Extraction and separation system for extracting medicinal materials

Abstract

The invention relates to the technical field of solvent extraction separation and process automation control, in particular to an extraction separation system for medicinal material extraction, which comprises a controlled object interface layer, an industrial internet of things acquisition unit and an edge execution and monitoring unit. Data such as solvent ratio is collected, and the on-line detection assembly outputs a first component indication and a second component indication of the crude extract and the purity of the recovered solvent. The cloud twinborn optimization platform comprises an extraction-recovery coupled hybrid automaton digital twinborn model library (comprising a counter-current extraction stage, a replacement flushing stage and stage switching guard conditions) and identifiability judgment, online identification, mixed integer double-layer optimization, anti-fact admission and reachability verification modules, and generates a strategy evidence packet; and the edge side verification is executed and is monitored in a consistent invariant manner, and the preset security policy is rolled back abnormally. According to the method, a strategy is generated through cloud twin identification optimization, and an evidence packet is issued to edge gating execution after verification; and the abnormal rollback security policy is monitored consistently.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of solvent extraction separation and process automation control, and particularly relates to an extraction separation system for medicinal material extraction. BACKGROUND

[0002] Solvent extraction separation is widely used in the separation and purification of multi-component systems. In the context of medicinal material extraction, factors such as batch differences of raw materials, complex component systems, and process stage switching make process parameters and quality indicators prone to fluctuation. Moreover, there is a coupling relationship between the extraction unit and the solvent recovery unit, which makes it difficult for control methods based on experience or single-point adjustment to balance product quality and solvent recovery constraints.

[0003] In the prior art, for example, CN113359512B discloses a rare earth extraction separation process component content digital twin feature analysis method. By obtaining field process data and constructing a digital twin simulation and analysis platform, the process state is analyzed and predicted to provide a basis for process control analysis. However, such schemes usually focus on model and data-driven simulation analysis and state evaluation. When process strategies based on model calculations are used for on-site execution, fluctuations in working conditions, model deviations, or measurement noise may cause the strategy to deviate from the expected value during execution and touch critical process constraints. Existing solutions lack a verifiable closed-loop mechanism for "strategy issuance-site execution", making it difficult to verify the executability of the strategy in meeting key constraints before issuance, and also difficult to identify deviations in a timely manner and take safety measures during execution.

[0004] Therefore, the main technical problem to be solved by the prior art is: in the solvent extraction separation process of medicinal material extraction, how to realize the safe issuance and stable execution of the process strategy based on model optimization under the conditions of batch differences of raw materials and fluctuations in working conditions, so that the strategy has verifiable before issuance, monitorable during execution, and safety disposal in abnormality, thereby avoiding the destruction of key quality and recovery constraints. SUMMARY

[0005] In order to overcome the above technical defects, the purpose of the present application is to provide an extraction separation system for medicinal material extraction. The present application generates a process strategy through cloud digital twin and online identification driven optimization, and issues it to the edge gate execution in the form of evidence package after completing the executability verification before issuance, while monitoring the consistency of the execution process and rolling back to the safety strategy in abnormality, thereby realizing the safe issuance and stable execution of the strategy.

[0006] The present application discloses an extraction separation system for medicinal material extraction, comprising:

[0007] The controlled object interface layer is used for interacting data and issuing control instructions with the solvent extraction section, the solid-liquid separation section and the solvent recovery section;

[0008] The industrial internet of things collection unit is connected with the controlled object interface layer and is used for collecting process data, the industrial internet of things collection unit includes an online detection component, and the process data at least includes solvent proportioning, flow, temperature, pressure, crude extract liquid first component indication, crude extract liquid second component indication and recovery solvent purity, wherein the crude extract liquid first component indication, the crude extract liquid second component indication and the recovery solvent purity are output by the online detection component;

[0009] The edge execution and guardianship unit is in communication with the industrial internet of things collection unit and is electrically connected with the controlled object interface layer, and is used for generating process state data flow and uploading to the cloud;

[0010] The cloud twin optimization platform is in communication connection with the edge execution and guardianship unit, the cloud twin optimization platform includes an extraction-recovery coupled hybrid automaton digital twin model library, a distinguishability judgment module, an online identification module, a mixed integer bi-level optimization module, a counterfactual admission module and a reachability verification module;

[0011] The extraction-recovery coupled hybrid automaton digital twin model is used for characterizing the evolution of discrete process stages and continuous state variables, the discrete process stages at least include countercurrent extraction stages and displacement flushing stages, and the extraction-recovery coupled hybrid automaton digital twin model includes stage switching guard conditions for limiting stage switching trigger conditions;

[0012] The distinguishability judgment module is used for triggering the online identification module to update the extraction-recovery coupled hybrid automaton digital twin model and generate a model version identification when the distinguishability condition is met;

[0013] The mixed integer bi-level optimization module jointly solves the stage sequence, the segmented solvent proportioning and the segmented collection window under the model corresponding to the model version identification, and meets the selectivity constraints, the recovery solvent purity constraints, the solvent inventory consistency constraints and the stage switching reachability constraints defined by the crude extract liquid first component indication and the crude extract liquid second component indication;

[0014] The counterfactual admission module and the reachability verification module generate a strategy evidence package for the solving result;

[0015] The edge execution and guardianship unit executes the control instructions after receiving and verifying the strategy evidence package, and performs runtime consistency monitoring based on the selectivity consistency invariants, the recovery purity consistency invariants and the solvent inventory consistency invariants, and rolls back to a preset safety strategy when the monitoring fails.

[0016] Preferably, the online detection component includes a first wavelength channel and a second wavelength channel of an ultraviolet detector and / or two target peak area channels of an online chromatographic detector, and the selectivity constraint is jointly defined by the indication of the first component of the crude extract and the indication of the second component of the crude extract.

[0017] Preferably, the selectivity constraint includes a selectivity index threshold constraint, the selectivity index being calculated from the first component indicator and the second component indicator of the crude extract, and the mixed integer bilayer optimization module ensures that the selectivity index satisfies the selectivity index threshold constraint within at least one segmented collection window.

[0018] Preferably, the identifiability determination module calculates the parameter sensitivity information of the target parameter set based on the process status data stream and determines whether the identifiability conditions are met; when the identifiability conditions are not met, the cloud twin optimization platform sends an identification incentive request to the edge execution and monitoring unit, and the edge execution and monitoring unit executes the identification incentive request without violating the constraints on the purity of the recycled solvent and the consistency of the solvent inventory.

[0019] Preferably, the identified stimulus request includes a solvent ratio step sequence and / or a flow rate step sequence, and the edge execution and monitoring unit limits the execution of the solvent ratio step sequence and / or flow rate step sequence to the displacement rinsing phase and writes the stimulus identifier into the process status data stream.

[0020] Preferably, the online identification module updates the parameter set in groups according to discrete process stages, including at least: updating the first parameter set related to extraction mass transfer in the countercurrent extraction stage, and updating the second parameter set related to recovery load in the displacement rinsing stage, and generating a group parameter version number respectively.

[0021] Preferably, the purity constraint of the recovered solvent includes a purity range constraint and a recovery load upper limit constraint, and the mixed integer bilayer optimization module satisfies both the purity range constraint and the recovery load upper limit constraint when generating the segmented solvent ratio.

[0022] Preferably, the solvent inventory consistency constraint is jointly limited by the solvent ratio, flow rate, and purity of the recovered solvent, and the edge execution and monitoring unit performs online consistency verification on the process status data stream based on the solvent inventory consistency constraint.

[0023] Preferably, the segmented collection window has a set of window boundary conditions, which includes at least a threshold condition based on a selectivity index and a prediction condition based on an extraction-recovery coupled hybrid automaton digital twin model. The edge execution and monitoring unit opens the corresponding segmented collection window when both the threshold condition and the prediction condition are met.

[0024] Preferably, the baseline strategy of the counterfactual admission module includes at least two types of baselines: historical strategies and conservative strategies from the same batch, and the counterfactual admission module uses the counterfactual simulation output of the most unfavorable baseline as the admission comparison benchmark.

[0025] Preferably, the reachability verification module performs reachability verification on the phase switching guard conditions and phase dwell time constraints in the extraction-recovery coupled hybrid automaton digital twin model, and the strategy evidence package further includes a reachability verification summary identifier and a guard condition consistency identifier.

[0026] Preferably, the strategy evidence package further includes an input data window identifier and a data summary identifier. The input data window identifier is used to identify the data window range of the process state data stream on which the generation stage sequence, segmented solvent ratio, and segmented collection window are based. The edge execution and monitoring unit refuses to execute control commands when the input data window identifier does not match.

[0027] Preferably, the edge execution and monitoring unit maintains a policy version stack, which includes at least the current policy version and the previous policy version. Rolling back to the preset security policy includes locking the stage sequence to the stage sequence specified by the preset security policy and applying a limit to the segmented solvent ratio.

[0028] Preferably, the edge execution and monitoring unit includes a disconnection fault tolerance module. When communication with the cloud twin optimization platform is interrupted, the disconnection fault tolerance module caches the process status data stream and retransmits it in the order of event time after communication is restored. During the communication interruption, only the process policy with the most recent valid policy evidence package or the preset security policy is allowed to be executed.

[0029] Preferably, the edge execution and monitoring unit includes an idempotent execution module, which writes the control instruction into a policy sequence number and performs a duplicate issuance determination, and the edge execution and monitoring unit associates the policy sequence number with the policy evidence package and writes it into the audit log.

[0030] Compared with existing technologies, the above technical solution has the following advantages:

[0031] 1. Achieve a verifiable and traceable closed loop for policy issuance and execution: By solidifying the data window used in policy generation, constraint set version, counterfactual admission conclusion, reachability verification conclusion, and integrity signature into the policy evidence package, and verifying each item at the edge before gating execution, a verifiable link is formed from policy generation to execution, reducing the risk of unverified policies directly affecting the field.

[0032] 2. Improve process stability and consistency under complex coupling conditions: By establishing a runtime consistency invariant monitoring mechanism for selectivity indicators, purity of recovered solvent, recovery load and solvent inventory closure error, and triggering limit and rollback treatment when the conditions are not met, the extraction-recovery coupling process can still maintain operation within the preset constraint range under disturbance and time delay conditions.

[0033] 3. Enhance the executability and cross-stage constraint satisfaction of the strategy: By defining stage modeling based on hybrid automata, stage dwell time constraints and guard conditions, and combining reachability verification to make consistency judgments on stage switching and constraint reachability domains, the executability and predictability of the strategy when switching between the countercurrent extraction stage and the displacement flushing stage are improved.

[0034] 4. Improve adaptability and strategy effectiveness under batch differences: By identifying and updating the model version identifier online, the parameters of the digital twin model are adaptively updated with changes in batch and operating conditions. After the update, joint optimization is performed to solve the problem, thereby improving the strategy matching degree under different batches of medicinal materials and different feed fluctuation conditions.

[0035] 5. Improve the reliability and consistency of collection window determination: By using threshold conditions, prediction conditions, holding time and minimum window duration as the collection window gating conditions, and making the determination at the edge side based on the process status data stream aligned with event time, the opening and closing of the collection window has a unified standard, reducing window boundary deviations caused by instantaneous noise or late data.

[0036] 6. Achieve consistent association between "process control events - sample retention - test results": By binding the sample retention number with the strategy sequence number and window number, and solidifying the A1 / A2 peak identification and integration rules, online indicator data, offline fingerprint spectrum and collected fraction results can be associated with specific windows and execution events in a unified manner, which facilitates quality traceability and inter-batch comparative analysis.

[0037] 7. Enhance robustness and continuous operation capability under cloud-edge collaboration: By using cached retransmission for fault tolerance during disconnection and policy source restriction during disconnection, and with idempotent execution to provide idempotent confirmation for repeated command issuance, the continuity of the process can be maintained and the uncertainty caused by repeated actions can be reduced in communication fluctuation scenarios.

[0038] 8. Enhance on-site maintainability and fault location efficiency: By performing structured audit records on evidence package verification results, stage switching events, window events, monitoring failure events, and rollback / recovery events, key nodes in the operation process have interpretable event trajectories, making it easier for maintenance personnel to locate the abnormal triggering links and reproduce the problem conditions. Attached Figure Description

[0039] Figure 1This is a schematic diagram of the overall system structure of an extraction and separation system for extracting medicinal materials according to the present invention;

[0040] Figure 2 This is a schematic diagram of the strategy generation and gating execution process;

[0041] Figure 3 This is a diagram illustrating runtime monitoring and rollback.

[0042] Figure 4 This is a schematic diagram comparing the curves of selective indicators;

[0043] Figure 5 This is a schematic diagram comparing the purity curves of the recovered solvent;

[0044] Figure 6 This is a schematic diagram comparing the closed-loop error curves of solvent inventory.

[0045] Figure 7 This is a schematic diagram of the fingerprint spectrum of medicinal material extract (HPLC response example, with the indicator peaks of active ingredients and impurity / background indicator peaks marked);

[0046] Figure 8 A diagram illustrating the results of fraction collection (example percentages of fractions W1, W2, and W3). Detailed Implementation

[0047] The advantages of the present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments.

[0048] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0049] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The singular forms “a,” “the,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0050] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0051] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0052] In the description of this invention, unless otherwise specified and limited, it should be noted that the terms "installation", "connection" and "linking" should be interpreted broadly. For example, they can refer to mechanical or electrical connections, or internal connections between two components. They can be direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0053] In the following description, suffixes such as "module," "part," or "unit" used to denote elements are used only for the convenience of the description of the invention and have no specific meaning in themselves. Therefore, "module" and "part" can be used interchangeably.

[0054] This embodiment provides an extraction and separation system for medicinal material extraction. It is designed to generate, validate, and distribute process strategies, implement gating, and ensure consistency during operation, even under conditions of batch variations in raw materials, feed fluctuations, solvent recovery load disturbances, and network latency. This ensures the extraction and separation process operates within traceable constraints. See also... Figure 1The extraction and separation system for medicinal material extraction includes a controlled object interface layer 100, an industrial IoT acquisition unit 200, an edge execution and monitoring unit 300, and a cloud-based twin optimization platform 400. The controlled object interface layer 100 is used for data and control command interaction with the solvent extraction section 10, the solid-liquid separation section 20, and the solvent recovery section 30. The industrial IoT acquisition unit 200 is used to collect process data and output online quality indicators. The edge execution and monitoring unit 300 is used to form a process status data stream, verify the strategy evidence package, perform gated execution of verified process strategies, and monitor and roll back runtime consistency. The cloud-based twin optimization platform 400 is used to complete online identification, joint optimization solution, counterfactual access and reachability verification based on a digital twin model, and generate a strategy evidence package. See also... Figure 2 The strategy evidence package is generated and distributed by the cloud-based twin optimization platform 400, and enters gating execution after verification by the edge execution and monitoring unit 300; see also Figure 3 The runtime consistency monitoring triggers a rollback and establishes an audit trail; see [link / reference]. Figure 4 to Figure 6 Examples of comparative curves showing the changes of key operating indicators over time are provided; see [link / reference]. Figure 7 and Figure 8 The fingerprint spectrum of the medicinal material extraction results and the sample collection results are shown.

[0055] Figure 4 This is used to represent the trajectory of the selectivity index over time under two operating modes: "Comparison Scheme (Symmetric Analysis / Optimized Distribution Only)" and "This Implementation Scheme (Evidence Package Gating + Admission Verification + Monitoring Rollback)," within the same batch and statistical period. The selectivity index is calculated as the ratio of the indicator data of the first component of the crude extract to the indicator data of the second component of the crude extract, reflecting the relative relationship between the online detection component's response to the target active ingredient and its response to impurities / background. The threshold line in the curve corresponds to the determination criteria for the "lower boundary of the selectivity threshold," and the fluctuation amplitude and continuous holding section of the curve correspond to the determination criteria related to the "guard holding time" and "collection window gating."

[0056] Figure 5 This is used to represent the change trajectory of recovered solvent purity over time under two operating modes within the same batch and the same statistical period. The recovered solvent purity is output from the online purity analysis channel on the recovery side. The upper and lower limits given in the curve correspond to the range determination criteria for the "lower limit of recovered solvent purity" and the "upper limit of recovered solvent purity." The percentage of time the curve falls within the range characterizes the operating record of the recovery side within the allowable purity range. The fluctuation of the curve near the range boundary characterizes the impact of load disturbances or changes in the recovery process state on the purity data.

[0057] Figure 6 This diagram illustrates the change in solvent inventory closure error over time under two different operating modes within the same batch and statistical period. The solvent inventory closure error is calculated from the total amount of solvent input, total amount of solvent recovered, and inventory change within the statistical window, reflecting the degree of closure of the solvent material balance within the statistical window. The threshold line in the diagram corresponds to the determination criteria for the "upper bound of inventory closure error." The section of the curve exceeding the threshold corresponds to the recording section in the runtime consistency monitoring where the "inventory consistency invariant is not satisfied," and is correlated with the time point of the rollback event in the audit log.

[0058] Figure 7 This example illustrates the response curve of the fingerprint spectrum of the medicinal extract obtained after detecting samples collected at a combined or representative time point within the collection window (the example is represented by the change in HPLC response with retention time). The A1 indicator peak is used for peak identification and integration of the "first component indicator data of the crude extract," and the A2 indicator peak is used for peak identification and integration of the "second component indicator data of the crude extract." By performing peak identification and area integration within a preset retention time window, indicator data for calculating selectivity indicators can be obtained, and this data is used to correlate the offline fingerprint spectrum results with the output of the online detection component in terms of caliber.

[0059] Figure 8 This diagram illustrates the compositional differences in target active ingredient indication and impurity / background indication between different collected fractions obtained through collection windows W1, W2, and W3. The bar or percentage value corresponding to each window in the diagram represents an example of the "active ingredient percentage" and "impurity indication percentage" in the sample within that window; these percentages can be obtained by normalizing the sample after content determination, or by normalizing the indicator peak areas (e.g., A1, A2, and the total response) according to a preset normalization rule. Figure 8 This is used to correspond to the collection window gating records, making it easier to establish a correlation between the "time boundary of window opening / closing" and the "fraction detection result corresponding to the window".

[0060] It should be noted that, in this embodiment, the interaction method between the controlled object interface layer 100 and the three work sections will be described in detail as follows:

[0061] The controlled object interface layer 100 is physically connected to the field controller, actuators, and detection equipment via industrial Ethernet, fieldbus, or a security isolation gateway, and logically provides a data subscription interface and a command issuance interface. The data subscription interface subscribes to at least solvent ratio data, flow rate data, temperature data, pressure data, valve position status data, pump speed status data, recovered solvent purity data, and recovery-side load-related data, and adds a collection timestamp to each data entry. The command issuance interface includes at least the issuance of solvent ratio setpoints, flow rate setpoints, stage switching commands, and collection window opening / closing commands, and receives execution receipts. To ensure traceability of the execution chain, the controlled object interface layer 100 writes a strategy sequence number to each issued command and requires the field controller to return an execution receipt containing the strategy sequence number. The execution receipt includes at least the actuator arrival time, arrival status, and failure reason code.

[0062] The solvent extraction section 10 is used to transfer the target active ingredient from the medicinal material to the solvent phase. The solid-liquid separation section 20 is used to separate the extracted liquid phase from the solid phase to ensure downstream recovery stability. The solvent recovery section 30 is used to recover solvent-containing materials to obtain recovered solvent and maintain the recovery-side load within the allowable range. In this embodiment, the three sections are considered as coupled objects. The coupling is at least reflected in the changes in the recovery-side load and the purity of the recovered solvent caused by changes in the solvent ratio and flow rate on the extraction side. In turn, changes in the recovery-side load limit the magnitude of changes in the solvent ratio and the frequency of stage switching on the extraction side.

[0063] It should be noted that, in this embodiment, the industrial IoT acquisition unit 200, the online detection component 210, and the medicinal material example will be described in detail as follows:

[0064] The industrial IoT acquisition unit 200 is used to collect process data, which includes at least solvent ratio data, flow rate data, temperature data, pressure data, crude extract first component indicator data, crude extract second component indicator data, and recovered solvent purity data. The industrial IoT acquisition unit 200 includes an online detection component 210, which outputs crude extract first component indicator data, crude extract second component indicator data, and recovered solvent purity data. In one embodiment, the online detection component 210 includes an ultraviolet detector 211 with a first wavelength channel and a second wavelength channel, wherein the first wavelength channel characterizes the target active ingredient response, and the second wavelength channel characterizes characteristic impurities or background response. In another embodiment, the online detection component 210 includes a chromatographic online detector 212 with two target peak area channels, wherein the first peak area channel characterizes the target active ingredient peak area, and the second peak area channel characterizes the characteristic impurity peak area. The recovered solvent purity data is output from an online refractive index, density, or chromatographic purity analysis channel, and recovered-side load-related data such as recovery-side flow rate, steam flow rate, electric heating power, or compressor power are simultaneously acquired.

[0065] To clarify the extraction scenario of medicinal materials, examples of medicinal materials used in this embodiment as extraction targets include, but are not limited to, Scutellaria baicalensis, Salvia miltiorrhiza, and Glycyrrhiza uralensis. The target active ingredient can be one of flavonoids, phenolic acids, or saponins. The online detection component 210 defines the indicator peak or indicator channel output corresponding to the target active ingredient as the first component indicator data of the crude extract, and defines the indicator peak or indicator channel output corresponding to impurities or background as the second component indicator data of the crude extract. The method for determining the indicator peak or indicator channel is given in the "Peak Identification and Integration Rules" section below.

[0066] For ease of explanation, the following symbol definitions are used: the indicator data of the first component of the crude extract is denoted as... The indicator data for the second component of the crude extract is denoted as The purity data of the recovered solvent is recorded as The extraction-side flow rate data is denoted as Solvent ratio data are recorded as The relevant data on the recovery side load are denoted as .

[0067] It should be noted that this embodiment will provide a detailed explanation of the definition and normalization method of the recovery load data, as follows:

[0068] To ensure that the meaning of the recovery load is measurable and comparable, this embodiment uses data related to the recovery-side load. Normalizing the rated values ​​yields the recovery load data. Let the relevant load data U on the recovery side represent the heating energy input or equivalent energy consumption per unit time, and let the corresponding rated value be... The recovery load data is then defined as:

[0069]

[0070] in, For the recovered load data, U represents the recovered load-related data. The load-related data is the data corresponding to the rated load; when steam flow rate is used as the load-related data U on the recovery side, the load-related data corresponding to the rated load is... The steam flow rate is the rated operating temperature; when the electric heating power is used as the load-related data U on the recovery side, the load-related data corresponding to the rated load is... This is the rated heating power.

[0071] It should be noted that, in this embodiment, the edge execution and monitoring unit 300 will be described in detail, including timing alignment, stage tracking, gated execution, idempotency, and disconnection fault tolerance, as follows:

[0072] The edge execution and monitoring unit 300 includes at least a time synchronization module 310, a stage tracking module 320, an evidence package verification module 330, an instruction orchestration module 340, a runtime monitoring module 350, a rollback module 360, a disconnection fault tolerance module 370, and an idempotent execution module 380.

[0073] The timing alignment module 310 is used to align data from different acquisition channels for event time and form a process status data stream. The sampling period is "sampling period", the maximum allowable time deviation for alignment is "maximum allowable time deviation for alignment", and the late threshold is "late threshold". When the timestamp difference between data from different channels at the same time is less than the maximum allowable time deviation for alignment, it is aligned by nearest neighbor. When it is greater than the maximum allowable time deviation for alignment but less than the late threshold, it is aligned by linear interpolation. When it is greater than the late threshold, it is marked as a late segment and written to the late identifier field.

[0074] The stage tracking module 320 is used to identify discrete process stages and maintain their status. Discrete process stages include at least a countercurrent extraction stage and a displacement rinsing stage. The stage tracking module 320 determines stage switching based on valve group switching status, solvent ratio switching events, collection window opening and closing events, and flow stability conditions. Flow stability conditions are determined using the flow variation coefficient within the "flow stability sliding window length." Defined as:

[0075]

[0076] in, The standard deviation of the flow rate within the sliding window. The average flow rate within the sliding window is used; flow stability is determined when the flow rate variation coefficient is not greater than the "flow stability threshold". The stage tracking module 320 encodes the stage identifier, stage start and end time, switching trigger event, and switching reason into the process status data stream.

[0077] The evidence package verification module 330 is used to verify the policy evidence package issued by the cloud twin optimization platform 400. The verification includes at least the input data window identifier matching verification, data digest identifier consistency verification, constraint set identifier consistency verification, verification result identifier validity verification, and integrity signature verification. When any verification fails, the evidence package verification module 330 outputs a rejection event to the instruction orchestration module 340 and writes the rejection reason to the audit log.

[0078] The instruction orchestration module 340 is used to convert the phase sequence, segmented solvent ratio, segmented flow rate and segmented collection window in the strategy evidence package into a control instruction sequence that can be recognized by the controlled object interface layer 100. The control instruction sequence is issued in the order of "solvent ratio first, then flow rate, switch phase after meeting guard conditions, open and close collection window after meeting window boundary conditions", and receives execution receipts.

[0079] The idempotent execution module 380 is used to determine the idempotency of repeatedly issued control commands. The idempotency determination uses the strategy sequence number and the step sequence number to form the idempotency key. When the repeated issuance of the same idempotency key is detected, only an acknowledgment is returned without triggering the executor to repeat the action, and the number of repeated issuances and the trigger source are recorded in the audit log.

[0080] The disconnection fault tolerance module 370 is used to handle the disconnection situation between the cloud and the edge. When the communication with the cloud twin optimization platform 400 is interrupted, the disconnection fault tolerance module 370 caches the process status data stream and retransmits it in the order of event time after the communication is restored. During the communication interruption, only the process policy with the most recent valid policy evidence package or the preset security policy is allowed to be executed, and the reception of new policy evidence packages that have not been verified is prohibited.

[0081] It should be noted that this embodiment will provide a detailed description of the cloud-based twin optimization platform 400: model, identifiability determination, online identification, joint optimization, counterfactual access, and reachability verification, as follows:

[0082] See Figure 1 and Figure 2 The cloud-based twin optimization platform 400 includes an extraction-recovery coupled hybrid automata digital twin model library 410, an identifiability determination module 420, an online identification module 430, a hybrid integer two-layer optimization module 440, a counterfactual access module 450, and an accessibility verification module 460.

[0083] A digital twin model of an extraction-recovery coupled hybrid automata is used to characterize the coupled evolution of discrete stages and continuous states. The discrete stages include at least a countercurrent extraction stage and a displacement flushing stage, while the continuous states include at least extraction mass transfer-related states and recovery load-related states. The model includes stage switching guard conditions and stage residence time constraints. In one embodiment, the stage residence time constraints are set as "countercurrent extraction stage residence time" and "displacement flushing stage residence time." In one embodiment, the stage switching guard conditions consist of a combination of conditions, including satisfying flow stability conditions, recoverable solvent purity being within the allowable range, selectivity indicators meeting requirements for a specified duration, and the arrival of a stage switching command. The selectivity holding time is the "guard holding time."

[0084] How to obtain the sensitivity matrix:

[0085] To achieve identifiability determination and online identification, this embodiment provides a sensitivity matrix. The method of obtaining the sensitivity matrix. In one implementation, the cloud-based twin optimization platform 400 uses a numerical difference method to calculate the sensitivity matrix. Let the objective parameter vector be... Let the model output feature vector be ,in It should include at least selectivity indicators, recovered solvent purity and recovery load data, or observations composed of these; for each parameter Apply small perturbation The outputs were obtained through simulation. and Then, the sensitivity column vector for this parameter is approximated by difference:

[0086]

[0087] in, Representing the sensitivity matrix The To ensure the stability of the difference approximation, Take 1% to 5% of the nominal value of the parameter.

[0088] Identifiability determination module and online identification module:

[0089] The identifiability determination module 420 is used to determine whether to trigger the online identification module 430 to update the model parameters. In one implementation, the identifiability determination employs an information matrix threshold criterion. The information matrix is ​​defined as:

[0090]

[0091] in, For information matrix, This is the sensitivity matrix. The identifiability condition is... and Joint judgment, among which The positive threshold value is [value]. , The number of target parameters.

[0092] When the identifiability condition is not met, the cloud-based twin optimization platform 400 issues an identification incentive request. The edge execution and monitoring unit 300 then performs a solvent ratio or flow rate step during the displacement rinsing phase. The solvent ratio step range is 2%–8% relative to the current ratio, and the flow rate step range is 3%–10% relative to the current flow rate. Each step lasts 30–180 seconds and is limited to the displacement rinsing phase. The online identification module 430 updates the parameter set in groups according to discrete stages. This includes updating the first parameter group related to extraction mass transfer and generating a first parameter version number during the countercurrent extraction phase, and updating the second parameter group related to recovery load and generating a second parameter version number during the displacement rinsing phase. The first and second parameter version numbers are then combined to generate a model version identifier, which is written into the extraction-recovery coupled hybrid automata digital twin model library 410.

[0093] Mixed-integer two-level optimization module and constraint definition:

[0094] The mixed-integer two-level optimization module 440 jointly solves the stage sequence, segmented solvent ratio, segmented flow rate, and segmented collection window under the model corresponding to the model version identifier, while satisfying selectivity constraints, recovered solvent purity constraints, solvent inventory consistency constraints, and stage switching reachability constraints. To define the selectivity constraints, the selectivity index is calculated using the following formula:

[0095]

[0096] in, As a selective indicator, This is the indicator data for the first component of the crude extract. This is the indicator data for the second component of the crude extract. The rate of change in selectivity is calculated using a difference approximation on the edge side:

[0097]

[0098] in, The differential time interval ranges from 1 min to 5 min. The purity constraint of the recovered solvent is set as an interval constraint consisting of the "lower limit of the purity of the recovered solvent" and the "upper limit of the purity of the recovered solvent". The recovery load constraint is set as an upper limit constraint on the recovery load data using the "upper limit of the recovery load".

[0099] Solvent inventory consistency constraints are defined using the inventory closure error definition, assuming the total solvent input within the statistical window is... The total amount recovered is The change in inventory was The inventory closure error is calculated using the following formula:

[0100]

[0101] in, For inventory closure error, To prevent extremely small positive numbers with a denominator of zero, the value is taken as... The upper limit of inventory closure error is "upper bound of inventory closure error".

[0102] The segmented collection window has a set of window boundary conditions. The set of window boundary conditions includes at least threshold conditions and prediction conditions. The threshold conditions include at least a selectivity index not less than the "lower limit of the selectivity threshold" and a selectivity change rate not greater than the "upper limit of the selectivity change rate". The prediction conditions include at least the condition that the predicted value of the selectivity index will not be lower than the "lower limit of the selectivity threshold" in the future "prediction time domain". The minimum collection window duration is the "minimum collection window duration". The edge execution and monitoring unit 300 only opens the corresponding collection window when the threshold conditions and prediction conditions are met simultaneously.

[0103] Counterfactual Admission Module and Most Unfavorable Baseline Criterion

[0104] The counterfactual admission module 450 is used to perform counterfactual simulation comparison between the solved process strategy and the baseline strategy and output the admission conclusion. The baseline strategy includes at least two types of baselines: historical strategies from the same batch and conservative strategies. To clarify the selection criteria for the most unfavorable baseline, this embodiment stipulates that the most unfavorable baseline is selected according to the principle of "prioritizing the degree of constraint violation, followed by the degradation of the target value": first, the degree of constraint violation of each baseline in the simulation time domain is compared. ,like Take the different The largest is the most unfavorable baseline. If they are the same, compare the overall target value. and take The largest is the most unfavorable baseline, where the degree of constraint violation is... It consists of the weighted integrals of each constraint outbound quantity, and the constraint outbound quantities include at least the following: , , , , The reachability verification module 460 performs reachability verification on policies that have passed counterfactual access. The reachability verification checks at least the consistency of phase switching guard conditions, phase dwell time constraints, and reachability domain boundaries related to recovery purity, and outputs a reachability verification summary identifier and a guard condition consistency identifier. After the counterfactual access and reachability verification are passed, the cloud-based twin optimization platform 400 generates a policy evidence package and distributes it to the edge execution and monitoring unit 300.

[0105] It should be noted that, for ease of understanding, in this embodiment, the degree of constraint violation will be described separately. To elaborate further, the details are as follows:

[0106] Severity of Constraints It consists of the weighted integrals of all constraint overbound quantities, and is defined as:

[0107]

[0108] in, For selective violation scoring, To recover the solvent purity violation integral term, To recover the load violation integral term, This is the integral term for inventory closure error violations; , , , For the corresponding weight coefficients, satisfying , , , ≥0 and =1.

[0109] Select performance metrics Defined as ,in, This is the indicator data for the first component of the crude extract. This is the indicator data for the second component of the crude extract; the lower limit of the selectivity threshold is... The purity of the recovered solvent is Its allowable range is The recovery load is Its upper limit is Inventory closure error is Its upper limit is The integral terms for each violation are given in the following normalized form:

[0110]

[0111]

[0112]

[0113]

[0114] The rules for determining the weighting coefficients are as follows: A set of basic weights is preset. And write it into the constraint set configuration, for example When a violation integral term is not included in the calculation of this strategy or the constraint is not enabled, its corresponding weight is reset to 0, and the remaining weights are normalized.

[0115]

[0116] in This is the set of constraints enabled for this strategy, ensuring that the sum of weights remains 1 even when the number of constraint outbound quantities changes.

[0117] It should be noted that this embodiment will describe in detail the methods and data sources for obtaining the total amount of solvent input, total amount of solvent recovered, and changes in inventory, as follows:

[0118] To ensure that the calculation of inventory closure error has a feasible data source, this embodiment uses the total amount of solvent input. Total amount recovered With inventory change The acquisition method is specified as follows. Total amount of solvent added. The result is obtained by integrating the flow rate data from the solvent input branch flow meter within the statistical window:

[0119]

[0120] in, Instantaneous flow rate of solvent entering the branch. Total recovered amount. The flow rate data from the solvent recovery reflux or discharge branch flow meter is integrated within the statistical window to obtain the following:

[0121]

[0122] in, Instantaneous flow rate of the solvent recovery branch. Inventory change. It is calculated from the difference in readings of the level gauge, mass gauge, or weighing scale of the recycled solvent storage tank or solvent mother tank at the start and end times of the statistical window:

[0123]

[0124] in, and These are the equivalent volume or equivalent mass conversion values ​​of the inventory at the start and end times of the statistical window; when using a level gauge combined with a tank calibration curve, the level is converted to volume; when using a mass gauge, the volume is converted according to density or uniformly calculated using mass caliber; to reduce measurement noise, the edge execution and monitoring units have 300 pairs of... and Take the average value of a short window of 10s to 60s around each time point.

[0125] It should be noted that this embodiment will describe the policy evidence package fields, source, and edge verification rules in detail, as follows:

[0126] To ensure verifiability before issuance and traceability during execution, this embodiment uses a policy evidence package to solidify the policy's source, data window, threshold version, and verification conclusion. The policy evidence package contains at least the fields shown in Table 1, and the evidence package verification module 330 of the edge execution and monitoring unit 300 verifies each item according to the rules shown in Table 1.

[0127] Policy Evidence Package Fields, Sources, and Edge Validation Rules Table 1

[0128]

[0129] It should be noted that this embodiment will provide a detailed description of runtime consistency monitoring, rollback handling, version stack, and audit logs:

[0130] See Figure 3 The runtime monitoring module 350 performs consistency monitoring based on a set of invariants. This set includes at least three invariants: selective consistency invariants, recovered purity consistency invariants, and solvent inventory consistency invariants. Selective consistency invariants include at least a selectivity index not less than the "lower limit of the selectivity threshold" and a selectivity change rate not greater than the "upper limit of the selectivity change rate." Recovered purity consistency invariants include at least recovered solvent purity within the range defined by the "lower limit of recovered solvent purity" and the "upper limit of recovered solvent purity," and recovered load data not exceeding the "upper limit of recovered load." Solvent inventory consistency invariants include at least inventory closure error not exceeding the "upper limit of inventory closure error." The runtime monitoring module 350 outputs monitoring conclusions at sampling intervals and binds these conclusions to the strategy sequence number, writing them to the audit log.

[0131] When any invariant is not satisfied, the runtime monitoring module 350 outputs a monitoring failure event and triggers the rollback module 360 ​​to perform rollback processing. The rollback module 360 ​​locks the stage sequence to the stage sequence specified by the preset security policy and applies a limit to the solvent ratio setting value. The limit adopts a restriction that the change in the current solvent ratio does not exceed 2% to 5%, and at the same time limits the flow rate setting value to 0.7 to 0.9 of the rated flow rate. Only after the purity of the recovered solvent and the inventory closure error are restored to the allowable range and the "guard hold time" is continuously maintained, is it allowed to reapply for a new policy evidence package. The edge execution and monitoring unit 300 maintains the policy version stack, which contains at least the current policy version and the previous policy version, and writes the rollback identifier, the invariant type triggered by the rollback, the trigger time, and the recovery time into the audit log.

[0132] It should be noted that, for ease of understanding, this embodiment will provide specific numerical examples for 8 hours: stage sequence, segmented set values, and collection window, as follows:

[0133] Below is a set of specific numerical examples for an 8-hour operation. Assume the start time of this batch is 0 minutes and the end time is 480 minutes. The cloud-based twin optimization platform 400 generates a strategy evidence package with the "input data window length" as its length and distributes it. The edge execution and monitoring unit 300 executes it after verification. Examples of stage sequences and segmentation settings are shown in Table 2, and examples of collection windows are shown in Table 3.

[0134] 8-hour phase sequence and segmented setting example table 2

[0135]

[0136] Sample Table 3 for 8-hour collection window

[0137]

[0138] When short-term noise occurs at the boundary of the candidate window, causing the threshold condition to be discontinuously met, the edge execution and monitoring unit 300 delays the opening of the window or closes it early, and writes the window adjustment record to the audit log to ensure that the threshold condition and the minimum collection window duration are met within the window.

[0139] It should be noted that this embodiment will provide a detailed description of the edge-side execution and auditing event examples, as follows:

[0140] Table 4 provides examples of instruction arrangement and key events when the edge side executes the strategy in Table 2.

[0141] Example Table 4 of Edge-side Execution and Audit Events

[0142]

[0143] It should be noted that this embodiment will describe the sample numbering rules, detection conditions, and peak identification integration rules in detail, as follows:

[0144] To link the medicinal material extraction results with the strategy execution chain, this embodiment stipulates the following rules for sample numbering, detection conditions, and peak identification integration.

[0145] Sample retention numbering rules: During each collection window, samples are taken from the collection pipeline or collection tank at fixed intervals and mixed into a window sample. The window sample is bound to the strategy sequence number and the sampling time interval is recorded to form a sample retention record. The sample retention number adopts a combination code of "batch number - strategy sequence number - window number" and the sample retention number is written into the audit log and sample label. The sample retention number and associated fields are shown in Table 5.

[0146] Sample Number and Related Fields Example Table 5

[0147]

[0148] Example of detection conditions: To generate a fingerprint chromatogram of the medicinal extract, this embodiment uses high-performance liquid chromatography (HPLC) to detect the retained sample. The retained sample is either the liquid phase sample after solid-liquid separation step 20 or a combined sample from the collection windows, preferably a combined sample from collection windows W1, W2, and W3. The combined samples are numbered according to "batch number - strategy serial number - window number" and then processed using 0.22... After filtration, the sample was diluted at a fixed ratio and used as the injection sample; the corresponding detection conditions are shown in Table 6.

[0149] Example Table 6 of Fingerprint Spectrum Detection Conditions

[0150]

[0151] Peak identification and integration rules: For each retained sample, baseline correction is performed first, and then the peak is integrated at the preset retention time center. , Peaks are identified and integrated within the nearby retention time window, with the retention time window half-width. The retention time is 0.2 min to 1.0 min; when the peak retention time falls within... Furthermore, when the peak shape satisfies the minimum peak width and minimum peak height thresholds, the area integral value of the peak is defined as the indicator data of the first component of the crude extract. When the peak retention time falls into Furthermore, when the peak shape meets the threshold, the area integral value of the peak is defined as the indicator data of the second component of the crude extract. In one implementation, the peak area integral is obtained by integrating the discrete sampling points of the "baseline-to-peak curve" using the trapezoidal integration method. The peak shape threshold and baseline correction rules are written into the detection configuration table, and the version number is recorded in the audit log. See also Figure 7 Examples are given and Indicator peak location marking; see Figure 8 The example provides a schematic diagram of the results for "percentage of effective components / percentage of impurities" in the fraction samples corresponding to collection windows W1, W2, and W3. The percentage of effective components can be determined by... The relative total response normalization or the result of content determination normalization can be used to determine the impurity indicator percentage. The relative total response is normalized or obtained by normalizing the impurity peak content.

[0152] It should be noted that, for ease of understanding, this embodiment will provide a comparison of runtime statistics and an example of parameter configuration, as follows:

[0153] To illustrate the differences in operational records under the same statistical caliber for the two operating modes, this embodiment summarizes and statistically analyzes the operational logs of three batches, each lasting 8 hours, to obtain operational statistical comparison results. These results are used to present the recording status of constraint events, window events, the proportion of recovered solvent purity ranges, inventory closure events, and disposal time under the two operating modes. The operational statistical comparison results are shown in Table 7. To illustrate the threshold and time settings used in the above statistics and curve plotting, this embodiment provides a set of preferred threshold and preferred time parameter configuration examples. These configuration examples are used to limit the judgment criteria for sampling, alignment, stability criteria, window gating, guard maintenance, stage dwell time, and inventory closure calculation. Parameter configuration examples are shown in Table 8.

[0154] Table 7 shows the statistical comparison results of the operation (mean of 3 batches).

[0155]

[0156] Example Table 8 of Optimized Threshold and Optimized Time Parameter Configuration

[0157]

[0158] It should be noted that, based on the statistical caliber of the three batches of operation logs in this embodiment, each batch lasting 8 hours, Table 7 presents the comparison results between the "Comparison Scheme (Symmetric Analysis / Optimization Deployment Only)" and the "Scheme of this Embodiment (Evidence Package Gating + Access Verification + Monitoring Rollback)" in terms of constraint events, window events, proportion of recovered solvent purity range, inventory closure events, and disposal time; Table 8 provides examples of the threshold and time parameter caliber used to form the above statistics and curves, and Figure 4 to Figure 6 Examples of time series curves corresponding to selectivity indicators, recovered solvent purity, and inventory closure error are provided in Table 7. As shown in the statistical results, under the same statistical caliber, the recorded values ​​for indicators such as the number of single-batch constraint violations, the number of inventory closure error exceeding limits, and the average recovery time for anomaly handling in this embodiment are lower than those in the comparative embodiment. Simultaneously, the recorded values ​​for the collection window effectiveness rate and the recovered solvent purity qualification rate in this embodiment are higher than those in the comparative embodiment, thus making it easier for the fraction samples retained within the collection window to meet the preset threshold caliber. Combined with… Figure 7 fingerprint pattern diagram and Figure 8 The illustration shows the collection results. In this embodiment, by binding the sample retention number with the strategy sequence number and window number and solidifying the peak identification integration rules, the indicator peak data of the medicinal material extraction results and the process control events form a consistent association record in the audit link, which facilitates the corresponding analysis and traceability of the segmented collection windows and the target effective component indicator peak responses in the extraction process of different batches of medicinal materials.

[0159] It should be noted that the embodiments of the present invention have better implementability and are not intended to limit the present invention in any way. Any person skilled in the art may use the above-disclosed technical content to change or modify it into equivalent effective embodiments. However, any modifications or equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of the technical solution of the present invention.

Claims

1. An extraction and separation system for extracting medicinal materials, characterized in that, include: The controlled object interface layer is used to exchange data with the solvent extraction section, solid-liquid separation section, and solvent recovery section and to issue control commands. An industrial IoT acquisition unit is connected to the interface layer of the controlled object and is used to acquire process data. The industrial IoT acquisition unit includes an online detection component. The process data includes at least solvent ratio, flow rate, temperature, pressure, crude extract first component indicator, crude extract second component indicator, and recovered solvent purity. The crude extract first component indicator, crude extract second component indicator, and recovered solvent purity are output by the online detection component. The edge execution and monitoring unit communicates with the industrial IoT acquisition unit and is electrically connected to the interface layer of the controlled object, and is used to generate process status data streams and upload them to the cloud. The cloud-based twin optimization platform is communicatively connected to the edge execution and monitoring unit. The cloud-based twin optimization platform includes an extraction-recovery coupled hybrid automaton digital twin model library, an identifiability determination module, an online identification module, a hybrid integer two-layer optimization module, a counterfactual access module, and an accessibility verification module. The extraction-recovery coupled hybrid automaton digital twin model is used to characterize the evolution of discrete process stages and continuous state variables. The discrete process stages include at least a countercurrent extraction stage and a displacement rinsing stage. The extraction-recovery coupled hybrid automaton digital twin model includes stage switching guard conditions to limit the stage switching trigger conditions. The identifiability determination module is used to trigger the online identification module to update the digital twin model of the extraction-recovery coupled hybrid automaton and generate a model version identifier when the identifiability condition is met; The mixed integer bilayer optimization module jointly solves the stage sequence, segmented solvent ratio and segmented collection window under the model corresponding to the model version identifier, and satisfies the selectivity constraint, recovered solvent purity constraint, solvent inventory consistency constraint and stage switching reachability constraint defined by the first component indicator and the second component indicator of the crude extract. The counterfactual access module and the reachability verification module generate a strategy evidence package based on the solution results; After receiving and verifying the policy evidence package, the edge execution and monitoring unit executes control instructions and performs runtime consistency monitoring based on the selective consistency invariant, the recovery purity consistency invariant, and the solvent inventory consistency invariant. If the monitoring fails, it rolls back to the preset security policy.

2. The extraction and separation system for medicinal material extraction according to claim 1, characterized in that, The online detection component includes a first wavelength channel and a second wavelength channel of an ultraviolet detector and / or two target peak area channels of an online chromatographic detector, and the selectivity constraint is jointly defined by the first component indicator of the crude extract and the second component indicator of the crude extract.

3. The extraction and separation system for medicinal material extraction according to claim 1, characterized in that, The identifiability determination module calculates the parameter sensitivity information of the target parameter set based on the process state data stream and determines whether the identifiability condition is met. When the identifiability condition is not met, the cloud twin optimization platform sends an identification incentive request to the edge execution and monitoring unit, and the edge execution and monitoring unit executes the identification incentive request without violating the recycled solvent purity constraint and the solvent inventory consistency constraint.

4. The extraction and separation system for medicinal material extraction according to claim 3, characterized in that, The identification stimulus request includes a solvent ratio step sequence and / or a flow rate step sequence, and the edge execution and monitoring unit limits the execution of the solvent ratio step sequence and / or the flow rate step sequence to the displacement rinsing phase and writes the stimulus identifier into the process status data stream.

5. The extraction and separation system for medicinal material extraction according to claim 1, characterized in that, The online identification module updates the parameter set in groups according to discrete process stages, including at least: updating the first parameter group related to extraction mass transfer in the countercurrent extraction stage, and updating the second parameter group related to recovery load in the displacement rinsing stage, and generating a group parameter version number respectively.

6. The extraction and separation system for medicinal material extraction according to claim 1, characterized in that, The purity constraints of the recovered solvent include a purity range constraint and a recovery load upper limit constraint, and the mixed integer two-layer optimization module satisfies both the purity range constraint and the recovery load upper limit constraint when generating the segmented solvent ratio.

7. The extraction and separation system for extracting medicinal materials according to claim 1, characterized in that, The solvent inventory consistency constraint is jointly defined by the solvent ratio, the flow rate, and the purity of the recovered solvent, and the edge execution and monitoring unit performs online consistency verification on the process status data stream based on the solvent inventory consistency constraint.

8. The extraction and separation system for medicinal material extraction according to claim 1, characterized in that, The edge execution and monitoring unit maintains a policy version stack, which includes at least the current policy version and the previous policy version. The rollback to the preset security policy includes locking the stage sequence to the stage sequence specified by the preset security policy and applying a limit to the segmented solvent ratio.

9. The extraction and separation system for medicinal material extraction according to claim 1, characterized in that, The edge execution and monitoring unit includes a disconnection fault tolerance module. When communication with the cloud twin optimization platform is interrupted, the disconnection fault tolerance module caches the process status data stream and retransmits it in the order of event time after communication is restored. During the communication interruption, only the process policy with the most recent valid policy evidence package or the preset security policy is allowed to be executed.

10. The extraction and separation system for medicinal material extraction according to claim 1, characterized in that, The edge execution and monitoring unit includes an idempotent execution module, which writes a policy sequence number for the control instruction and performs a duplicate issuance determination. The edge execution and monitoring unit also associates the policy sequence number with the policy evidence package and writes it into the audit log.

Images