Solvent reuse release gating and abnormity isolation control method and device and storage medium

Abstract

The invention relates to release gating and exception isolation control in a solvent recycling process. The system collects at least two online parameters of the pretreatment section, the membrane separation / contact section and the discharge section, and calculates SPC / EWMA and other memory type statistics; and meanwhile, setting two-level thresholds of a process threshold and a release threshold for each upper limit type index and each lower limit type index. And when any one index exceeds a process threshold or the memory type statistic exceeds a control limit, the driving valve group is switched to a small tank buffer backflow branch connected with a discharging device in parallel, CIP cleaning is started in an interlocking manner, and a release port is locked. After it is confirmed through retesting that all the indexes meet the release threshold and the memory type statistic returns to the control limit and is kept for a preset time, locking is relieved, release is recovered, and meanwhile a release record with a timestamp and batch tracing data are generated. And after cleaning and resetting, a plurality of subsequent batches are set as observation batches, and a tightened control limit is adopted for judgment. The invention further provides a control device for executing the method and a computer readable storage medium.

Description

Technical Field

[0001] This invention belongs to the field of process control and quality release technology, specifically involving a control method, device and storage medium that uses memory statistics (SPC / EWMA) and dual threshold gating in a solvent recycling system, and interlocks it with valve group / tank topology, cleaning in place (CIP) and traceability records. Background Technology

[0002] Continuous or quasi-batch solvent recycling processes typically require quality release and operational stability under multiple parameter constraints. Typical monitored parameters include online moisture, particulate matter, total organic carbon (TOC), temperature, pressure, and flow rate. The common engineering practice of using a "single threshold comparison + manual release" approach can meet basic control requirements when disturbances exhibit abrupt changes. However, when deviations originate from factors such as membrane fouling, feedstock fluctuations, reflux retention, or sampling delays and exhibit a slow shift, timely identification and early intervention are often difficult, easily leading to problems such as delayed release criteria, erroneous releases, or overly conservative shutdowns.

[0003] To improve sensitivity to minute drifts, some systems add buffer zones or introduce statistical process methods (such as SPC control charts and EWMA) in addition to single thresholds. However, when the statistics on the detection side and the equipment topology on the execution side are not coordinated—for example, the trigger / termination conditions of the three-way valve assembly, the small tank buffer reflux branch, and the Clean in Place (CIP) are not programmed—even if an "abnormal trend" is detected, it is difficult to translate it into physical isolation and effective reset of the suspected material, leading to consequences such as recontamination and slow reset.

[0004] Meanwhile, many devices lack a state machine that can be uniformly scheduled at the DCS / PLC level and cross-device interlocks: for scenarios of "minor overruns but not reaching the prohibition level for release," there is often a lack of a flexible path of "first switching back flow - temporary storage - retesting - re-judgment"; the start, duration, and end conditions of CIP lack consistent procedural coupling with the valve / pump status, prolonging recovery time and increasing the risk of residual contamination. In addition, release records and batch traceability mostly rely on manual registration, and key elements such as timestamps, valve / pump status, and parameter snapshots are inconsistent, making post-event review and compliance difficult. In summary, although existing technologies have attempted to introduce statistical criteria or buffer zones, due to the lack of integrated linkage control between valve group / tank topology and CIP, it is still difficult to simultaneously achieve the overall goals of early anomaly identification, suspicious batch isolation, rapid cleaning and reset, and traceable release. Summary of the Invention

[0005] This invention addresses continuous or quasi-batch operation scenarios involving solvent recycling. It tackles pain points such as the difficulty in timely detection of small-scale drift accumulation over long periods, the challenge of physically isolating suspicious materials from the main line, the disconnect between cleaning / resetting and monitoring criteria, and inconsistent release and traceability standards. The invention proposes a comprehensive technical solution that forms a closed-loop control layer across six stages: detection, judgment, isolation, resetting, release, and traceability.

[0006] On the detection side, the system collects at least two types of online parameters (which can be upper or lower limit indicators) at a preset sampling period and calculates memory-type statistics at the control layer, including SPC control chart statistics and / or EWMA statistics, thereby distinguishing between short-term fluctuations and long-term deviations. On the decision side, a dual threshold system consisting of a process threshold and a release threshold is established for each indicator, and it is used in conjunction with statistical control limits for judgment: when the measured value exceeds the process threshold, or the statistical value exceeds its control limit, an abnormal trend is identified. This judgment result is not only used to trigger an alarm, but also directly drives the execution side action—the three-way valve group switches to the small tank return branch connected in parallel with the discharge, the release path is locked, and CIP is started according to a preset sequence. By programmatically coupling the detection signal, valve group topology, and CIP, suspicious materials are transferred in real time and isolated from the main line, and the subsequent cleaning and retesting conditions are also consistent with the judgment model.

[0007] Once CIP is completed and retesting confirms that all indicators have reached the release threshold, and the statistics return to control and remain stable for a preset time, the system automatically releases the lockout and resumes release. To prevent rebounds caused by short-term residues or baseline drift after cleaning, this invention sets up an observation batch during the recovery phase: within a limited batch, control limits are tightened or memory weights are increased, and only if there are no limit-crossing events during the observation period does it exit into regular release mode. The entire process is automatically archived using fields such as timestamp, batch number, valve / pump status, online parameter snapshot, electronic signature, and traceability number, forming an auditable traceability chain.

[0008] To adapt to long-term drift caused by seasonal changes, equipment replacement, or operating condition switching, this invention further provides a baseline reassessment process to robustly update the mean and variance and verify the out-of-limit rate (ARL). Simultaneously, it provides an anomaly level-action matrix to unify handling strategies under different trigger intensities (e.g., short-range / standard / enhanced CIP, observation batch number, and control limit tightening magnitude), ensuring consistent interlocking behavior across equipment. The above method can be deployed on a PLC / DCS or industrial PC architecture. The control device consists of a processor, memory, and I / O interfaces. The I / O interfaces are electrically connected to online sensors, valve groups, pumps, human-machine interfaces, and data storage. The program stored in the storage medium completes state machine coordination between functional modules such as data acquisition, statistical judgment, interlocking execution, release, and archiving during processor execution. This invention focuses on the implementation of control layer logic and data archiving, without limiting specific membrane components, chemical formulations, or interface appearance.

[0009] Beneficial effects By programmatically coupling memory-based statistical judgment with valve assembly / tank topology and CIP, this invention integrates "anomaly detection" with "how to handle it": First, memory-based statistics are more sensitive to gradual shifts, and when combined with dual-threshold gating, they can issue executable decision signals in advance, suppressing false releases and shortening detection delays under the same noise conditions. Second, valve group switching and small tank reflux achieve immediate physical isolation of suspected materials. The start, duration and end conditions of CIP are aligned with the judgment model, avoiding the fragmented process of "alarm - manual - line stop - re-evaluation", which helps to reduce the risk of recontamination and mixing. Third, through the dual constraints of "retesting and passing + holding time" and the tightening of the judgment of observation batches, the system can restore stable production more quickly while ensuring quality; Fourth, standardized fields are used to automatically archive release and event logs, forming a read-only and consistent traceability evidence chain to support compliance audits and post-event reviews. The accompanying baseline reassessment and anomaly level matrix ensure that the solution maintains controllable over-limit rates and downtime during long-term operation and multi-condition switching, and has good engineering portability and large-scale deployment capabilities.

[0010] In summary, this invention is not simply an algorithm superposition, but rather constructs a closed and self-consistent control loop that integrates detection, judgment, isolation, reset, release, and traceability. It takes into account the technical objectives of early identification, isolation, fast reset, and traceability, and is applicable to various solvent reuse and similar multi-parameter release scenarios. Attached Figure Description

[0011] Figure 1 The control system framework and I / O point diagram are as follows: Controller 201 is communicatively connected to the human-machine interface 207 and data storage 208; Controller 201 is electrically connected to the online sensor 203, the three-way valve group 204, and the pump 206 via the I / O interface 202. The three-way valve group 204 switches between the main discharge outlet 209 and the parallel small tank 205 (return branch); Pump 206 is used for material conveying and drives CIP cleaning 310.

[0012] Figure 2 The event / state machine diagram shows: Normal 501, Release 502, Abnormal Isolation 503, CIP Reset 504 and Observation Batch 505. Arrows indicate state transitions and conditions.

[0013] Figure 3 Here is a schematic diagram of the HMI interface layout: Threshold / Control Limit Configuration Area 511, Trend Curve Display Area 512, Alarm / Event Log and Traceability Export Area 513.

[0014] Figure 4 This is a schematic diagram of dual-threshold gating and EWMA timeline: showing the changes in measured values ​​and statistics, as well as the trigger and recovery nodes corresponding to the process threshold and release threshold.

[0015] Figure 5 is a schematic diagram of the release record and traceability data structure: the release record data structure 530 includes a timestamp field 536, a batch number field 535, a valve position / pump position field 534, an online parameter snapshot field 533, an electronic signature field 532, and a traceability number field 531, and is archived to the data storage 208.

[0016] Reference Table of Figure Labels Detailed Implementation

[0017] Terminology and Objects "Process threshold" is used to trigger abnormal paths in advance; "Release threshold" is the final quality release threshold; "Observation batch" is the transition batch after CIP reset and uses tightened control limits; "Upper limit indicators" are unfavorable when their index values ​​increase (such as moisture, particle size, TOC), while the lower limit indicators are the opposite. Monitored parameters include, but are not limited to, at least two of the following: online moisture, particle size, TOC, temperature, pressure, and flow rate.

[0018] (I) System Structure and Signal Mapping like Figure 1 As shown, the controller 201 is electrically connected to the online sensor 203, valve assembly 204, pump 206, and HMI 207 via I / O interface 202. Data storage 208 is used for archiving release records and event logs. Valve assembly 204 switches between the main discharge line and the parallel return branch of the small tank 205; pump 206 provides power for conveying and CIP. The system is compatible with continuous and semi-batch topologies.

[0019] (ii) Decision and Interlocking Logic See Figure 2 and Figure 4 Statistics and control limits: Collect monitored parameters Xt with sampling period Ts, calculate SPC and / or EWMA statistics Zt, set process thresholds and release thresholds for upper limit type / lower limit type indicators respectively, and the statistics and measured values ​​jointly participate in the gating judgment; the priority of abnormal level and the handling actions are executed according to Table 3; when long-term drift or operation condition change triggers baseline re-evaluation, check and switch according to the process in Table 4.

[0020] Processing monitored parameters In this embodiment, the monitored parameter Xt is smoothed and trend determined using the exponentially weighted moving average (EWMA) statistic Zt. Let 0 < λ ≤ 1 be the smoothing coefficient, then Zt is calculated recursively using the following formula: , t≥1, where Z0 can be the average value of historical steady-state periods or the first online measurement. A smaller λ makes Zt more sensitive to historical data, which is beneficial for identifying slow drifts; a larger λ makes Zt respond faster to new data.

[0021] (III) Dual Threshold Gating Strategy To avoid frequent switching caused by a single threshold, this embodiment adopts a dual-threshold gating strategy: a process threshold Th_proc is set during the online monitoring phase to trigger anomaly isolation and reflow path; a release threshold Th_rel is set during the recovery and release phase to determine whether the process has stabilized to a release level. The typical rule is: when Xt or Zt exceeds the process threshold Th_proc, the system determines that the batch has an anomaly risk, triggers valve group reflow, and performs CIP cleaning as needed; after anomaly handling, the system only allows switching from the observation batch state back to the release state if Xt and Zt do not exceed the release threshold Th_rel within L consecutive sampling periods, and Zt shows no continuous deviation trend.

[0022] Abnormal isolation: When any indicator exceeds the process threshold but does not exceed the release threshold, or when the statistic exceeds the corresponding control limit, the drive valve group 204 switches to the return branch of the small tank 205 and starts CIP, and the release port 209 is locked; if the release threshold is exceeded, release is prohibited and the return and CIP interlock is maintained. CIP and Reset: CIP adopts a two-stage process (homogeneous cleaning solvent, polar cleaning solvent), and ends with rinsing with ultrapure water (UPW); CIP is terminated after the retest meets the qualification conditions; Observation batch: After CIP, the first few batches are set as observation batches, and the control limits are tightened and / or the weighted statistics are enhanced to verify stability; the number of observation batches can be 1 to 3 batches, and the control limits can be tightened by about 2% to 10% relative to the normal release, which is 5% in this embodiment; when the statistics and measured values ​​of N consecutive batches (N is 1 to 3) do not exceed the limits, the normal release is automatically restored; Release and Archiving: Release is granted after all indicators meet the release threshold and the statistics remain stable within the control limits for a preset time Th, and a read-only release record and event log are generated in data storage 208; Anomaly Handling Level Matrix: For ease of project implementation, the correspondence between anomaly levels and handling actions is shown in Table 3.

[0023] (iv) Examples of Statistical Control Limit Calculation and Parameter Examples SPC: μ and σ are estimated based on historical steady-state baseline data, with control limits set at μ ± 3σ; EWMA: Z0 is taken as the steady-state mean or the first measured value; control limits are set accordingly. .

[0024] This manual uses the following standardized notation: measured values ​​are denoted as Xt, EWMA statistics as Zt, smoothing factor λ, release threshold L, and holding time Th; the "threshold / control limit" is uniformly configured in "threshold / control limit configuration area 511" of HMI 207. The smoothing factor λ can be 0.1 to 0.3, preferably around 0.2. Units are mg·kg. -1 mg·L -1 μm, particles·mL -1 The above symbols and units have the same meaning throughout the text and in the accompanying figures.

[0025] Hold time Th: preferably 2 to 10 sampling periods, to reduce jitter misjudgment; Baseline reassessment: When long-term drift, formula switching or replacement of critical equipment leads to μ and σ mismatch, the baseline reassessment process in "Other Implementation Methods" shall be performed (see Table 4).

[0026] (v) Human-Machine Interface (HMI) and Parameter Window See Figure 3 The HMI 207 provides threshold / control limit configuration, trend display, alarm and event logs, electronic signatures, and batch traceability export. Parameter window (fixed example): Release threshold (upper limit): Moisture ≤ 300 mg·kg -1 Particle size (≥0.05μm) ≤1.0 particles / mL -1 TOC ≤ 10 mg·L -1 Process threshold (10% tightening): Moisture 270 mg·kg -1 0.9 particles per mL -1 TOC 9 mg·L -1 Statistical process parameters: EWMA λ=0.2; SPC at 3σ; observation batch quantity 3. The field mappings of the alarm / event log and traceability export area 513 correspond one-to-one with Table 1 for auditing and review. The above parameter window values ​​are an exemplary configuration of one embodiment, which can be adjusted and fixed according to the on-site SOP, and do not constitute a limitation on the claims.

[0027] like Figure 3As shown, the human-machine interface 207 includes at least a threshold / control limit configuration area 511, a trend curve display area 512, and an alarm, event log, and traceability export area 513. The threshold / control limit configuration area 511 is used to set process thresholds, release thresholds, and parameters of the EWMA algorithm (e.g., smoothing factor λ, observation batch length N, alarm threshold Th, and release threshold L, etc.); the trend curve display area 512 is used to display the real-time changing trends of the monitored parameter Xt and the EWMA statistic Zt over time; the alarm, event log, and traceability export area 513 is used to display alarm and event logs, record electronic signatures of operations, and provide the function of exporting traceability records by batch number or traceability number. The field mapping relationships can be given in the relevant field description table.

[0028] Example 1: Parameterization and Reproducibility of Normal Operation See Figures 1 to 5 The sampling period Ts = 1 min; a baseline is established with M = 120 sampling points to obtain the mean moisture content. Standard deviation Choose λ=0.2, L=3.0, and set the EWMA control limits to: That is, UCL≈240, LCL≈220. During the 24-hour operation, Xt fluctuated between 210 and 240, while Zt remained within the control limits; the log showed that the "process threshold 270 mg·kg" was not triggered. -1 The state machine cycles through "normal - release", and the release records are archived according to the fields in Table 1.

[0029] Example 2: Disturbance and Abnormal Paths, Reset and Observation Batch See Figure 2 and Figure 4 At t0, the feed moisture content rises to 600–800 mg·kg⁻¹. -1 At t0+12 min, Zt crosses the UCL and Xt > 270 mg·kg -1 (Process threshold); The controller issues an interlock: V204 switches to the reflux position, small tank 205 receives material, and the CIP sequence is started; CIP adopts a "two-stage and ultrapure water (UPW) rinsing" process, for example, circulating the same cleaning solvent for 30 minutes, circulating the polar cleaning solvent for 20 minutes, and rinsing with ultrapure water (UPW) for 10 minutes; After the retest is qualified, CIP is released and 3 batches are entered into the observation batch. During the observation period, the control limit is tightened by 5% (or L=2.8). After all 3 batches do not exceed the limit, normal release is restored.

[0030] Example 3: Deviation of Lower Limit Indicator: Insufficient Flow Triggers Backflow and Intervention The circulation flow rate Q was set as a lower limit indicator, with the release threshold Q ≥ Qrel and the process threshold Q ≥ Qrel - ΔQ (ΔQ being a margin of 5% to 15%). During a certain operation, an increase in flow resistance after membrane replacement at time t1 caused Q to drop below Qrel - ΔQ and EWMA to fall below LCL. The system maintained release lockout and switched backflow, performing a short-path CIP (50% of the above process duration). After retesting and confirming that Q had recovered to Qrel and that moisture / particles / TOC all met the release threshold and remained stable at Th = 10 min, release was resumed.

[0031] Example 4: Bypass Sampling Verification and Lockout Release In a situation where the EWMA is close to the UCL but the measured value is still below the process threshold, the system triggers the "bypass sampling" path: the sampling valve collects a bypass sample and sends it to the laboratory KF / TOC for verification; during the verification, the release port 209 remains locked; when the verification result is qualified and the online statistic returns to the control limit and maintains Th, the system automatically releases the lock and releases the sample; if the verification is unqualified, the reflux is maintained and the sample enters the CIP.

[0032] Example 5: Conservative Strategies for Key Sensor Failure When the moisture sensor experiences communication loss at time t2 (no new data for two consecutive sampling periods) or the reading remains unchanged (variance ≈ 0), the system enters conservative mode: it closes the release and switches backflow; it issues a "calibration request" prompt and prohibits release until the two-point calibration and comparison are completed; after calibration, it is necessary to verify that the statistic and the measured value are within the control limits and thresholds within Th before the release can be restored.

[0033] Example 6: Actuator Not Reached and CIP Interrupt Handling If V204 fails to reach the set value (e.g., exceeding the limit by 2 seconds) after the command is issued, the system maintains the release interlock and cuts off the pumping to the release branch, records the failure event, and triggers the redundant valve V204B (if configured). If any valve / pump fails to reach the set speed or the liquid level / temperature fails to reach the set value during the CIP phase, the current phase is interrupted and the system returns to the safe state to avoid recontamination caused by continuing cleaning under unmet conditions. After recovery, the corresponding phase needs to be timed again and a complete closed-loop record needs to be formed in the log.

[0034] Safety and Abnormal Handling When the critical sensor 203 fails or its data is abnormal (out of bounds, stationary, communication loss), a conservative strategy is implemented: the release port 209 is locked, backflow is maintained, and an alarm is triggered; the lockout is released after calibration and retesting. Release is prohibited if valve group 204 or pump 206 has not reached its position / speed; release is maintained while the lockout is active during bypass sampling verification.

[0035] Devices and storage media The controller 201 can be implemented using a programmable logic controller (PLC) or an industrial PC. The storage medium contains a program for executing the control method. The I / O interface 202 is electrically connected to the online sensor 203, the three-way valve group 204, the pump 206, and the human-machine interface (HMI) 207, respectively, to run a state machine and cross-device interlocking logic, completing dual threshold judgment, valve position switching, CIP triggering, retesting, and release archiving. The program modules of the control device correspond to the method steps S1 to S6, including a data acquisition module (S1), a statistics and judgment module (S2 to S3), an interlocking execution module (S4), a release and archiving module (S5), and an observation batch management module (S6). When the processor executes the program stored in the storage medium, it realizes the coordinated control between the above modules.

[0036] Release records and event logs are written in read-only mode, and traceability numbers and electronic signatures support an immutable audit chain, ensuring the integrity and consistency of batch traceability.

[0037] Table 1. Example of release record data fields Table 2. Examples of Threshold and Control Limit Settings Other implementation methods 1) Anomaly Level - Action Matrix: To unify the handling strategies under different triggering situations, anomalies are divided into warning (L0), statistical limit violation (L1), exceeding process threshold (L2), and exceeding release threshold (L3). The valve and pump actions, CIP versions, release strategies, observation batches, and archiving requirements corresponding to each level are shown in Table 3; this matrix can be further refined according to the on-site SOP.

[0038] Table 3 Anomaly Level - Action Matrix 2) Baseline Reassessment Process: When long-term drift, formula switching, replacement of critical equipment, or seasonal fluctuations lead to μ and σ mismatch, robust statistics are used for baseline reassessment. The process includes seven steps: triggering, freezing, sampling, estimation, verification, switching, and observation, forming an audit loop in the log, as shown in Table 4.

[0039] Table 4 Baseline Reassessment Process Example of multi-parameter joint judgment (M-of-N): Taking three upper limit indicators {moisture, particle size, TOC} as an example, the decision window is set to W = 5 sampling periods, M = 2, and N = 3. The binary flag bj(t) for indicator j ∈ {1, 2, 3} at the t-th sampling time is defined as follows: if Xt,j > process threshold, or the corresponding EWMA Zt,j exceeds the upper control limit UCLj, then bj(t) = 1; otherwise, it is 0. If the following conditions are met within any sliding window of length W... If this condition is met at two consecutive sampling times, an abnormal path is triggered.

[0040] Numerical example: Moisture process threshold of 270 mg·kg -1 0.9 particles per mL -1 TOC 9 mg·L -1 λ=0.2, L=3.0; within the window from t0 to t0+5 min, if at times t0+3 and t0+4 the following occurs: {Moisture: Xt>270 and Zt exceeds UCL, Particles: Xt≤0.9 but Zt is under control, TOC: Xt>9 or Zt exceeds UCL}, then at least two indicators are set to 1, satisfying 2-of-3. Based on this, the system switches to "small tank reflux + interlock CIP" and determines it as L2 or L1 action according to Table 3.

[0041] Example 7 Performance Evaluation Methods and Statistical Standards To facilitate reproduction and auditing, this embodiment provides a unified definition, calculation method, and statistical caliber for the process control performance indicators involved. Statistics are based on the sampling period Ts and batch window W, with the time of interference recorded as t0; anomaly triggering is based on any triggering condition in step S3 of claim 1.

[0042] Table 5 Statistical Performance Indicators – Definitions and Statistical Scope Table 6. Process Running Events – Statistical Scope Data processing flow: (1) Run or simulate according to the baseline strategy and the strategy of this invention respectively, and record events and timestamps; (2) Calculate the indicators according to the standards in Tables 5 and 6, and report the mean and interval; (3) Provide statistics on ARL1 and MTTD for small-shift scenarios respectively; (4) The above evaluation method is only used for reproducing experiments and auditing, and does not limit the technical solution of the present invention.

[0043] Under the same raw materials and equipment conditions, and based on the aforementioned specifications, this invention helps reduce the false release rate and shorten the detection and recovery time by using memory-type statistics + dual-threshold gating triggering of the side valve group backflow and CIP interlocking; the "retest qualified + hold time" and observation batch mechanism are used to verify recovery stability; and the archiving and auditing links improve traceability completeness. The specific improvement range is subject to subsequent operation / simulation results and does not constitute a limitation of this specification.

[0044] Industrial applicability This solution can be quickly integrated with existing sensors, valve assemblies, and PLC / DCS. Parameter windows and thresholds can be adjusted according to on-site SOPs. It is suitable for solvent reuse and similar multi-parameter quality release scenarios, and facilitates large-scale deployment and auditing.

Claims

1. A control method for release gating and anomaly isolation in a solvent recycling system, characterized in that, include: S1: Collect at least two of the online parameters set in the pretreatment section, membrane separation / contact section, and discharge section; S2: Calculate memory statistics for the online parameters. The memory statistics include statistical process control (SPC) control chart statistics and / or exponentially weighted moving average (EWMA) statistics. In this text, "and / or" means "can be used alone or in combination" and does not constitute a limitation on the specific implementation. S3: Set two levels of thresholds, namely process threshold and release threshold, for upper limit type indicators and lower limit type indicators respectively, and establish gate control judgment rules. When any of the following conditions are met, an anomaly judgment is triggered: i) the measured value exceeds the corresponding process threshold; ii) the memory type statistic exceeds its control limit. S4: When any indicator exceeds the process threshold but does not exceed the release threshold, or when the memory statistic exceeds its control limit, output a control signal to drive the valve group to switch to the small tank buffer return branch connected in parallel with the discharge pipeline, and interlock the start of the cleaning in place (CIP) sequence, while locking the release path pointing to the discharge section. S5: After retesting and confirming that all indicators meet the release threshold and the memory statistics are restored to within the control limit and maintained for a preset time, the release path is unlocked and the release is opened, generating the corresponding batch release record and batch traceability file. S6: After completing the on-site cleaning and reset, the first few batches are set as observation batches. The observation batches are judged using relatively tight control limits. After the preset exit conditions are met, the normal release strategy is restored.

2. The method according to claim 1, characterized in that, The EWMA statistic satisfies: And Z0 is taken as the steady-state mean or the first measurement value; its control limit is , where μZ and σ are derived from baseline data, and L is the threshold coefficient; where λ takes values ​​of 0.1 to 0.3, and L takes values ​​of 2.5 to 3.

5.

3. The method according to claim 1 or 2, characterized in that: For upper limit indicators, the process threshold is tightened by 5% to 15% relative to the release threshold; for lower limit indicators, the process threshold is relaxed by 5% to 15% relative to the release threshold; in engineering applications, the process threshold can be tightened by 5% to 15% relative to the release threshold, and set in combination with noise level and process margin.

4. The method according to any one of claims 1 to 3, characterized in that, The in-situ cleaning includes: sequentially performing a homologous cleaning solvent stage and a polar cleaning solvent stage, and finishing with an ultrapure water (UPW) rinse.

5. The method according to any one of claims 1 to 4, characterized in that: The number of observation batches is 1 to 3, and the tightened control limit is tightened by 2% to 10% relative to the conventional control limit; and the exit condition for the observation batch is that no control limit exceedance occurs within N consecutive batches (N is 1 to 3).

6. The method according to any one of claims 1 to 5, characterized in that, The execution sequence of the abnormal path includes: switching the valve group to the small tank buffer reflux branch, small tank buffer reflux, performing on-site cleaning, retesting, and selecting to release or continue reflux according to the judgment result.

7. The method according to any one of claims 1 to 6, characterized in that: The release record includes at least the following fields: parameter snapshot, valve position status, pump position status, batch number, timestamp, electronic signature, and traceability number, and is archived in read-only mode; the control method supports bypass sampling verification while the release path is locked, and the release path remains locked during the verification; and the online parameters include at least one of trace moisture and particle count.

8. A control device for implementing the method according to any one of claims 1 to 7, characterized in that, include: The device includes a processor, a memory, and an I / O interface, wherein the I / O interface is electrically connected to online sensors, valve groups, pumps, and a human-machine interface; the memory stores a computer program, which, when executed by the processor, is used to perform the method described in any one of claims 1 to 7; the human-machine interface is used to configure process thresholds and release thresholds, display trends of monitored parameters, event logs and release records, and supports batch association export and electronic signatures; the control device also provides one-button CIP interlock control and records the CIP start / end time and related valve and pump status.

9. A computer-readable storage medium having a computer program stored thereon, the program, when executed by a processor, implementing the method of any one of claims 1 to 7.

10. A solvent recycling system for implementing the method according to any one of claims 1 to 7, characterized in that, include: The control device includes online sensors, valve groups, pumps, and a human-machine interface electrically connected to the control unit. The control unit is configured to drive the valve group to switch to the small tank buffer reflux branch connected in parallel with the discharge pipeline and perform CIP interlock control, and to control the discharge release or abnormal reflux based on the gating judgment results of the online parameters and memory statistics.

Images