Chromatograph three-way purging sealing performance test and industrial data acquisition system

Abstract

The invention relates to the technical field of fluid state measurement and equipment leakproofness nondestructive testing, in particular to a chromatographic instrument three-way purging leakproofness testing and industrial data acquisition system which comprises a data acquisition module, a benchmark reconstruction module, a parameter injection module, a residual error extraction module, a coupling judgment module and a feedback control module. The system takes control pulse or valve position switching time as a unified time reference, synchronously aligns pressure, flow and valve driving electrical time sequence data, and reconstructs an ideal process reference waveform under an unbiased assumption; the core of the method is that flow resistance attenuation or viscous damping parameters are injected into a reference model to generate a cleaning failure simulation waveform, the similarity is calculated by comparing the residual error deviation vectors of real data and simulation data relative to an ideal reference, and then control parameters are output; according to the method, the problem of misjudgment caused by sampling offset is solved, a conservative mode is started when a data link is abnormal, the situation that sensor signal loss is misjudged as valve body cleaning failure is effectively avoided, and the integrity and reliability of judgment in the industrial purging process are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of fluid state measurement and non-destructive testing of equipment sealing, specifically a chromatograph three-way purging sealing test and industrial data acquisition system. Background Technology

[0002] In the continuous analysis process of online chromatographs in petrochemical plants, the purging and cleaning effect of the three-way valve directly affects the cleanliness of the pipeline and the stability of the analysis results after the sample gas is switched. Therefore, accurate acquisition and judgment of the purging process is an important foundation for ensuring the reliable operation of the equipment.

[0003] Existing purging monitoring methods have limitations. For example, they often rely on single pressure values, single flow peaks, or simple threshold alarms for judgment. Due to the simultaneous influence of pump pulsation, temperature fluctuations, valve aging, residue adhesion, and asynchronous sampling sequences in industrial settings, it is difficult to fully reflect the true differences in the purging process. Especially when micro-leakage, residue, and normal switching disturbances are superimposed, false alarms and missed alarms are prone to occur, which cannot effectively support subsequent cleaning control and maintenance decisions. More importantly, existing systems lack accurate quantitative measurement and non-destructive testing methods for valve sealing micro-leakage defects and pipeline residual contamination physical states. It is difficult to measure and extract the true physical deviations during dynamic fluid evolution, thus failing to achieve measurement-level accurate state testing and evaluation. Summary of the Invention

[0004] The purpose of this invention is to provide a system for testing the three-way purge seal of a chromatograph and for industrial data acquisition, in order to solve the problems mentioned in the background art. Specifically, the technical solution of this invention is as follows: A three-way purging seal test system for chromatographs and an industrial data acquisition system, including: The data acquisition module is used to collect pressure time-series data, flow time-series data, and valve drive electrical time-series data of the fluid passage during the industrial purging and cleaning process. It is aligned with the control pulse trigger time or valve position switching time as a unified time reference to generate real-time process data that characterizes the state of the purging and cleaning process. The benchmark reconstruction module is communicatively connected to the data acquisition module. It is used to construct an ideal process benchmark model under the assumptions of no leakage influence, no residual influence, and no mechanical delay influence based on the preset purging flow rules and the valve drive electrical timing data, and generate an ideal process benchmark waveform. The parameter injection module is communicatively connected to the benchmark reconstruction module and is used to inject preset bypass flow resistance attenuation parameters and / or nonlinear viscous damping parameters into the ideal process benchmark model to generate process simulation waveforms that characterize the cleaning failure state. The residual extraction module is communicatively connected to the data acquisition module, the benchmark reconstruction module, and the parameter injection module. It is used to generate a real deviation vector reflecting the process deviation based on the real-time process data and the ideal process benchmark waveform, and to generate a theoretical deviation vector based on the process simulation waveform and the ideal process benchmark waveform. The coupling decision module is communicatively connected to the residual extraction module. It is used to calculate the similarity based on the trajectory distance and distribution distance determined by the actual deviation vector and the theoretical deviation vector after dimensionality reduction and feature extraction, and to generate the purging and cleaning process control parameters based on the similarity. The feedback control module is communicatively connected to the coupling decision module and is used to adjust subsequent sampling parameters and / or output cleaning control commands to the industrial upper control layer based on the purging and cleaning process control parameters.

[0005] Preferably, the data acquisition module includes: The pressure acquisition unit is used to acquire transient pressure signals in the purging path that reflect the impact effect of the clean airflow; The flow acquisition unit is used to acquire transient flow signals that reflect the flow capacity of the cleaning medium in the purging path; The electrical acquisition unit is used to acquire valve drive coil current signals and control pulse signals; The timing alignment unit, connected to the pressure acquisition unit, the flow acquisition unit, and the electrical acquisition unit, is used to align the acquired timing signals with the rising edge of the control pulse or the valve position switching trigger moment as a unified time zero point to generate the real-time process data.

[0006] Preferably, the benchmark reconstruction module is used to construct an ideal process benchmark model based on preset purge flow rules and valve drive electrical timing data, under theoretical assumptions of no leakage influence, no residual influence, and no mechanical delay influence. The reference reconstruction module is also used to output the ideal pressure decay waveform, the ideal flow change waveform, and the ideal electrical response waveform corresponding to the valve position switching process, so as to jointly constitute the ideal process reference waveform.

[0007] Preferably, the parameter injection module is used to inject the bypass flow resistance attenuation parameter, which is used to change the equivalent flow resistance of the bypass branch, into the ideal process reference model to simulate the micro-leakage cleaning failure path caused by seal aging. The parameter injection module is also used to inject the nonlinear viscous damping parameter, which is used to change the flow damping characteristics of the main channel, into the ideal process benchmark model to simulate the impact of channel residue on the cleaning and recovery process. The parameter injection module generates the process simulation waveform based on the injected model.

[0008] Preferably, the residual extraction module includes: The real-time deviation generation unit is used to perform differential processing on the real-time process data and the ideal process reference waveform to generate the real-time deviation vector; The theoretical deviation generation unit is used to perform differential processing on the process simulation waveform and the ideal process reference waveform to generate the theoretical deviation vector; The feature compression unit, connected to the real deviation generation unit and the theoretical deviation generation unit, is used to perform dimensionality reduction processing on the real deviation vector and the theoretical deviation vector based on statistical projection or feature selection, so as to generate real deviation feature quantities and theoretical deviation feature quantities for coupling decision, respectively.

[0009] Preferably, the coupling decision module is used to calculate the trajectory distance corresponding to the actual deviation feature and the theoretical deviation feature using a dynamic time warping algorithm; The coupling decision module is further used to calculate the distribution distance corresponding to the actual deviation feature and the theoretical deviation feature using Mahalanobis distance; The coupling decision module normalizes the trajectory distance and the distribution distance, and then performs weighted fusion based on the normalized trajectory distance and distribution distance to generate the similarity.

[0010] Preferably, the coupling decision module is further configured to base its decision on a preset reporting threshold and a preset suppression threshold, and under the condition that the reporting threshold is greater than the suppression threshold: When the similarity is greater than or equal to the reporting threshold, a cleaning anomaly event is generated; When the similarity is less than or equal to the inhibition threshold, a cleanliness compliance event is generated; A clean review event is generated when the similarity is greater than the suppression threshold and less than the reporting threshold.

[0011] Preferably, the feedback control module is further configured to send a purging cleaning anomaly reporting command or a valve body cleaning fault command to the industrial upper control layer when the cleaning anomaly event is generated. The feedback control module is also used to maintain the current collection cycle and suppress abnormal reporting when the cleaning compliance event is generated; The feedback control module is further configured to adjust the sampling frequency, sampling window, or the range of values ​​for the bypass flow resistance attenuation parameter and / or nonlinear viscous damping parameter in the parameter injection module when the cleaning review event is generated, and trigger a re-decision.

[0012] Preferably, the system further includes a data communication module; The data communication module is communicatively connected to the coupling decision module and / or the feedback control module, and is used to upload the purging and cleaning process control parameters, the cleaning abnormal events, and the cleaning control commands. The data communication module includes at least one of an industrial Ethernet interface, a serial communication interface, or a fieldbus interface.

[0013] Preferably, the pressure timing data, the flow timing data, and the valve drive electrical timing data are real-time process data formed from real-time sampling signals during the industrial purging process; The purging and cleaning process control parameters include parameters for characterizing micro-leakage cleaning failure, parameters for characterizing residual cleaning failure, and parameters for characterizing cleaning compliance. The feedback control module outputs valve cleaning and maintenance instructions, purging and enhanced cleaning instructions, or continue operation instructions based on the parameters used to characterize the micro-leakage cleaning failure state, the parameters used to characterize the residual cleaning failure state, or the parameters used to characterize the cleaning compliance state.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. By synchronously acquiring pressure timing data, flow timing data, and valve drive electrical timing data during the industrial purging and cleaning process through the data acquisition module, and aligning them with the control pulse trigger time or valve position switching time as a unified time reference, the valve action, fluid impact, and flow recovery processes can be recorded under the same event chain. This avoids misreading caused by sampling clock offset, sampling asynchrony, or isolated judgment of a single sensor value, and improves the completeness of the representation of the actual purging process. By using the valve position switching trigger time as the zero point when the control pulse is missing to deal with the signal loss problem, the reliability of industrial field data acquisition and subsequent judgment is improved. 2. By constructing an ideal process baseline model based on preset purging flow rules and valve-driven electrical timing data under the assumptions of no leakage, no residual effects, and no mechanical delay effects, and generating ideal pressure decay waveforms, ideal flow change waveforms, and ideal electrical response waveforms, it is possible to first clarify the event structure that should be presented under the healthy state of the equipment, thereby avoiding the historical average waveform from masking abnormal problems due to gradual equipment wear, seasonal temperature drift, and baseline characteristic drift, thus improving the stability and comparability of subsequent deviation interpretations. By reselecting or modifying the ideal template according to the current operating conditions when process parameters change, purging media changes, or the equipment is put into operation for the first time, the ideal baseline can always revolve around the sequential relationships and boundary constraints under the healthy state, thereby improving the adaptability of the baseline reference under different operating conditions. 3. By injecting bypass flow resistance attenuation parameters into the ideal process baseline model through the parameter injection module to simulate the micro-leakage cleaning failure path caused by seal aging, and injecting nonlinear viscous damping parameters to simulate the impact of path residue on the cleaning recovery process, it is possible to deconstruct anomalies that originally only manifested as waveform deviations into micro-leakage and residue failure mechanisms with clear physical characteristics. This solves the problem of difficulty in distinguishing between micro-leakage, residue and normal switching disturbances when they are superimposed, and improves the identification of valve seal aging and path contamination accumulation. By introducing two types of parameters simultaneously to form a composite simulation waveform when necessary, and marking it as an unknown anomaly when it cannot be stably coupled, it is possible to avoid incorrectly applying new cleaning failures to existing templates, thereby improving the engineering prudence of failure classification. Attached Figure Description

[0015] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a schematic diagram of the chromatograph three-way purge sealing test and industrial data acquisition system module; Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0017] Example 1: A three-way purging seal test system for chromatographs and an industrial data acquisition system, including: The data acquisition module is used to collect pressure time-series data, flow time-series data, and valve drive electrical time-series data of the fluid passage during the industrial purging and cleaning process. It is aligned with the control pulse trigger time or valve position switching time as a unified time reference to generate real-time process data that characterizes the state of the purging and cleaning process. The baseline reconstruction module communicates with the data acquisition module and is used to construct an ideal process baseline model under the assumptions of no leakage influence, no residual influence, and no mechanical delay influence based on the preset purging flow rules and valve drive electrical timing data, and generate an ideal process baseline waveform. The parameter injection module, which is connected in communication with the benchmark reconstruction module, is used to inject preset bypass flow resistance attenuation parameters and / or nonlinear viscous damping parameters into the ideal process benchmark model to generate process simulation waveforms that characterize the clean failure state. The residual extraction module communicates with the data acquisition module, the benchmark reconstruction module, and the parameter injection module. It is used to generate a real deviation vector reflecting the process deviation based on real-time process data and ideal process benchmark waveform, and to generate a theoretical deviation vector based on process simulation waveform and ideal process benchmark waveform. The coupling decision module is connected to the residual extraction module. It is used to calculate the similarity based on the trajectory distance and distribution distance determined by the feature quantity extracted after dimensionality reduction of the actual deviation vector and the theoretical deviation vector, and to generate the purging and cleaning process control parameters based on the similarity. The feedback control module, which communicates with the coupling decision module, is used to adjust subsequent sampling parameters and / or output cleaning control commands to the industrial upper-level control layer based on the control parameters of the purging and cleaning process.

[0018] This embodiment provides an implementation mechanism for a three-way gas chromatograph purging and sealing test and an industrial data acquisition system. Specifically, the system is deployed in an online gas chromatograph analysis cabinet in a continuously operating petrochemical plant, switching between three states: process sample gas introduction, purging and cleaning, and standby isolation. Since the equipment is in an industrial environment with pump pulsation, pipeline vibration, temperature fluctuation and valve aging, relying solely on a single pressure value or flow peak can easily misjudge normal switching disturbances as contamination risks and may also lead to missed detection of actual cleaning failures caused by micro-leakage and residue. Therefore, this system does not directly perform isolated threshold judgment on the field waveform, but instead constructs an ideal purging process reference synchronized with valve action, and then injects suspected failure factors into this reference to form simulation anomalies. The field deviation and simulation deviation are coupled and compared to output control parameters corresponding to the actual working conditions. This comparison and calculation process is essentially a non-destructive testing and inspection method based on multi-source physical parameter acquisition and dynamic feature measurement. By testing and measuring the deviation vector in the time distribution dimension, it achieves accurate quantitative characterization of internal hidden physical defects. In the detailed description, the data acquisition module synchronously acquires the pressure timing, flow timing, and valve drive electrical signal in the purging passage each time the three-way valve receives a switching command. Among them, pressure changes reflect the impact and replacement capacity of the purging gas on the contents of the passage, flow changes reflect the overall conductivity of the passage and the ability of the cleaning medium to pass through, and electrical signals reflect when the valve core begins to be stressed and when the switching is completed. Placing these three types of information on the same time reference can avoid misjudging premature pressure peaks as abnormalities due to sampling clock offset. The baseline reconstruction module constructs an ideal process baseline model based on preset purge flow rules and drive electrical timing. The preset purge flow rules are: the fluid follows a turbulent resistance model within a limited pipe diameter, and the process of the purge gas from valve opening to establishing stable flow satisfies a preset first-order inertial element response law. Specifically, the system transfer function of the first-order inertial element response law can be expressed as: G(s) = K / (Ts+1) Where K is the steady-state gain of the fluid response, T is the inertial time constant established by the fluid, and s is the Laplace operator; This model corresponds to a state where the equipment is structurally intact, there are no residual deposits inside the flow path, and the valve body operates without delay. The above characteristics represent the pressure decay, flow recovery, and electrical response relationships that the equipment should exhibit in a healthy state. By pre-defining an ideal standard response model under complex industrial conditions, the deviations of subsequent process data are technically relevant. The parameter injection module introduces two types of failure factors into the ideal process baseline model. One is bypass flow resistance attenuation, which is used to represent the tiny leakage branches formed after the seal wears. This type of problem will cause the pressure difference that should be established quickly during the purging stage to be diverted, resulting in slower recovery or tailing. Specifically, slower recovery means that the pressure difference establishment time exceeds the preset time threshold, and tailing means that the time constant of the waveform returning to the baseline value increases. The second is nonlinear viscous damping, which is used to express the flow stagnation caused by sample gas residue, condensate adhesion, or dead volume retention. Such problems usually do not manifest as instantaneous spikes, but rather as significant time delays and stagnation in the path recovery process after purging, and elongation of the waveform tail. The specific method of parameter injection is as follows: the bypass flow resistance attenuation parameter is used as the equivalent flow splitting coefficient of the leakage branch and introduced into the main fluid control equation of the ideal process reference model to reduce the pressure difference in the main path; specifically, the main fluid control equation is:

[0019] in, To characterize the transient value of flow rate, A constant coefficient characterizing the flow capacity of a valve. The effective pressure difference of the purge gas across the main path is represented by the bypass flow resistance attenuation parameter injected into the model, which is reflected as the effective pressure difference in the equation under the same input pressure. The reduction is proportional; The nonlinear viscous damping parameter is used as a time-varying function of the flow resistance coefficient and superimposed on the fluid recovery stage of the ideal process reference model to simulate the delay effect; the process simulation waveform obtained after parameter injection provides a theoretical reference for how the waveform would change if a certain failure did occur. The residual extraction module generates two types of deviations: one is the actual deviation of the real-time process from the ideal baseline, which includes the mixed effects of real failure factors and industrial noise; the other is the theoretical deviation of the simulation anomaly from the ideal baseline, which mainly retains the morphological characteristics of a certain failure path. Subsequently, the coupling decision module does not use a single peak value as the sole basis for decision-making, but examines the degree of similarity between the field deviation and the theoretical deviation in terms of evolution trajectory and distribution structure. If the two are similar in terms of time unfolding mode and overall discrete characteristics, it indicates that the field anomaly is more likely to come from the injected failure mechanism rather than random environmental interference. Finally, the feedback control module adjusts the subsequent sampling strategy based on the decision result, or outputs cleaning control commands to the upper programmable logic controller, distributed control system, or analysis cabin control station. As an anomaly protection mechanism, when the field experiences signal loss within a preset duration from the sensor, incomplete sampling window, missing valve drive signal, or asynchronous pressure and flow sampling, the system prioritizes determining that the data from the current round does not meet reliable comparison conditions. In this case, instead of directly issuing a cleaning anomaly conclusion, it enters a conservative mode: maintaining the most recent valid control parameters, shortening the next round of acquisition, or triggering supplementary acquisition. If complete aligned data cannot be formed for several consecutive rounds, a data link anomaly is reported instead of valve body cleaning failure, thereby avoiding misjudging measurement faults as process faults. For example, in the online chromatography cabinet of an ethylene plant, the three-way valve switches to the purge position after each sample gas analysis. During a certain operating cycle, the field pressure waveform shows fluctuations in amplitude within a preset noise tolerance range. If the monitoring system only extracts amplitude characteristics for judgment, it is easily misjudged as being caused by gas source pump pulsation. However, this system, after aligning the waveform with an ideal reference, finds that the tail of the deviation and the theoretical anomaly after the injection bypass flow resistance attenuation meet a preset similarity threshold matching condition. Specifically, let the preset noise tolerance range be... The actual fluctuation amplitude of the on-site pressure waveform is After injecting a bypass flow resistance attenuation parameter with an equivalent shunt coefficient of 0.02 into the benchmark model, the comprehensive similarity S between the actual deviation characteristic and the theoretical deviation characteristic was calculated to be 0.88, which is greater than the preset reporting threshold of 0.85. Meanwhile, the electromagnetic drive current did not show obvious mechanical jamming characteristics. Therefore, the result was attributed to the tendency of sealing micro-leakage rather than simply external noise, and control parameters for valve body maintenance were generated for the industrial upper control layer. The purpose of this step is to couple the fluid dynamics response characteristics, valve actuation mechanism and data discrimination process during the three-way purging process of the chromatograph, so as to distinguish between micro-leakage, residue and environmental noise, and improve the reliability of the online chromatography system in identifying the cleanliness status in complex industrial sites.

[0020] The data acquisition module includes: a pressure acquisition unit for acquiring transient pressure signals reflecting the impact effect of the clean airflow in the purging path; a flow acquisition unit for acquiring transient flow signals reflecting the flow capacity of the clean medium in the purging path; an electrical acquisition unit for acquiring valve drive coil current signals and control pulse signals; and a timing alignment unit, connected to the pressure acquisition unit, flow acquisition unit, and electrical acquisition unit, for aligning the acquired timing signals with the rising edge of the control pulse or the valve position switching trigger moment as a unified time zero point to generate real-time process data.

[0021] This embodiment provides a data acquisition mechanism for the transient process of purging. Specifically, in the above-mentioned online chromatography analysis cabinet scenario of petrochemical plant, if only pressure signals are collected, although the transient pressure fluctuations after the purging gas impact can be obtained, it is difficult to determine whether the fluctuations are due to narrowing of the passage, micro-leakage shunting, or simply the valve core not being in place. If only current signals are collected, it can only be known whether the drive is issued and whether the coil is energized, but it cannot be confirmed whether the gas has been truly replaced. Therefore, this embodiment adopts simultaneous acquisition of pressure, flow, and electrical signals, and forms a unified process record under the same event through time alignment. In the description, the pressure acquisition unit is set at a key node in the purge path, such as the position between the downstream of the three-way valve and the chromatographic injection line, to acquire the pressure transient after switching; this signal can reflect whether a sufficient pressure difference is formed after the purge gas enters the path to drive the residual sample gas out; the flow acquisition unit is arranged in the main purge path or the return branch to record the flow changes during switching; if the flow build-up lag time exceeds the preset normal response time window, it usually indicates that there is adhering residue, local blockage or abnormal damping in the main path; The electrical acquisition unit collects the control pulses output by the programmable logic controller and the valve drive coil current; the control pulse represents the moment when the control system issues the intention to act, while the rise, hold and fall of the coil current reflect the actual execution process of the valve drive. The timing alignment unit sets the rising edge of the control pulse or the valve position switching trigger time to a unified zero point. Thus, if in one event the pressure waveform experiences a main peak within a preset duration after the zero point, followed by a stable flow rate, while in another event the current has established normally but the pressure and flow response times exceed a preset delay threshold, the two can be identified as different types of problems. For example, the same purging cycle corresponds to three data segments, denoted as segment P, segment Q, and segment E. If segment E has entered a stable energized state, while segments P and Q show a timing lag on the time axis, it is closer to mechanical hysteresis or flow path obstruction. If segment P deviates continuously at the tail after initial establishment, while segment Q exhibits a weak and continuous discharge, it is closer to micro-leakage. As an anomaly protection mechanism, if the control pulse is lost, but the valve position feedback sensor can still output a switching trigger signal, the system will use the valve position switching moment as the zero point; if both are missing, the data in this round will only be archived and will not be used for subsequent coupling decisions; if the pressure sensor is saturated by a short-term impact, the flow and electrical signals will be allowed to align first, and this round will be marked as a pressure channel degradation; after multiple consecutive degradations, the system will prompt for maintenance of the data acquisition hardware. For example, during the night shift operation of the same online chromatograph, after the programmable logic controller issues a purge control pulse, the electrical acquisition unit records a rapid increase in coil current, indicating that the drive has been activated; the pressure acquisition unit records the purge impact peak within the first preset time window, and the flow acquisition unit records the establishment of the main flow path; after the three data streams are aligned to a unified zero point, the sequence of valve action, gas impact, and flow recovery can be clearly shown; if the valve core is slightly stuck the next day, although the control pulse is still normal and the current is present, the pressure peak and flow establishment are shifted later overall, and this difference can be retained to provide a basis for subsequent judgment; The purpose of this step is to construct reliable real-time process data through multi-source synchronous acquisition and a unified time reference, thereby achieving a complete characterization of the purging physical process and avoiding misjudgments caused by a single sensor value.

[0022] The benchmark reconstruction module is used to construct an ideal process benchmark model based on preset purging flow rules and valve drive electrical timing data, under the theoretical assumptions of no leakage influence, no residual influence, and no mechanical delay influence. The benchmark reconstruction module is also used to output the ideal pressure decay waveform, ideal flow change waveform, and ideal electrical response waveform corresponding to the valve position switching process, so as to jointly constitute the ideal process benchmark waveform.

[0023] This embodiment provides an ideal process benchmark reconstruction mechanism. Specifically, based on the aforementioned multi-source acquisition, if the historical average waveform is directly used as a normal reference, it will be affected by the cumulative effects of gradual equipment wear, seasonal temperature drift, and gas source fluctuations. As the equipment operating cycle increases, abnormal features are easily smoothed out and then masked by the new average value. Therefore, this embodiment does not use simple averaging, but constructs a theoretical ideal process benchmark model based on the purging flow rules and drive timing. In the detailed description, the benchmark model corresponds to the following engineering conditions: the three-way valve core switches without additional delay, the seal is intact with no bypass leakage, there is no significant residual dead volume inside the purging passage, and the purging air source is stable and can fully establish flow. Based on this assumption, the system generates ideal pressure decay waveform, ideal flow change waveform, and ideal electrical response waveform. The ideal pressure waveform reflects the natural process of pressure establishment and release after switching, the ideal flow waveform reflects the flow establishment and stabilization process after the main passage is turned on, and the ideal electrical waveform reflects the normal drive profile of coil energization, holding, and release. These three ideal waveforms together constitute the event template when the equipment is healthy and the purging is compliant. For example: In a standard purging event, the time window is divided into an initial segment, an intermediate segment, and a recovery segment; the ideal electrical waveform should quickly enter the energized state in the initial segment, the ideal pressure waveform should form a stable impact and attenuation in the intermediate segment, and the ideal flow waveform should be established before the recovery segment; if the deviation of the field data in these key segments has a specific direction, it can be mapped to a specific failure mechanism more directly; Furthermore, in this embodiment, the absence of dead volume effect is a specific expression of the absence of residue effect. It is used to emphasize that when constructing an ideal template, the additional storage and release effects caused by adhesion, retention, condensation, or local cavities are not brought into the baseline model. In other words, in the context of this embodiment, the residue effect can be manifested as the release lag after medium adhesion or as the displacement residue caused by dead volume. Both are considered as non-ideal factors that should be excluded in the ideal model. After this treatment, the health assumptions and modeling conditions in this embodiment are consistent in engineering meaning, avoiding the misunderstanding of dead volume as a condition independent of residue. Furthermore, when the reference reconstruction module outputs the ideal waveform, it does not require the three types of waveforms to have completely fixed absolute amplitudes under any operating condition. Instead, it requires them to satisfy consistent sequential relationships and boundary constraints under the current operating condition template. For example, the electrical response should appear before the fluid response, there should be an interpretable temporal causality between pressure buildup and flow buildup, and the recovery segment should complete the main stabilization within the expected window. In this way, even if there are reasonable fluctuations in the pressure level of the purging medium or the ambient temperature, the system still builds a benchmark around the event structure that should exist under a healthy state, rather than directly absorbing external fluctuations into the ideal template. As an anomaly protection mechanism, during the initial operation phase of the equipment or when the process parameters are being adjusted, the system can first generate an initial ideal template based on the equipment design parameters, pipe diameter, valve type, rated purging gas source conditions, and control cycle. After accumulating a stable number of qualified cycles that meet the preset threshold, the template is then corrected for boundary conditions using these qualified cycles. If the process switching causes changes in the type of purging medium, pressure level, or valve action cycle, a matching ideal template is reselected to avoid using the ideal waveform of low-pressure conditions to compare with high-pressure conditions. For example, in the above-mentioned online chromatograph, under normal shift conditions, when the three-way valve is switched from the sample gas position to the purge position, the ideal electrical response should first show a complete adsorption process, the ideal pressure waveform should show an impact peak within a limited observation period and gradually decline, and the ideal flow waveform should be established smoothly; if the subsequent actual data shows that the electrical response is normal, but the tail of the pressure waveform is prolonged and the flow establishment is slow, the system can identify it as an anomaly that deviates from the ideal reference, rather than regard this state as a new normal average value; The purpose of this step is to provide a stable and interpretable physical reference for subsequent anomaly extraction, thereby converting complex waveforms in the field into deviation descriptions relative to the healthy state.

[0024] The parameter injection module is used to inject bypass flow resistance attenuation parameters, which are used to change the equivalent flow resistance of bypass branches, into the ideal process benchmark model to simulate the micro-leakage cleaning failure path caused by seal aging. The parameter injection module is also used to inject nonlinear viscous damping parameters, which are used to change the flow damping characteristics of the main path, into the ideal process benchmark model to simulate the impact of path residue on the cleaning recovery process. The parameter injection module generates process simulation waveforms based on the injected model.

[0025] This embodiment provides a parameter injection mechanism for cleaning failure mechanisms. Specifically, with only an ideal process baseline, the system can only know that the field deviates from the healthy state, but cannot further explain which type of failure the deviation is closer to. Especially in the chromatograph purging scenario, micro-leakage and residue can cause abnormal pressure and flow waveforms. If the failure path is not deconstructed mechanistically, it is easy to cause confusion between the valve body seal maintenance requirements and the purging parameter optimization requirements. Therefore, this embodiment actively injects failure parameters with different specific attributes on the ideal baseline to construct a process simulation waveform that can be compared. In the description, the bypass flow resistance attenuation parameter is used to express micro-leakage. This parameter is used to characterize that the purging driving force that should have been completed by the main passage is partially diverted through an unexpected discharge branch with an equivalent cross-sectional area smaller than the preset cross-sectional threshold. The bypass flow resistance attenuation parameter is used to quantify the admittance characteristics of the unexpected discharge branch. As a result, the pressure differential in the main passage is not fully established, the pressure attenuation will show a certain tail, and the flow may also show a diversion characteristic with an amplitude less than the set normal flow fluctuation threshold. Such anomalies are often seen in seal ring hardening, valve seat wear, or seal edge fatigue caused by long-term switching. Nonlinear viscous damping parameters are used to express residues. This parameter characterizes the presence of sample gas adsorption layers, condensate, particle deposits, or local dead volumes within the pathway. When the purging medium passes through, it needs to overcome additional flow resistance, and this resistance exhibits nonlinearity with changes in flow rate, amount of deposits, and local geometry. Compared to microleakage, residues are more likely to manifest as slow recovery of the main pathway, elongated waveform tails, and a gradual approach to normal only after repeated purging. The parameter injection module generates process simulation waveforms based on these two mechanisms, forming theoretical characteristic models for different failure modes. For example: using an ideal waveform as a reference, a micro-leakage factor is injected to obtain a micro-leakage simulation waveform, and a residual factor is injected to obtain a residual simulation waveform. The micro-leakage simulation waveform, compared to the ideal waveform, shows that the pressure difference is diverted and the tail-end venting is continuous. The residual simulation waveform, compared to the ideal waveform, shows that the main passage is established slowly and the recovery process is viscous. If the field waveform is closer to the micro-leakage simulation waveform, the maintenance focus is on valve sealing. If it is closer to the residual simulation waveform, the maintenance focus is on enhanced purging or passage cleaning. Furthermore, to maintain consistency in symbols and terminology throughout the text, the aforementioned ideal waveform, micro-leakage simulation waveform, and residual simulation waveform are used as descriptive names and will no longer be represented by single-letter abbreviations, thereby avoiding confusion with symbols used in state classification parameters or similarity calculations later in the text. Moreover, the micro-leakage simulation waveform and residual simulation waveform here refer to specific manifestations of process simulation waveforms under different parameter injection conditions. Although they are distinguished in name, they do not add new modules or new data types at the system structure level. The above definitions are only used to distinguish different simulation conditions and should not be interpreted as physical limitations on the hardware structure. As an anomaly protection mechanism, if a certain round of field anomalies has mixed characteristics of micro-leakage and residual, the parameter injection module can introduce both types of parameters at the same time to form a composite simulation waveform. If the composite simulation still cannot form a stable coupling with the field deviation, it will not be forcibly classified as a certain known failure, but will be marked as an unknown anomaly, triggering a review or manual maintenance check. This can avoid incorrectly applying new process anomaly characteristics to the existing template. For example, in the aforementioned petrochemical site, after the equipment had been running for six months, the purging recovery slowed down during the low-temperature period at night. After the system injected micro-leakage factors and residual factors into the ideal benchmark respectively, it was found that the field deviation was closer to the residual simulation, indicating that the problem was more likely to be caused by the adhesion or condensation of heavy components in the sample gas, rather than valve seal failure. If after running for several more weeks, the field deviation began to show both tail-end diversion and recovery viscosity, the system could further adopt a composite injection method, indicating that the valve had both a seal aging trend and passage contamination accumulation. The purpose of this step is to map abstract abnormal deviations into interpretable physical failure paths, thereby enabling the differentiation between two types of cleaning failure states: valve body micro-leakage and passage residue.

[0026] The residual extraction module includes: a real deviation generation unit, used to perform differential processing on real-time process data and ideal process reference waveform to generate a real deviation vector; a theoretical deviation generation unit, used to perform differential processing on process simulation waveform and ideal process reference waveform to generate a theoretical deviation vector; and a feature compression unit, connected to the real deviation generation unit and the theoretical deviation generation unit, used to perform dimensionality reduction processing on the real deviation vector and the theoretical deviation vector based on statistical projection or feature selection to generate real deviation feature quantity and theoretical deviation feature quantity for coupling decision, respectively.

[0027] This embodiment provides a dual-track residual extraction and feature compression mechanism. Specifically, after obtaining the ideal benchmark and multiple types of simulation waveforms, if the complete original waveform is still directly compared point by point, on the one hand, a large amount of environmental noise will be mixed into the discrimination process, and on the other hand, non-critical features will mask the real fault evolution features. Therefore, this embodiment first generates two types of deviations, and then compresses the part of the deviation that best reflects the change in process state into a feature quantity. In the description, the real deviation generation unit differs between the real-time process data and the ideal process reference waveform to obtain the real deviation vector; it expresses the quantitative index of the amplitude deviation generated by the field process relative to the healthy ideal state and the evolution trend in this time domain; the theoretical deviation generation unit differs between the process simulation waveform and the ideal process reference waveform to obtain the theoretical deviation vector. The differential processing is specifically as follows: under a unified time zero-point reference, the time domain amplitude of the corresponding sampling points of the real-time process data and the ideal process reference waveform is subtracted point by point to obtain the local deviation value at each moment; it expresses the theoretical deviation direction and deviation profile if a certain failure actually exists; both types of deviations are established relative to the same ideal reference, so they are comparable. The feature compression unit further extracts key information that can represent cleaning failure from the deviations; for example, it can retain the initial impact deviation, build-up delay deviation, tailing deviation, pressure and flow coupling deviation, electrical and fluid response misalignment deviation, etc.; this is not a simple deletion of data, but a collection of physical features directly related to the cleaning state. For ease of explanation, this embodiment uses a simplified data model for description: Assume that a certain round of actual deviation consists of six local deviations, among which only the establishment hysteresis of the second segment, the pressure tail of the fourth segment, and the pressure-flow decoupling of the fifth segment can best reflect the fault. Then, after feature compression, these three types of features are retained as subsequent decision inputs, while the segments that are strongly correlated with noise but weakly correlated with faults are weakened. As an anomaly protection mechanism, if the characteristic energy of a single-round deviation data is lower than the preset lower limit, it indicates that the matching degree between the on-site process and the ideal state has met the set requirements. In this case, the feature compression unit can directly output low deviation features for subsequent judgment as approaching the standard. If the characteristic energy of a certain round deviation exceeds the preset upper limit and the distribution features are divergent, and there is a lack of consistency among the three types of deviations: pressure, flow, and electrical, external interference or acquisition anomalies should be considered first. The system can request additional sampling instead of immediately reporting the fault. For example, in the same chromatograph, after a purging cycle, the real-time pressure curve deviates from the ideal reference only in the tail section where the amplitude difference is less than the preset tolerance, while the flow rate curve exhibits recovery hysteresis in the middle and later sections; the real deviation generation unit extracts these two deviation regions; at the same time, the residual simulation waveform also deviates from the ideal reference mainly in the same region; after feature compression, the system retains the two main features of recovery hysteresis in the middle and later sections and viscous tailing in the tail section, and uses them as inputs for subsequent coupling decisions; The purpose of this step is to convert the complex original waveform into deviation characteristics that better represent the evolution of the fault, thereby reducing noise interference and improving the pertinence of subsequent decisions. Pressure timing data, flow timing data, and valve drive electrical timing data are real-time process data formed by real-time sampling signals during the industrial purging process. Among them, the purging and cleaning process control parameters include parameters for characterizing micro-leakage cleaning failure state, parameters for characterizing residual cleaning failure state, and parameters for characterizing cleaning compliance state. The feedback control module outputs valve cleaning and maintenance commands, purging enhanced cleaning commands, or continue operation commands based on the parameters for characterizing micro-leakage cleaning failure state, parameters for characterizing residual cleaning failure state, or parameters for characterizing cleaning compliance state.

[0028] This embodiment provides a state-based control parameter output mechanism. Specifically, in industrial settings, simply outputting anomalies is often insufficient to guide subsequent handling because the maintenance directions for the three states—valve micro-leakage, residual passage, and true compliance—are completely different. If all are handled using the same strategy, it may lead to ineffective shutdowns, over-purging, or missed structural damage. Therefore, this embodiment further reduces real-time process data into state parameters with specific attributes and outputs differentiated control commands accordingly. In the description, the pressure timing, flow timing, and drive electrical timing are all derived from real-time sampling signals during the industrial purging process. This means that the system determines the actual behavior of the equipment in the current cycle, rather than the results of offline analysis afterward. The control parameters obtained by the coupled decision include at least three types of state representations: one type is used to characterize the micro-leakage cleaning failure state, reflecting the aging of seals, wear of valve seats, or bypass leakage trends; another type is used to characterize the residual cleaning failure state, reflecting dead volume retention, attached contamination, or purging recovery lag; and the third type is used to characterize the cleaning compliance state, indicating that the conditions for continuing to put the next sample gas analysis into the current purging cycle have been met. The feedback control module outputs different instructions based on different states. If the micro-leakage state is dominant, it outputs a valve cleaning and maintenance instruction, focusing on guiding maintenance personnel to check the sealing structure, valve core fit, and valve seat integrity. If the residual state is dominant, it outputs a purging enhancement cleaning instruction, such as extending the purging time, increasing the purging flow rate, or increasing the number of repeated purging cycles. If the cleanliness standard is clearly met, a continue operation command is output, allowing the chromatograph to enter the next analysis cycle. A simplified logic control model is used for explanation: the state set includes three categories: micro-leakage state, residual state, and normal standard met state; the control module outputs maintenance actions for micro-leakage state, enhanced purging actions for residual state, and continue operation actions for normal standard met state. Furthermore, to maintain the uniqueness of the symbol meanings throughout the text, the above three types of states are described using complete terminology in this embodiment, instead of using single-letter abbreviations, in order to avoid overlap with the simulation waveform names, similarity symbols, or other local schematic symbols mentioned above; and, the normal compliance state here is consistent with the clean compliance state in this embodiment in terms of engineering meaning, both indicating that the purging wheel meets the conditions for continued operation, and is only an explanatory expansion for ease of description, without introducing new state categories; As an anomaly protection mechanism, if the same round of data simultaneously shows strong micro-leakage cleaning failure and residual cleaning failure, the system can output joint suggestions based on the composite state, such as first performing an enhanced purging, and then arranging to check the valve seal within the nearest window period; if none of the three states have a significant advantage, the system will output a review or conservative operation suggestion, rather than forcibly classifying it into a certain category; if the quality of the real-time sampling signal deteriorates, causing the state parameters to be unreliable, the control module will reduce the level of automatic action and only give a prompt to the operation and maintenance personnel. For example, in the online chromatograph of the above-mentioned petrochemical plant, when the water content of the sample gas is lower than the preset water content benchmark during the high-temperature period of the day shift, the system outputs a cleanliness standard status multiple times, and the feedback control module continuously issues a command to continue operation; after entering the night shift, the tendency of heavy components in the sample gas to condense increases, the system begins to identify the residual cleaning failure status, and issues a purging enhanced cleaning command, automatically extending the purging time; after a period of operation, if a certain valve position still shows a micro-leakage characteristic after enhanced purging, the system further outputs a valve cleaning and maintenance command, prompting the replacement of the seal within the planned maintenance window.

[0029] The purpose of this step is to transform real-time detection results into maintenance-oriented state parameters and control actions, thereby achieving a closed loop from anomaly detection to guiding handling.

[0030] Example 2: The coupling decision module is used to calculate the trajectory distance corresponding to the feature quantity using the dynamic time warping algorithm; the coupling decision module is also used to calculate the distribution distance corresponding to the feature quantity using Mahalanobis distance; the coupling decision module normalizes the trajectory distance and distribution distance, and then performs weighted fusion based on the normalized trajectory distance and distribution distance to generate similarity.

[0031] This embodiment provides a decision mechanism oriented towards fault morphology coupling. Specifically, the single-dimensional distance measurement method has limitations under complex working conditions: if only time trajectories are compared, when the site is affected by the beat drift with an offset less than the preset tolerance, even if the anomaly mechanisms are the same, they may be mistakenly considered dissimilar; if only the overall distribution is compared, two anomalies with different mechanisms but similar amplitude distributions may be confused. Therefore, this embodiment introduces both trajectory distance and distribution distance, and then performs a fusion judgment. In the following descriptions, trajectory distance is used to measure the degree of similarity between actual and theoretical deviation characteristics in terms of temporal evolution; for the purging process, although micro-leakage and residue can both cause anomalies, the start time, peak time and decay process of their anomalies differ in the temporal distribution; dynamic time warping is suitable for handling situations with similar forms but slightly different rhythms. Distribution distance is used to measure the degree of similarity between two sets of features in the overall statistical structure. It focuses more on the synergistic relationship between multidimensional features, such as whether pressure tail deviation always occurs at the same time as flow recovery hysteresis, and whether electrical response has a consistent offset relationship with fluid changes. Mahalanobis distance can take into account the correlation between various features. The feature covariance matrix on which Mahalanobis distance is calculated is pre-calculated based on the feature set extracted from multiple historical valid samples of the equipment under healthy operating conditions. It is used to characterize the intrinsic statistical correlation between multidimensional feature quantities under normal operating conditions. A simplified example can be used for illustration: Let the on-site characteristic be F1, the micro-leakage theoretical characteristic be F2, and the residual theoretical characteristic be F3; if F1 and F2 have a timing deviation at the time of the appearance of the tail signal trailing characteristic, but the overall evolution order is consistent, then the trajectory distance between F1 and F2 is less than the preset trajectory distance threshold; if F1 and F2 also maintain consistency in the linkage relationship of pressure, flow, and electrical deviation, then their distribution distance is also less than the preset distribution distance threshold; after normalization and weighted fusion, the similarity of F1 to F2 is higher than that to F3, and the system tends to judge it as a micro-leakage anomaly; if the trajectories are similar but the distribution differences are large, it indicates that the on-site situation is more likely to be a combination of cycle fluctuations and external disturbances, and it is not appropriate to directly classify it as a known failure type; Furthermore, to avoid the ambiguity that greater distance necessarily equates to higher similarity, this embodiment generates similarity by first fusing distance and then converting direction. Specifically, the normalized trajectory distance and distribution distance are first fused into a comprehensive distance using preset weights. Then, the comprehensive distance is mapped to a similarity index in the same direction, ensuring that the closer the distance is to the theoretical failure template, the higher the similarity; conversely, the further the distance deviates from the theoretical failure template, the lower the similarity. This can be expressed as follows:

[0032] The similarity used for subsequent event determination is denoted as... The trajectory distance after normalization is denoted as... The distribution distance after normalization is denoted as The corresponding fusion weights are denoted as follows: and and satisfy After this processing, the decision-making direction in this embodiment, which reports an anomaly when the similarity is high and judges it as compliant when the similarity is low, has a direct and consistent engineering meaning. Furthermore, normalization does not require the use of a single fixed algorithm. Its core requirement is to compress distances of different dimensions and ranges into a common interval that can be compared in a weighted manner. For example, it can be scaled according to the historical effective distance boundary, the calibration sample distance boundary, or the controlled simulation boundary under each working condition template. If the current distance exceeds the boundary, it is treated as saturation outside the boundary. This preserves the comparability between different failure templates and avoids the unreasonable dominance of a certain distance in the fusion due to differences in distance dimensions. In abnormal situations, if the length of the on-site feature is insufficient, the key segment is missing, or the calculation of a certain type of distance loses its reference significance, the system will reduce the weight of that type of distance, prioritize the use of another type of reliable distance, and mark the result of this round as a low-confidence judgment; if neither type of distance can form a stable conclusion, the system will output the verification status instead of a deterministic fault conclusion. For example, during the continuous operation of the aforementioned chromatograph, the actual deviation of a certain purging event and the residual theoretical deviation show a time axis offset smaller than the preset time tolerance at the point of recovery lag in the middle and later stages, but the overall evolution profile is basically the same. Simultaneously, the linkage between pressure tailing and slow flow recovery is also consistent with the residual theoretical deviation. In the specific calculation, the preset time tolerance is set to 50ms, and the actual time axis offset is 20ms. The calculated normalized trajectory distance... Normalized distribution distance Take the preset fusion weights and The similarity is obtained based on the aforementioned formula. The system thus obtained a high similarity greater than the preset reporting threshold and judged the round as a residual trend. In another event, although the pressure fluctuation amplitude exceeded the preset amplitude threshold, its distribution structure was inconsistent with any theoretical failure template and was more like the background disturbance caused by the periodic pulse of the air source pump. Therefore, it was not falsely reported as a cleaning failure. The purpose of this step is to improve the robustness of discrimination by using the dual constraints of time trajectory and statistical distribution, thereby enabling the distinction between process anomalies and random industrial noise.

[0033] The coupling decision module is also used to generate a cleaning anomaly event based on a preset reporting threshold and a preset suppression threshold, and under the condition that the reporting threshold is greater than the suppression threshold: when the similarity is greater than or equal to the reporting threshold, a cleaning compliance event is generated; when the similarity is less than or equal to the suppression threshold, a cleaning review event is generated; when the similarity is greater than the suppression threshold and less than the reporting threshold, a cleaning review event is generated.

[0034] This embodiment provides a hierarchical event output mechanism. Specifically, after similarity calculation is completed, if only a single threshold is set, two types of problems will occur in edge cases: first, occasional noise slightly above the threshold may trigger false alarms; second, real anomalies slightly below the threshold but existing for multiple rounds may be ignored. Therefore, this embodiment sets two boundaries, a reporting threshold and a suppression threshold, to form three event levels: anomaly, compliance, and verification. In the detailed description, the reporting threshold is used to determine that the field deviation is close enough to a certain type of theoretical failure, which can be considered as a cleaning anomaly; the suppression threshold is used to determine that the field deviation is significantly different from the theoretical failure, which can be considered as the current round of purging meeting the standard; the area between the two is not directly given a final conclusion, but is defined as the review area. The reporting threshold and suppression threshold are set based on the historical calibration sample set. Specifically, the reporting threshold is determined based on the minimum similarity lower boundary between the known samples that have experienced actual cleaning failures and the theoretical deviation vector, while the suppression threshold is determined based on the maximum similarity upper boundary between the historical healthy samples and the theoretical deviation vector. This hierarchical logic conforms to the conservative control principle in industrial settings: clear anomalies are reported promptly, clear normal conditions are avoided from interference, and ambiguous edge cases are subject to further observation. Using a simplified example, a single similarity judgment result can be denoted as S. If S falls within the high similarity interval, it indicates that the actual deviation and the theoretical failure template are strongly coupled, generating a clean anomaly event. If S falls within the low similarity interval, it indicates that even if there are fluctuations, they conform to the characteristics of normal switching disturbances or are within the preset noise tolerance, generating a clean compliance event. If S falls within the transition verification interval, it indicates that there are certain abnormal signs on site, but the evidence is not yet sufficient, generating a clean verification event. This avoids the system from mistakenly triggering control commands due to insufficient judgment margin under boundary conditions. Furthermore, in this embodiment, the threshold object is the similarity after direction unification, rather than the original distance value; that is, high similarity indicates that the field deviation is close to a certain type of failure template, and low similarity indicates that the field deviation is not close to that type of failure template; thus, the reporting threshold is always set in a higher similarity range, and the suppression threshold is always set in a lower similarity range, with a clear hysteresis band between the two, to avoid the same equipment jumping back and forth between abnormal and compliant under boundary conditions; Furthermore, the verification zone is not only a static numerical range, but can also be conservatively confirmed by combining the results of consecutive rounds; for example, when a single round enters the middle zone, the verification is output first; if multiple rounds move towards the high zone, the anomaly confidence is increased; if multiple rounds fall back to the low zone, the compliance status is restored; this does not change the division of the dual threshold events, but rather provides an engineering supplement to the usage of the verification zone, thus making it more adaptable to the field conditions where pump pulsation, temperature drift, and gradual changes in process composition coexist. In abnormal situations, if the reported threshold and suppression threshold are configured incorrectly, the high and low boundaries are reversed, or the difference between the two is less than the preset safety margin, the system will perform configuration verification before activation. If it fails, it will maintain the previous version of valid parameters and alert maintenance personnel to make corrections. If the similarity remains in the review area for a long time, the system can increase the number of review rounds or switch to manual confirmation mode instead of repeating automatic judgment indefinitely. For example, during a week of continuous operation of the online chromatograph, the similarity S of most purging rounds is between 0.45 and 0.55, falling into the low region below the preset inhibition threshold of 0.60. The system outputs this as a cleanliness compliance event without triggering additional intervention from the main control system. When the ambient temperature decreases, causing a phase change in the sample gas components and enhancing condensation characteristics, the similarity of two consecutive rounds reaches 0.72 and 0.78, respectively, entering the intermediate region between the preset reporting threshold of 0.85 and the inhibition threshold of 0.60. The system first generates a cleanliness verification event and shortens the next detection cycle. In the subsequent third round, the similarity reaches 0.89, entering the high region above the reporting threshold. Only then does the system generate a cleanliness anomaly event, thereby avoiding premature shutdown or false alarms due to a single edge fluctuation. The purpose of this step is to establish a more robust event decision boundary through dual-threshold hierarchical output, thereby enabling differentiated management of clearly abnormal, clearly normal, and pending confirmation states.

[0035] Example 3: The feedback control module is also used to send a purging cleaning anomaly reporting command or a valve body cleaning fault command to the industrial upper control layer when a cleaning anomaly event is generated; the feedback control module is also used to maintain the current acquisition cycle and suppress anomaly reporting when a cleaning compliance event is generated; the feedback control module is also used to adjust the sampling frequency, sampling window, or the bypass flow resistance attenuation parameter and / or nonlinear viscous damping parameter value range in the parameter injection module and trigger a re-decision when a cleaning review event is generated.

[0036] This embodiment provides a feedback control mechanism oriented towards event classification. Specifically, without subsequent handling, the system can only output abnormal statuses and cannot automatically adjust the field equipment. Especially in the scenario of continuous online monitoring of chromatographs, the handling actions corresponding to different event levels should be different: clear abnormalities require prompt intervention, clear compliance should avoid ineffective operations, and status verification requires further confirmation through more refined data. Therefore, this embodiment implements differentiated feedback control based on event type. In the detailed description, when a cleaning anomaly event is generated, the feedback control module sends a purging cleaning anomaly reporting command to the industrial upper control layer, and further sends a valve body cleaning fault command if necessary. Such commands can trigger extended purging time, switching to backup analysis channels, reducing the confidence level of the analysis results at that point, or arranging for maintenance teams to check the valve seal and valve seat status. When a cleaning compliance event is generated, the feedback control module maintains the current acquisition cycle, does not change the established operating rhythm, and suppresses anomaly reporting to avoid frequent actions of the control system due to normal fluctuations. When a cleaning verification event is generated, the feedback control module temporarily suspends the output of a definite fault diagnosis conclusion, but instead increases the sampling frequency, extends the sampling window, or appropriately expands the range of failure parameter values ​​in the parameter injection module in order to capture more sufficient evidence in the next round of judgment. A simplified example can be used to illustrate this: If the result of a certain round is abnormal, the system outputs instruction set A; if it meets the standard, the system maintains instruction set B; if it requires review, the system switches to instruction set C; A emphasizes reporting and intervention, B emphasizes maintenance and suppression, and C emphasizes encrypted data collection and re-decision; through this diversion, the control actions are consistent with the strength of the fault evidence. In abnormal situations, if communication with the industrial upper control layer is temporarily interrupted, the feedback control module first saves the events and suggested actions locally and makes the minimum necessary adjustments to the local acquisition strategy; after communication is restored, the data is retransmitted in batches; if the system still cannot escape the intermediate state after multiple checks, it can switch to conservative operation according to the preset safety strategy, such as extending the purging time and prompting manual confirmation, rather than remaining in an uncertain state for a long time. For example, in the aforementioned chromatograph scenario, after an event in the early morning is determined to be a cleaning verification, the system automatically increases the sampling frequency for the next round and extends the observation window to the longer recovery period after valve switching. After a second determination, if the similarity value increases and exceeds the preset reporting threshold and enters the abnormal zone, the feedback control module sends a purging cleaning abnormality reporting command to the distributed control system and suggests switching the analysis point to the backup valve group. Conversely, in most normal shifts, the system continuously outputs compliant events with the sampling cycle remaining unchanged, avoiding unnecessary alarm burden on the upper-level system. The purpose of this step is to directly translate the judgment results into executable industrial control actions, thereby achieving an integrated closed loop of detection, verification, and disposal.

[0037] Example 4: The system also includes a data communication module; the data communication module is connected to the coupling decision module and / or feedback control module for uploading purging and cleaning process control parameters, cleaning abnormal events and cleaning control commands; wherein, the data communication module includes at least one of an industrial Ethernet interface, a serial communication interface or a fieldbus interface.

[0038] This embodiment provides an industrial communication connection mechanism. Specifically, if the system only completes the judgment locally without communicating with the industrial upper control layer, maintenance terminal, and historical database, it cannot support production interlocking, remote operation and maintenance, and long-term status tracking. Especially in the online analysis cabin of petrochemical plants, the chromatograph usually needs to be connected to a programmable logic controller, distributed control system, edge gateway, or asset management platform. Therefore, this embodiment sets up a data communication module to reliably upload process control parameters, abnormal events, and control commands. In the detailed description, the data communication module can adopt at least one of the following: industrial Ethernet interface, serial communication interface, or fieldbus interface; inside the equipment cabinet, the results output by the coupled decision module and feedback control module are first organized into structured data, such as the current event type, micro-leakage tendency parameters, residual tendency parameters, suggested action status, and timestamp; and then transmitted to the industrial upper control layer through an adapted industrial protocol; for linkage scenarios requiring rapid response, industrial Ethernet can be used first; for the renovation of existing equipment, serial communication can be retained; for analysis cabins that have already deployed fieldbuses, they can also be directly connected to the existing network; From an engineering perspective, the communication module not only serves to upload data but also provides a foundation for cross-cycle tracking. Information such as multiple rounds of verification events, the slow upward trend of micro-leakage parameters, and the concentrated occurrence of residual tendencies during low-temperature periods can only be used for planned maintenance and operational condition correlation analysis after entering the upper-level system. As an anomaly protection mechanism, if communication is interrupted, the system first caches key events and control parameters locally and stores them hierarchically according to event priority; for high-priority anomaly events, they can be retransmitted through redundant links; if redundant links are also unavailable, local audio-visual prompts or on-site maintenance interface prompts are triggered; after communication is restored, historical records are retransmitted in chronological order to prevent the loss of key operation and maintenance information. For example, in the above-mentioned online chromatograph, this system communicates with the control station of the analysis cabin via industrial Ethernet, uploading the compliance, verification or abnormal events after each purging to the distributed control system; at the same time, it transmits detailed process control parameters to the local maintenance terminal via a serial interface; when a cleaning abnormality occurs one night, the distributed control system immediately receives the abnormality report and switches to the backup analysis channel, and the maintenance personnel can view the trend of the gradual increase of micro-leakage parameters over the past week on the terminal the next day, thereby arranging valve body maintenance; The purpose of this step is to integrate the local identification results into the entire industrial control and operation and maintenance system, thereby achieving data connectivity between on-site detection and higher-level management.

[0039] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A chromatograph three-way purging seal test and industrial data acquisition system, characterized in that, include: The data acquisition module is used to collect pressure time-series data, flow time-series data, and valve drive electrical time-series data of the fluid passage during the industrial purging and cleaning process. It is aligned with the control pulse trigger time or valve position switching time as a unified time reference to generate real-time process data that characterizes the state of the purging and cleaning process. The benchmark reconstruction module is communicatively connected to the data acquisition module. It is used to construct an ideal process benchmark model under the assumptions of no leakage influence, no residual influence, and no mechanical delay influence based on the preset purging flow rules and the valve drive electrical timing data, and generate an ideal process benchmark waveform. The parameter injection module is communicatively connected to the benchmark reconstruction module and is used to inject preset bypass flow resistance attenuation parameters and / or nonlinear viscous damping parameters into the ideal process benchmark model to generate process simulation waveforms that characterize the cleaning failure state. The residual extraction module is communicatively connected to the data acquisition module, the benchmark reconstruction module, and the parameter injection module. It is used to generate a real deviation vector reflecting the process deviation based on the real-time process data and the ideal process benchmark waveform, and to generate a theoretical deviation vector based on the process simulation waveform and the ideal process benchmark waveform. The coupling decision module is communicatively connected to the residual extraction module. It is used to calculate the similarity based on the trajectory distance and distribution distance determined by the actual deviation vector and the theoretical deviation vector after dimensionality reduction and feature extraction, and to generate the purging and cleaning process control parameters based on the similarity. The feedback control module is communicatively connected to the coupling decision module and is used to adjust subsequent sampling parameters and / or output cleaning control commands to the industrial upper control layer based on the purging and cleaning process control parameters.

2. The chromatograph three-way purge sealing test and industrial data acquisition system as described in claim 1, characterized in that, The data acquisition module includes: The pressure acquisition unit is used to acquire transient pressure signals in the purging path that reflect the impact effect of the clean airflow; The flow acquisition unit is used to acquire transient flow signals that reflect the flow capacity of the cleaning medium in the purging path; The electrical acquisition unit is used to acquire valve drive coil current signals and control pulse signals; The timing alignment unit, connected to the pressure acquisition unit, the flow acquisition unit, and the electrical acquisition unit, is used to align the acquired timing signals with the rising edge of the control pulse or the valve position switching trigger moment as a unified time zero point to generate the real-time process data.

3. The chromatograph three-way purge sealing test and industrial data acquisition system as described in claim 1, characterized in that, The benchmark reconstruction module is used to construct an ideal process benchmark model based on the preset purge flow rules and the valve drive electrical timing data, under the theoretical assumptions of no leakage influence, no residual influence and no mechanical delay influence. The reference reconstruction module is also used to output the ideal pressure decay waveform, the ideal flow change waveform, and the ideal electrical response waveform corresponding to the valve position switching process, so as to jointly constitute the ideal process reference waveform.

4. The chromatograph three-way purge sealing test and industrial data acquisition system as described in claim 1, characterized in that, The parameter injection module is used to inject the bypass flow resistance attenuation parameter, which is used to change the equivalent flow resistance of the bypass branch, into the ideal process reference model to simulate the micro-leakage cleaning failure path caused by seal aging. The parameter injection module is also used to inject the nonlinear viscous damping parameter, which is used to change the flow damping characteristics of the main channel, into the ideal process benchmark model to simulate the impact of channel residue on the cleaning and recovery process. The parameter injection module generates the process simulation waveform based on the injected model.

5. The chromatograph three-way purge sealing test and industrial data acquisition system as described in claim 1, characterized in that, The residual extraction module includes: The real-time deviation generation unit is used to perform differential processing on the real-time process data and the ideal process reference waveform to generate the real-time deviation vector; The theoretical deviation generation unit is used to perform differential processing on the process simulation waveform and the ideal process reference waveform to generate the theoretical deviation vector; The feature compression unit, connected to the real deviation generation unit and the theoretical deviation generation unit, is used to perform dimensionality reduction processing on the real deviation vector and the theoretical deviation vector based on statistical projection or feature selection, so as to generate real deviation feature quantities and theoretical deviation feature quantities for coupling decision, respectively.

6. The chromatograph three-way purge sealing test and industrial data acquisition system as described in claim 5, characterized in that, The coupling decision module is used to calculate the trajectory distance corresponding to the actual deviation feature and the theoretical deviation feature using a dynamic time warping algorithm; The coupling decision module is further used to calculate the distribution distance corresponding to the actual deviation feature and the theoretical deviation feature using Mahalanobis distance; The coupling decision module normalizes the trajectory distance and the distribution distance, and then performs weighted fusion based on the normalized trajectory distance and distribution distance to generate the similarity.

7. The chromatograph three-way purge sealing test and industrial data acquisition system as described in claim 1, characterized in that, The coupling decision module is further configured to, based on a preset reporting threshold and a preset suppression threshold, and under the condition that the reporting threshold is greater than the suppression threshold: When the similarity is greater than or equal to the reporting threshold, a cleaning anomaly event is generated; When the similarity is less than or equal to the inhibition threshold, a cleanliness compliance event is generated; A clean review event is generated when the similarity is greater than the suppression threshold and less than the reporting threshold.

8. The chromatograph three-way purge sealing test and industrial data acquisition system as described in claim 7, characterized in that, The feedback control module is also used to send a purging cleaning anomaly reporting command or a valve body cleaning fault command to the industrial upper control layer when the cleaning anomaly event is generated. The feedback control module is also used to maintain the current collection cycle and suppress abnormal reporting when the cleaning compliance event is generated; The feedback control module is further configured to adjust the sampling frequency, sampling window, or the range of values ​​for the bypass flow resistance attenuation parameter and / or nonlinear viscous damping parameter in the parameter injection module when the cleaning review event is generated, and trigger a re-decision.

9. The chromatograph three-way purge sealing test and industrial data acquisition system as described in claim 8, characterized in that, The system also includes a data communication module; The data communication module is communicatively connected to the coupling decision module and / or the feedback control module, and is used to upload the purging and cleaning process control parameters, the cleaning abnormal events, and the cleaning control commands. The data communication module includes at least one of an industrial Ethernet interface, a serial communication interface, or a fieldbus interface.

10. The chromatograph three-way purge sealing test and industrial data acquisition system as described in claim 1, characterized in that, The pressure timing data, the flow timing data, and the valve drive electrical timing data are real-time process data formed from real-time sampling signals during the industrial purging process; The purging and cleaning process control parameters include parameters for characterizing micro-leakage cleaning failure, parameters for characterizing residual cleaning failure, and parameters for characterizing cleaning compliance. The feedback control module outputs valve cleaning and maintenance instructions, purging and enhanced cleaning instructions, or continue operation instructions based on the parameters used to characterize the micro-leakage cleaning failure state, the parameters used to characterize the residual cleaning failure state, or the parameters used to characterize the cleaning compliance state.

Images