reactor's circulation pipeline and calculates the total steviol glycoside mass fraction and the proportion of higher-order steviol glycosides. The system generates a current conversion indicator based on a comparison of the representative value with the target composition trajectory within the initial range. If the representative value remains stable within the initial target range for three consecutive acquisition cycles, the system considers the initial stage to have met expectations and prepares to enter the intermediate stage. Upon initiation of the intermediate stage, the production control system gradually increases the enzyme retention membrane circulation pump speed according to a buffering strategy, raising the microfiltration membrane flux target to 25–30 L / (m²). 2 (·h), simultaneously turn on the glycoside fractionation membrane extraction pump, and gradually increase the fractionation membrane flux to the middle of the allowable range, for example, 24–28 L / (m 2 ·h), and simultaneously open the retentate reflux branch and the permeate discharge and remixing branch. By adjusting the diversion valve, stabilize the retentate reflux ratio at 60% to 70% of the total retentate flow rate, control the permeate discharge ratio at 20% to 30%, and return the remaining part as remixing flow back to the reactor to maintain the reaction liquid level and composition balance.

[0113] During the mid-term operation, the production control system collects data on the reactor temperature, pH, liquid level, pump speed, pipeline flow rate, and membrane inlet and outlet pressure at a 10-30 second rhythm to form representative values ​​for each cycle. These values ​​are then compared with the parameter ranges in the phased operation strategy. When the representative value of the microfiltration membrane flux is found to be below the lower limit of the target range for two consecutive collection cycles and the transmembrane pressure difference is still far below the upper limit of the safe operation, the system automatically increases the speed of the circulating pump by a preset step. When the flux of the glycoside fractionation membrane approaches the upper limit of the allowable range and the transmembrane pressure difference approaches 70% of the safe upper limit, the speed of the lead-out pump is reduced first, and a "flux approaching the safe upper limit" alert event is recorded to prevent excessive membrane fouling or equipment overload.

[0114] When the total steviol glycoside mass fraction estimated online approaches 70% and the proportion of higher-order steviol glycosides approaches 80%, and remains in the upper half of the mid-term target range for several consecutive collection cycles, the system switches to the late stage according to the phased operation strategy. The substrate feed flow rate and donor sugar addition flow rate are gradually reduced until only the maintenance amount is retained or the donor sugar addition is completely stopped. At the same time, the flux targets of the microfiltration membrane and glycoside fractionation membrane are increased to the middle to high range of their respective allowable ranges. The retentate reflux ratio is slowly reduced to 40% to 50%, and the permeate discharge ratio is increased to 50% to 70% in order to accelerate the enrichment and output of high-purity glycosides.

[0115] Throughout the entire operation, the production control system uses a 60-second control cycle. During this cycle, it smooths and merges the reaction liquid state quantity, membrane flux value, retentate return ratio, permeate discharge and remixing ratio, etc., to form dynamic operating condition segments. These segments are then aligned with static operating condition information to generate timestamped enzyme-membrane coupling operating condition fingerprint snapshots such as FP-B20260301-01-v1.3-20260301T10:30, which are then stored in the operating condition fingerprint database.

[0116] During a certain mid-term control period, the system detected that the estimated total steviol glycoside mass fraction was slightly below the lower limit of the target range, while the proportion of higher-order steviol glycosides remained in the middle of the target range, and the membrane flux and retentate reflux ratio were both within the target range. After deviation detection for three consecutive control periods, the system first made corrections according to the priority order of "substrate first, donor sugar first": moderately increasing the feed flow rate of the raw material extract by 8% and the donor sugar addition flow rate by 10%, ensuring that the adjustment range of both in a single period did not exceed 10% of the current value. After the adjustment, in the subsequent two control periods, the estimated total steviol glycoside mass fraction returned to the target range, the proportion of higher-order steviol glycosides continued to rise along the target trajectory, and the system did not trigger the linkage adjustment of membrane flux or retentate reflux ratio during this period.

[0117] During another later stage of operation, the system detected that the estimated proportion of higher-order steviol glycosides had reached the upper part of the target range, while the total steviol glycoside mass fraction was slightly higher than the upper limit of the target range. After three consecutive control cycles, the substrate and donor sugars had dropped to the lower limit allowed by the process specifications, making further reduction unsuitable. The system then entered the second stage of correction: by reducing the retentate recirculation ratio by 5%–10%, increasing the target flux of the glycoside fractionation membrane by 5%–10%, and increasing the amount of high-purity permeate and retentate supplied without exceeding the transmembrane pressure differential safety limit and flux decay rate threshold. After several control cycles, both the total steviol glycoside mass fraction and the proportion of higher-order steviol glycosides fell back into the target range. The system recorded the parameter adjustments, execution status numbers, and corresponding rule version numbers throughout the correction process in the batch operation log.

[0118] Once the enzyme catalysis process enters its later stages, the production control system continuously monitors the estimated total steviol glycoside mass fraction and the proportion of higher-order steviol glycosides based on the endpoint requirements in the target composition trajectory. At a certain point, the online estimation results show that the total steviol glycoside mass fraction has stabilized in the range of 95%–97%, the proportion of higher-order steviol glycosides has stabilized in the range of 86%–89%, the proportion of bitter glycosides is below 4%, and the mass fraction of small molecule sugars is below 2.5%. Simultaneously, the representative flux values ​​of the microfiltration membrane and the glycoside fractionation membrane are in the middle of their respective flux target ranges, the transmembrane pressure difference is below 80% of the operational safety limit, the retentate recirculation ratio has been narrowed to 20%–30% according to the endpoint strategy, and the permeate discharge ratio has increased to 70%–80%. If these conditions remain stable for five consecutive control cycles without any new deviations or safety rule rejections, the endpoint determination logic outputs a "endpoint conditions met" result and generates a "endpoint candidate" record.

[0119] The system then enters the endpoint confirmation observation window, maintaining the substrate and donor sugar dosages unchanged for 10 minutes during which protective adjustments are only allowed in case of safety risks. At the end of the observation window, if all key quantities remain within the endpoint target range, the system generates a "Endpoint Confirmation Passed" record, switches the batch status from "Running" to "Endpoint Stop Feeding Preparation," and sends instructions to the feeding unit to stop enzyme and donor sugar feeding. Under the safety shutdown logic, the on-site feeding unit reduces the feeding pump speed to zero and closes the relevant valves, returning a status number of zero. The system updates the batch status to "Endpoint Enrichment" and writes the stop feeding timestamp to the operating condition fingerprint database.

[0120] After entering the "endpoint enrichment" state, the system maintains the enzyme retention membrane flux in the lower-middle range of the global flux allowable range, for example, 22–26 L / (m²). 2 •h), maintain the flux of the glycoside fractionation membrane in the upper-middle range of the global flux allowable range, for example, 28–32 L / (m 2 •h), and control the retentate reflux ratio at 15%–25% and the permeate discharge ratio at 75%–85%. On the main line of the retentate from the glycoside fractionation membrane, the system opens the electric switching valve leading to the product collection tank. Under the premise that the buffer time of 6–12 minutes has ended and the endpoint conditions are still met, the retentate is introduced into the product collection tank to start collecting high-purity steviol glycoside concentrate.

[0121] During the collection phase, the system collects the inlet flow rate and online composition estimation signals of the collection tank at 60-second control cycles. The system requires that the estimated total steviol glycoside mass fraction at the collection tank inlet be no less than 95% and the proportion of higher-order steviol glycosides be no less than 86% for multiple consecutive control cycles. If any indicator falls below the target lower limit for three consecutive control cycles, the system immediately shuts down the collection branch, recirculates all the retentate, and then readjusts the parameters according to the phased operation strategy and constraints before re-determining whether to resume collection. In this batch operation, the collection phase lasted approximately 50 minutes, accumulating several hundred liters of high-purity higher-order steviol glycoside concentrate. The online estimated composition indicators consistently remained in the upper part of the target range, and the deviations between the subsequent laboratory test results and the online estimates were all within 10%.

[0122] After the entire batch is completed, process and quality management personnel retrieve the enzyme-membrane coupling operation fingerprint records, target composition trajectory records, membrane operation constraint records, phased operation strategy records, and batch operation logs for this batch. By comparing the evolution of operating conditions in the early, middle, and late stages with the final product quality, it is confirmed that the process under this set of steps can stably achieve target composition trajectory control based on the raw material extract operating condition fingerprint, membrane operation safety zone constraint, and online linkage correction on the actual production line, and obtain a concentrated product that meets the specifications of high-purity high-order steviol glycosides at the final stage.

[0123] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0124] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0125] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center containing one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0126] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0127] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0128] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0129] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0130] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0131] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0132] In conclusion, 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 within the protection scope of the present invention.

Claims

1. A method for preparing high-purity steviol glycosides using enzyme-catalyzed membrane separation, characterized in that, include: S1. Collect the composition of raw material extract, impurities, enzyme activity and membrane baseline performance to generate an enzyme-membrane coupling condition fingerprint that includes the grades of raw materials, enzymes, membranes and target products. S2. Based on the operating condition fingerprint, set the target glycoside composition and concentration constraints, and determine the safe operating zone for membrane flux, transmembrane pressure difference and crossflow velocity, and construct the target composition trajectory and membrane operation constraints; S3. Based on the target composition trajectory and membrane operation constraints, the enzyme catalysis process is divided into an early stage, a middle stage and a late stage, and a phased operation strategy is formed by giving reaction parameters and membrane parameter control ranges for each stage. S4. Construct an enzyme reaction vessel, an enzyme retention membrane circuit, and a glycoside fractionation membrane circuit system. Perform enzyme catalysis in the system. Adjust the flux of the enzyme retention membrane and the start and stop of the glycoside fractionation membrane according to the operating strategy. Reflux part of the glycoside fractionation membrane retentate back into the enzyme reaction vessel and divide the glycoside fractionation membrane permeate into a discharge stream and a return stream. S5. Collect reaction solution and membrane flux according to the control cycle to update the operating condition fingerprint. When the operating condition fingerprint does not meet the above constraints, first adjust the substrate and donor sugar addition, and then adjust the reflux ratio of the retentate and the membrane flux to complete the linkage correction. S6. When the operating condition fingerprint meets the above constraints, stop the addition of enzyme and donor sugar, keep the glycoside fractionation membrane circuit running, and collect the glycoside fractionation membrane retentate as a high-purity steviol glycoside concentrate.

2. The method for preparing high-purity steviol glycosides using enzyme-catalyzed membrane separation according to claim 1, characterized in that, S1 includes: Register the batch number, raw material grade, and dry leaf moisture content upon receiving raw materials; At the outlet of the extraction tank, the flow rate, conductivity, pH and temperature of the extractant were collected, and abnormalities were removed and the stability of the extraction conditions was determined. After the extraction conditions are determined to be stable, the composition of the extract, the results of impurity detection, the results of enzyme activity test, and the baseline performance of the membrane are written into the enzyme membrane coupling condition fingerprint record according to the batch number.

3. The method for preparing high-purity steviol glycosides using enzyme-catalyzed membrane separation according to claim 2, characterized in that: Enzyme membrane coupled operation condition fingerprint records are stored in a shared database of the production control system and manufacturing execution system using batch number and operation condition fingerprint version number as indexes. When generating enzyme membrane coupled operation condition fingerprints, the generation time, operator identity and detection equipment code are recorded. When an enzyme-membrane co-processing fingerprint record is detected for the same batch under the same operating condition fingerprint version number, the first record is retained and duplicate generation is prevented. If an enzyme-membrane co-processing fingerprint is not generated after a preset delay, the batch is marked as having expired operating condition information and enzyme catalysis is prevented from being automatically started according to the operating condition of that batch.

4. The method for preparing high-purity steviol glycosides by enzyme-catalyzed membrane separation according to claim 1, characterized in that, S2 include: Based on the target product grade, raw material extract composition, enzyme activity range and membrane baseline performance in the enzyme-membrane coupling condition fingerprint, records matching the batch are extracted from the product specification library and the membrane baseline performance test record library, respectively. The target glycoside composition range, membrane flux operation safety zone, transmembrane pressure difference operation safety zone, and crossflow velocity operation safety zone are set according to the stages. The target glycoside composition range and the corresponding operation safety zone of each stage are combined to generate target composition trajectory configuration record and membrane operation constraint record, and written into the configuration database with batch number and trajectory version number as index.

5. The method for preparing high-purity steviol glycosides using enzyme-catalyzed membrane separation according to claim 1, characterized in that, S3 includes: The production control system divides the enzyme catalysis process into early, middle and late stages based on the target composition trajectory configuration record and membrane operation constraint record, and determines the identifier of each stage according to the conversion indication quantity. For each stage, the control range of reaction parameters including substrate feed flow rate, donor sugar addition flow rate, enzyme dosage, reaction temperature, pH, and stirring intensity is determined, as well as the control range of membrane parameters including membrane flux, transmembrane pressure difference, and crossflow velocity. A staged operation strategy record with batch number and trajectory version number as keys is generated and stored in the configuration database.

6. The method for preparing high-purity steviol glycosides by enzyme-catalyzed membrane separation according to claim 1, characterized in that, S4 include: A system that includes an enzyme reaction vessel, an enzyme retention membrane circuit, and a glycoside fractionation membrane circuit on the same production line; The glycoside fractionation membrane circuit sends the retentate from the glycoside fractionation membrane back to the enzyme reactor through the retentate reflux branch, and divides the permeate from the glycoside fractionation membrane into an outflow and a return flow. The production control system collects signals of enzyme reactor temperature, pH, liquid level, and flow and pressure of each loop according to a phased operation strategy. It generates pump speed adjustment commands and valve opening adjustment commands, which are executed by the field control unit according to batch number and command sequence number. The system also records phased operation logs based on the returned status number.

7. The method for preparing high-purity steviol glycosides by enzyme-catalyzed membrane separation according to claim 1, characterized in that, S5 include: The production control system sets a control cycle, and within each control cycle, it collects the temperature, pH, estimated total steviol glycoside mass fraction, estimated proportion of higher-order steviol glycosides, enzyme retention membrane flux, and glycoside fractionation membrane flux of the enzyme reactor. The above representative values, along with the retentate reflux ratio, permeate discharge ratio, permeate remixing ratio, and current stage identifier, form a dynamic operating condition segment. This segment is then aligned with the static operating condition information to generate an enzyme-membrane coupling operating condition fingerprint snapshot. This snapshot is written into the operating condition fingerprint database according to the batch number and timestamp, recording the target composition trajectory version number, membrane operation constraint version number, and phased operation strategy version number.

8. The method for preparing high-purity steviol glycosides by enzyme-catalyzed membrane separation according to claim 7, characterized in that: The production control system compares the estimated total steviol glycoside mass fraction, the estimated proportion of higher-order steviol glycosides, the enzyme retention membrane flux, the glycoside fractionation membrane flux, and the retentate reflux ratio with the target composition trajectory, membrane operation constraints, and phased operation strategies in the enzyme-membrane co-operation condition fingerprint snapshot according to the control cycle. When the constraints are not met, the feed pump speeds for substrate addition and donor sugar addition are adjusted according to the preset step size. If the constraints are still not met and the feed pump speed reaches the limit range of the process specification, the retentate reflux ratio and membrane flux are adjusted according to the preset sequence. The judgment result, adjustment amount and status number are written into the batch operation log library.

9. The method for preparing high-purity steviol glycosides by enzyme-catalyzed membrane separation according to claim 1, characterized in that, S6 include: When the enzyme-membrane co-processing fingerprint updated within a preset number of continuous control cycles meets the endpoint stage constraint, the production control system stops enzyme addition and donor sugar addition and switches the batch status to endpoint enrichment. The glycoside fractionation membrane loop is kept running under the enrichment state at the endpoint, and the flux of the enzyme retention membrane and the flux of the glycoside fractionation membrane are controlled according to the target flux range of the later stage. The reflux ratio of the retentate is controlled according to the ratio range of the endpoint stage. The retentate of the glycoside fractionation membrane is collected through the inlet branch of the product collection tank as a high-purity steviol glycoside concentrate.

Images