Use monitoring method and system of knife gate valve

Abstract

The invention discloses a use monitoring method and system for a knife gate valve, and relates to the technical field of knife gate valve monitoring, and the method comprises the steps: integrating a multi-mode sensor array at the root of a valve rod of the knife gate valve, an adjacent region of a sealing surface and in a valve cavity flow channel, and collecting multi-source signals in real time; performing dynamic fusion processing on the multi-source signal, and determining a feature extraction algorithm combination based on the current valve rod temperature and historical working condition data to extract a sealing surface damage feature, a transmission chain wear feature and a particle impact feature; constructing a dynamic anomaly judgment model, updating a clustering boundary of a vibration and torque characteristic space in real time, and performing decoupling diagnosis on a sealing surface dry friction risk, thread clearance expansion and gear wear in combination with time-frequency domain fluctuation analysis of a gear meshing frequency harmonic attenuation rate and an angular acceleration; and generating a three-level maintenance strategy according to the diagnosis result, dynamically adjusting the output torque of the execution mechanism and optimizing the switching frequency of the valve.

Description

Technical Field

[0001] This application relates to the technical field of knife gate valve monitoring. Specifically, it relates to a method and system for monitoring the use of a knife gate valve. Background Art

[0002] A knife gate valve is a key device used to control the on-off of fluid in an industrial pipeline system. It realizes the opening and closing of the pipeline through the reciprocating movement of the valve flap along the axis of the valve stem, and is widely used in fields such as chemical industry, energy, and sewage treatment. If the failure mode of the knife gate valve, such as wear of the sealing surface, cannot be detected in time, it may cause abnormal pipeline system pressure, fluid leakage, and even safety accidents. Therefore, the monitoring of the operating state of the knife gate valve is crucial.

[0003] Existing monitoring methods mostly rely on a single sensor (such as only monitoring the operating torque or vibration signal) or a fixed threshold alarm logic, and it is difficult to cover the coupled effects of multi-dimensional failure causes.

[0004] In response to the above problems, no effective solution has been proposed yet. Summary of the Invention

[0005] The embodiments of this application provide a method and system for monitoring the use of a knife gate valve to solve the above technical problems.

[0006] This application provides a method for monitoring the use of a knife gate valve, including:

[0007] Integrate a multi-modal sensor array at the root of the valve stem, the adjacent area of the sealing surface, and the valve cavity flow channel of the knife gate valve to collect multi-source signals in real time; the multi-source signals include vibration signals, operating torque signals, valve cavity pressure pulsation signals, valve stem temperature signals, and angular acceleration signals;

[0008] Perform dynamic fusion processing on the multi-source signals, and determine a combination of feature extraction algorithms based on the current valve stem temperature and historical operating condition data to extract sealing surface damage features, transmission chain wear features, and particle impact features;

[0009] Construct a dynamic anomaly determination model, update the clustering boundary of the vibration and torque feature spaces in real time, and combine the harmonic attenuation rate of gear meshing frequency and the time-frequency domain fluctuation analysis of angular acceleration to decouple and diagnose the dry friction risk of the sealing surface, the expansion of thread clearance, and gear wear;

[0010] Generate a three-level maintenance strategy according to the diagnosis results, dynamically adjust the output torque of the actuator, and optimize the valve opening and closing frequency.

[0011] Further, the extraction of the particle impact feature includes:

[0012] Perform envelope spectrum analysis on the vibration signal, and extract the peak amplitude and distribution density of the impact feature peaks within a preset frequency band;

[0013] Based on the sliding window density clustering algorithm, a dynamic amplitude threshold interval is generated with historical normal operating condition data of a preset period, and when the impact characteristic peak amplitude exceeds the dynamic amplitude threshold interval and the distribution density reaches a preset clustering density, it is determined that there is a dent on the sealing surface that exceeds a preset damage size;

[0014] When the cumulative deviation of the dynamic amplitude threshold interval exceeds a preset ratio within a preset monitoring period, a warning of wear of the sealing surface wear-resistant coating is triggered;

[0015] The window length of the sliding window density clustering algorithm is dynamically adjusted according to the fluctuation amplitude of the particle medium concentration, and when the concentration fluctuation exceeds a first preset fluctuation threshold, the window length is shortened.

[0016] Furthermore, the decoupling diagnosis of the sealing surface dry friction risk includes:

[0017] When the valve stem temperature exceeds the critical temperature of the sealing grease, the vibration signal is subjected to Hilbert transformation to extract the phase mutation point of the stiffness characteristic, and the slope change of the operating torque rising rate in the start-stop stage is simultaneously calculated;

[0018] Based on the dynamic allocation of weight coefficients of the temperature gradient, the phase mutation point and the torque slope are fused and calculated, wherein when the temperature is in the first interval, the torque slope weight ratio is a first preset value, and when the temperature is in the second interval, the fluctuation entropy value of the valve chamber pressure pulsation signal is introduced as a weight item;

[0019] The fusion calculation results are screened for frequency domain correlation, and the mutation points whose frequency coupling degree with the valve chamber pressure pulsation signal is lower than the first preset threshold are eliminated to generate a sealing surface lubrication failure warning.

[0020] Furthermore, the decoupling diagnosis of gear wear includes:

[0021] Decomposing the valve stem angular acceleration signal in the uniform speed stage in the time-frequency domain, calculating the standard deviation and comparing it with the second preset fluctuation threshold to mark the abnormal fluctuation segment;

[0022] Extracting a preset subharmonic component of the actuator gear meshing frequency, calculating its amplitude attenuation rate and comparing it with a preset attenuation threshold;

[0023] Locating the wear of the primary or secondary gear according to the temporal correlation between the occurrence time of the abnormal fluctuation segment and the attenuation of the harmonics;

[0024] The calculation of the harmonic amplitude attenuation rate includes bandpass filtering and sliding window normalization processing on the preset subharmonic components. When the ratio of the current amplitude to the historical average amplitude exceeds the preset sampling period and the cumulative decrease exceeds the preset attenuation threshold, the gear wear is determined.

[0025] Further, the dynamic adjustment rule of the sliding window density clustering algorithm includes:

[0026] When the fluctuation amplitude of the granular medium concentration exceeds the first preset fluctuation threshold, shorten the window length to a preset short period, and dynamically expand the clustering radius based on the standard deviation of the impact characteristic peak amplitude;

[0027] The expansion ratio of the clustering radius is positively correlated with the amplitude standard deviation to suppress misjudgment caused by sudden changes in the medium flow rate.

[0028] Further, the frequency domain correlation screening includes:

[0029] Perform multi-scale wavelet decomposition on the valve cavity pressure pulsation signal, and extract high-frequency sub-band components related to cavitation bubble collapse and particle impact;

[0030] Construct a mutual information entropy matrix between the phase mutation points and the high-frequency sub-band components, and calculate the correlation strength between the energy distribution of each sub-band component and the phase mutation frequency;

[0031] When the correlation strength is lower than the second preset threshold, it is determined as fluid noise interference, and an anti-interference weight coefficient is generated to correct the fusion calculation result;

[0032] Combined with the secondary verification of the corrected fusion index and the torque slope weight, generate a confidence score for the lubrication failure of the sealing surface, and trigger an alarm when the score exceeds the preset risk threshold.

[0033] Further, the analysis of the time series correlation includes:

[0034] Perform multi-resolution dynamic time warping on the harmonic attenuation event and the abnormal fluctuation section to align the time series offsets at different time scales;

[0035] Construct a causal graph based on the fault propagation model, and analyze the causal strength and time delay between the starting moment of harmonic attenuation and the abnormal fluctuation section;

[0036] If the causal strength exceeds the preset correlation threshold and the delay time is within the preset window, it is determined as an associated event of the same transmission chain fault;

[0037] According to the propagation order of the gear meshing path in the causal graph, locate the fault source as the first-stage gear or the second-stage gear.

[0038] Further, the dynamic adjustment of the output torque of the actuator includes:

[0039] Based on the real-time reduction of the sealing surface contact stiffness index, generate an initial compensation torque through a fuzzy controller;

[0040] An incremental online learning module is introduced to dynamically adjust the membership function parameters in the fuzzy rules according to the historical stiffness recovery rate after compensation and the current working condition parameters.

[0041] When it is detected that the stiffness index has not recovered to the preset safety range for multiple consecutive cycles, the reinforcement learning module is activated to optimize the torque gradient increment with the goal of maximizing the compensation effect.

[0042] The optimized compensation torque is written into the actuator control protocol, and the stiffness and torque conversion coefficient database is updated synchronously.

[0043] Furthermore, the sampling rate of the vibration signal is not less than 2000Hz, and the frequency response range of the valve cavity pressure pulsation signal covers the cavitation bubble collapse frequency band.

[0044] As an optional solution, the optimization of the valve opening and closing frequency includes:

[0045] Based on multi-dimensional data such as gear wear rate, temperature, and historical opening and closing times, a hybrid model of hidden Markov and reinforcement learning is constructed.

[0046] Through the hybrid model of hidden Markov and reinforcement learning, the gear life attenuation curve in different temperature ranges is predicted, and the safety boundary of the maximum allowable opening and closing frequency is dynamically calculated.

[0047] When the real-time opening and closing frequency approaches the safety boundary, an adaptive simulated annealing algorithm is used to generate a frequency reduction path and adjust it to within the safety threshold in stages within a preset time.

[0048] The change in the gear wear rate after each frequency reduction is recorded and fed back into the model to optimize the frequency control strategy for the next cycle.

[0049] This application provides a usage monitoring system for a knife gate valve, including:

[0050] A multi-source signal acquisition module for integrating a multi-modal sensor array at the root of the valve stem, in the adjacent area of the sealing surface, and in the valve cavity flow channel of the knife gate valve to collect multi-source signals in real time; the multi-source signals include vibration signals, operating torque signals, valve cavity pressure pulsation signals, valve stem temperature signals, and angular acceleration signals.

[0051] A feature extraction module for dynamically fusing and processing the multi-source signals, and determining a combination of feature extraction algorithms based on the current valve stem temperature and historical working condition data to extract sealing surface damage features, transmission chain wear features, and particle impact features.

[0052] The decoupling diagnosis module is used to construct a dynamic anomaly determination model, update the clustering boundaries of the vibration and torque feature spaces in real time, and perform decoupling diagnosis on the dry friction risk of the sealing surface, the expansion of the thread clearance, and gear wear by combining the harmonic attenuation rate of the gear meshing frequency and the time-frequency domain fluctuation analysis of the angular acceleration;

[0053] The adjustment and optimization module is used to generate a three-level maintenance strategy according to the diagnosis results, and dynamically adjust the output torque of the actuator and optimize the valve opening and closing frequency.

[0054] Based on the embodiments provided in this application, by integrating sensors at the root of the valve stem, in the area adjacent to the sealing surface, and in the valve cavity flow channel, multi-source signals such as vibration, torque, and pressure pulsation are synchronously collected to construct a monitoring system covering key parts, solve the problem of monitoring blind spots in key parts of the prior art, and realize the collaborative perception of multi-dimensional failure modes. Dynamically adjust the feature extraction algorithm based on the real-time working conditions, fuse multi-dimensional features such as sealing surface damage, transmission chain wear, and particle impact, break through the adaptability limitations of traditional fixed thresholds for complex working conditions (such as high temperature, particle medium), and improve the extraction accuracy of abnormal features and the matching degree with the working conditions. Through the dynamic anomaly determination model and the analysis of the gear meshing frequency and angular acceleration fluctuations, component-level decoupling diagnosis of the dry friction of the sealing surface, thread clearance, and gear wear is realized, accurately locate the specific faulty components, and provide clear guidance for maintenance. Dynamically output a three-level maintenance strategy according to the diagnosis results, and improve the pertinence and efficiency of maintenance by adjusting the actuator torque, optimizing the switching frequency, etc., and reduce the safety risks and operation and maintenance costs caused by failure modes. Brief Description of the Drawings

[0055] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, and constitute a part of this application. The illustrative embodiments and descriptions thereof of this application are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0056] Figure 1 is a flowchart of an optional method for monitoring the use of a knife gate valve according to an embodiment of this application;

[0057] Figure 2 is a flowchart of an optional method for monitoring the use of a knife gate valve according to an embodiment of this application;

[0058] Figure 3 is a structural diagram of an optional system for monitoring the use of a knife gate valve according to an embodiment of this application.

[0059] The realization of the objectives, functional characteristics, and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed Embodiments

[0060] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0061] Optionally, as Figure 1 shown, the present application provides a method for monitoring the use of a knife gate valve, including:

[0062] S101, integrating a multi-modal sensor array at the root of the valve stem, in the adjacent area of the sealing surface and in the valve cavity flow channel of the knife gate valve to collect multi-source signals in real time; the multi-source signals include vibration signals, operating torque signals, valve cavity pressure pulsation signals, valve stem temperature signals, and angular acceleration signals;

[0063] Optionally, in this embodiment, the specific position of the valve stem root: the flange transition area where the valve stem is connected to the valve flap (axial distance of 2-5 cm from the valve flap sealing surface), or the outer peripheral surface of the valve stem bearing support section (near the input end of the actuator reducer). Deploy a strain torque sensor and a three-axis angular acceleration sensor at the root of the valve stem, including: circumferentially pasting 4 groups of strain gauges (evenly distributed at 90°) at the root of the valve stem, and measuring the torsional stress of the valve stem in real time through a Wheatstone bridge (measurement range 0-500 N·m); using high-temperature resistant glue to paste the sensor module on the outside of the valve stem bearing, with the axis aligned with the center axis of the valve stem, and collecting the angular acceleration fluctuations within the range of ±15° / s² during the switching process (resolution 0.01° / s²).

[0064] Optionally, in this embodiment, the adjacent area of the sealing surface includes the adjacent area of the valve flap sealing surface and the adjacent area of the valve seat sealing surface. Among them, the adjacent area of the valve flap sealing surface includes: a 5-15 mm annular area on the radial outside: on the outer periphery of the metal matrix at the edge of the valve flap sealing surface, a micro vibration sensor is fixed by laser welding to collect the high-frequency vibration (200-2000 Hz) when the sealing surface contacts; the stress concentration area on the back of the sealing surface: 2 mm below the valve flap sealing surface (along the thickness direction), a micro pressure sensor array (single sensor size 1×1 mm) is embedded to monitor the contact stress distribution of the sealing surface. The adjacent area of the valve seat sealing surface includes: the upstream flow channel wall surface of the valve seat sealing surface: on the upstream side (axial distance of 1-3 cm) of the edge of the valve seat sealing surface, a high-frequency pressure pulsation sensor (frequency response 100 kHz) is integrally encapsulated by 3D printing to capture the transient pressure wave when the cavitation bubble collapses; the bottom of the flow channel in the projection area of the sealing surface: at the bottom of the flow channel directly below the valve seat sealing surface (vertical distance of 5-10 cm from the sealing surface), an ultrasonic liquid level sensor is installed to monitor the solid deposition height in the particle-containing medium.

[0065] Optionally, in this embodiment, the valve chamber flow channel includes: a valve chamber inlet section flow channel: An electromagnetic particle concentration sensor is installed in the straight pipe section of the upstream flow channel of the valve (at a distance of 5 times the pipe diameter from the valve seat sealing surface) to monitor the volume concentration of particles > 50μm in the medium through the change in conductivity; the flow-through surface of the valve disc flow channel: In the fluid scouring area when the valve disc is opened (such as the valve disc guiding inclined plane), a thin-film temperature sensor (with a thickness of 50μm) is used to measure the medium temperature, and the surface of the sensor is coated with a wear-resistant coating (such as tungsten carbide) to adapt to the working condition where the particle flow velocity ≤ 15m / s.

[0066] S102, perform dynamic fusion processing on multi-source signals, determine the combination of feature extraction algorithms based on the current valve stem temperature and historical working condition data, so as to extract the sealing surface damage feature, transmission chain wear feature, and particle impact feature;

[0067] S103, construct a dynamic anomaly determination model, update the clustering boundary of the vibration and torque feature space in real time, and combine the harmonic attenuation rate of the gear meshing frequency and the time-frequency domain fluctuation analysis of the angular acceleration to decouple and diagnose the dry friction risk of the sealing surface, the expansion of the thread clearance, and gear wear;

[0068] S104, generate a three-level maintenance strategy according to the diagnosis result, and dynamically adjust the output torque of the actuator and optimize the valve opening and closing frequency.

[0069] Based on the embodiments provided in this application, by integrating sensors at the valve stem root, adjacent areas of the sealing surface, and within the valve chamber flow channel, multi-source signals such as vibration, torque, and pressure pulsation are synchronously collected to construct a monitoring system covering key parts, solve the problem of monitoring blind spots in key parts of the prior art, and realize the collaborative perception of multi-dimensional failure modes. Dynamically adjust the feature extraction algorithm based on the real-time working condition, fuse multi-dimensional features such as sealing surface damage, transmission chain wear, and particle impact, break through the adaptability limitations of traditional fixed thresholds for complex working conditions (such as high temperature, particle medium), and improve the extraction accuracy and working condition matching degree of abnormal features. Through the dynamic anomaly determination model and the analysis of gear meshing frequency and angular acceleration fluctuation, realize the component-level decoupling diagnosis of dry friction of the sealing surface, thread clearance, and gear wear, accurately locate the specific faulty component, and provide clear guidance for maintenance. Dynamically output a three-level maintenance strategy according to the diagnosis result, and improve the pertinence and efficiency of maintenance by adjusting the actuator torque, optimizing the switching frequency, etc., and reduce the safety risk and operation and maintenance cost caused by the failure mode.

[0070] Further, as Figure 2 shown, the extraction of the particle impact feature includes:

[0071] S201, perform envelope spectrum analysis on the vibration signal, and extract the impact feature peak amplitude and distribution density within the preset frequency band;

[0072] Optionally, in this embodiment, for the high-frequency vibration generated by the impact of 0.5-2 mm particles on the sealing surface, the preset frequency band is set to 100-500 Hz (covering the area where the particle impact energy is concentrated and filtering out the low-frequency noise of the conventional vibration of the valve).

[0073] S202. Based on the sliding window density clustering algorithm, generate a dynamic amplitude threshold interval with the historical normal condition data of a preset period. When the amplitude of the impact characteristic peak exceeds the dynamic amplitude threshold interval and the distribution density reaches the preset clustering density, it is determined that there is a dent on the sealing surface that exceeds the preset damage size.

[0074] Optionally, in this embodiment, the preset period may include but is not limited to 1 hour, 10 minutes, etc. Specifically, the initial length of the sliding window is 1 hour (covering the average particle concentration under stable conveying conditions), and it is shortened to 10 minutes during abnormal fluctuations (to quickly respond to sudden high-concentration particle impacts).

[0075] The preset clustering density may include but is not limited to 5 per second, 6 per second, etc. When the distribution density of the impact characteristic peak in the vibration signal exceeds 5 per second within a unit time (corresponding to the sealing surface being subjected to more than 5 effective particle impacts per second), it is determined that there is a dent of ≥0.5 mm. The preset damage size may include but is not limited to 0.5 mm, 0.6 mm, etc.

[0076] S203. When the cumulative offset of the dynamic amplitude threshold interval exceeds a preset ratio within the preset monitoring period, trigger a warning for the wear of the wear-resistant coating on the sealing surface; among them, the window length of the sliding window density clustering algorithm is dynamically adjusted according to the fluctuation amplitude of the particle medium concentration. When the concentration fluctuation exceeds the first preset fluctuation threshold, the window length is shortened.

[0077] Optionally, in this embodiment, the first preset fluctuation threshold may include but is not limited to 20%, 30%, etc. For example, when the fluctuation amplitude of the particle concentration in the pipeline exceeds 20% within 5 minutes (such as suddenly increasing from 15% to 18%), it is determined as a significant concentration change and the sliding window shortening mechanism is triggered.

[0078] The preset ratio may include but is not limited to 15%, 20%, etc. For example, when the cumulative offset of the dynamic amplitude threshold interval exceeds 15% within 8 hours (such as shifting from 20 mV to 23 mV), trigger a warning for the wear of the wear-resistant coating (corresponding to the critical value of the coating thickness wear of 0.1 mm).

[0079] Based on the embodiments provided in this application, in order to solve the problem of monitoring the impact damage of the sealing surface under the working condition of the medium containing particles, this embodiment realizes the accurate extraction of the particle impact characteristics by combining the envelope spectrum analysis with the dynamic sliding window density clustering. Specifically, the vibration signal is subjected to envelope spectrum analysis, focusing on the impact characteristic peak within the preset frequency band, effectively filtering out the low-frequency background noise, capturing the high-frequency transient signal generated by the particle impact, and solving the problem of missed detection of high-frequency impact signals by the traditional method; the sliding window length is dynamically adjusted based on the concentration of the particle medium, and the window is automatically shortened when the concentration fluctuation intensifies, so as to adapt to the frequent changes of the impact events in real time and avoid the delayed recognition of the characteristics caused by the fixed window; the dynamic amplitude threshold interval is generated by the density clustering algorithm, and the sealing surface dent is determined by combining the impact peak distribution density, rather than relying on the fixed threshold, which can adapt to the differences in the degree of particle erosion under different working conditions, significantly improve the ability to identify the early damage of the sealing surface in a complex flow field environment, and provide a reliable basis for the early warning of the wear-resistant coating loss.

[0080] Furthermore, the decoupled diagnosis of the risk of dry friction on the sealing surface includes:

[0081] When the valve stem temperature exceeds the critical temperature of the sealing grease, the vibration signal is Hilbert transformed to extract the phase mutation point of the stiffness characteristic, and the slope change of the operating torque rising rate in the start-stop stage is calculated synchronously;

[0082] Optionally, in this embodiment, the critical temperature of the sealing grease may include but is not limited to 120°C, 125°C, etc. For example, for bentonite-based sealing grease commonly used in high-temperature working conditions, the critical temperature is set to 120°C (above this temperature, the sealing grease begins to harden and fail, and loses its lubricating effect).

[0083] Based on the dynamic allocation of weight coefficients of temperature gradient, the phase mutation point and the torque slope are integrated and calculated, wherein when the temperature is in the first interval, the torque slope weight ratio is a first preset value, and when the temperature is in the second interval, the fluctuation entropy value of the valve cavity pressure pulsation signal is introduced as a weight item;

[0084] Optionally, in this embodiment, the first interval is ≤120°C (normal temperature and medium temperature conditions), and the torque slope weight accounts for 70% (focusing on mechanical friction characteristics); the second interval is >120°C (high temperature conditions), and the fluctuation entropy value of the valve chamber pressure pulsation signal is introduced as the weight item (the weight is adjusted to 50% torque slope + 50% entropy value).

[0085] The fusion calculation results are screened for frequency domain correlation, and the mutation points whose frequency coupling degree with the valve chamber pressure pulsation signal is lower than the first preset threshold are eliminated to generate a sealing surface lubrication failure warning.

[0086] Optionally, in this embodiment, the first preset threshold may include but is not limited to 0.6, 0.7, etc. For example, when the frequency coupling degree between the phase mutation point and the valve cavity pressure pulsation signal is lower than 0.6 (such as the correlation between the vibration signal mutation frequency and the pressure pulsation frequency < 60%), it is determined as irrelevant fluid noise.

[0087] Based on the embodiments provided in this application, for the dry friction risk caused by the failure of the sealing grease in a high-temperature environment, this embodiment proposes a temperature gradient weight allocation and multi-signal fusion diagnosis method. Specifically,

[0088] When the valve stem temperature exceeds the critical temperature of the sealing grease, the stiffness characteristic phase mutation point of the vibration signal is extracted through Hilbert transform to accurately locate the non-linear change of the sealing surface contact state, and the slope of the torque rise rate is synchronously monitored to capture the abnormal increase trend of the torque at the initial stage of dry friction; the weight coefficient is dynamically adjusted based on the temperature gradient, focusing on the torque slope analysis at low temperatures and introducing the fluctuation entropy value of the valve cavity pressure pulsation (reflecting the stability of the flow field) at high temperatures to avoid the diagnostic deviation of a single signal under extreme working conditions; the pseudo-mutation points coupled with fluid noise are screened out through frequency domain correlation, and only the features strongly related to the lubrication failure of the sealing surface are retained, solving the false alarm problem caused by signal noise superposition in traditional methods under high temperature and high pressure, realizing multi-dimensional cross-validation of the lubrication state of the sealing surface, and improving the early warning confidence level.

[0089] Furthermore, the decoupled diagnosis of gear wear includes:

[0090] Perform time-frequency domain decomposition on the valve stem angular acceleration signal in the uniform motion stage, calculate the standard deviation and compare it with the second preset fluctuation threshold to mark the abnormal fluctuation section;

[0091] Optionally, in this embodiment, the second preset threshold may include but is not limited to 0.5° / s², 0.6° / s², etc. For example, when the standard deviation of the valve stem angular acceleration signal in the uniform motion stage exceeds 0.5° / s² (corresponding to the mechanical response with a valve stem thread clearance ≥ 0.1 mm), it is marked as an abnormal fluctuation section.

[0092] Extract the preset sub-harmonic components of the gear meshing frequency of the actuator, calculate its amplitude attenuation rate and compare it with the preset attenuation threshold;

[0093] Optionally, in this embodiment, the preset attenuation threshold may include but is not limited to 20%, 30%, etc. For example, when the cumulative amplitude reduction of the sub-harmonic components of the gear meshing frequency exceeds 20% (such as from 50 mV to 40 mV), it is determined that the gear is worn (simulating the signal attenuation of 0.3 mm of tooth surface wear).

[0094] Locate the wear of the first-stage or second-stage gear according to the temporal correlation between the occurrence time of the abnormal fluctuation section and the harmonic attenuation;

[0095] Among them, the calculation of the harmonic amplitude attenuation rate includes band-pass filtering and sliding window normalization processing of the preset sub-harmonic components. When the ratio of the current amplitude to the historical average amplitude exceeds the preset sampling period and the cumulative decline exceeds the preset attenuation threshold, gear wear is determined.

[0096] Optionally, in this embodiment, the preset sampling period may include but is not limited to 50 switching operations, 60 switching operations, etc. For example, the sampling period is set to 50 switching operations (covering the regular maintenance period of the gearbox, about 1 week of industrial operation time).

[0097] Based on the embodiments provided in this application, for the positioning problem of gear wear in the actuator, this embodiment realizes component-level diagnosis through angular acceleration time-frequency domain analysis and harmonic attenuation time-series correlation. Specifically,

[0098] Perform time-frequency domain decomposition on the angular acceleration signal in the uniform motion stage, calculate the standard deviation to identify abnormal fluctuation segments, and quantify the irregular motion of the valve stem transmission chain (such as thread clearance or gear meshing deviation); extract the sub-harmonic components of the gear meshing frequency, analyze the gear tooth surface wear degree through the amplitude attenuation rate, and combine band-pass filtering and sliding window normalization to eliminate the interference of working condition fluctuations on the harmonic signal and accurately capture the stiffness degradation caused by gear wear; based on the time-series correlation analysis of abnormal fluctuation segments and harmonic attenuation, locate the first-stage or second-stage gears through the fault propagation path (such as the attenuation signal of the first-stage gear appears earlier than that of the second-stage gear), realizing the leap from "equipment-level alarm" to "component-level positioning", and providing a direct basis for the targeted maintenance of the gearbox.

[0099] Furthermore, the dynamic adjustment rules of the sliding window density clustering algorithm include:

[0100] When the fluctuation amplitude of the particle medium concentration exceeds the first preset fluctuation threshold, shorten the window length to the preset short period, and dynamically expand the clustering radius based on the standard deviation of the impact characteristic peak amplitude;

[0101] Optionally, in this embodiment, the preset short period may include but is not limited to 10 minutes, 8 minutes, etc. For example, when the concentration fluctuation exceeds the limit, the sliding window length is shortened to 10 minutes (to adapt to the frequent impacts of high-concentration particles, such as the intermittent conveying of mine slurry pipelines).

[0102] The expansion ratio of the clustering radius is positively correlated with the amplitude standard deviation to suppress misjudgment caused by sudden changes in the medium flow rate.

[0103] Optionally, in this embodiment, the expansion ratio of the clustering radius may include but is not limited to 5%, 10%, etc. For example, dynamically adjust according to the standard deviation of the impact peak amplitude. When the standard deviation increases by 10% (such as from 10 mV to 11 mV), the clustering radius expands by 5% (to suppress misjudgment of false impact signals caused by sudden changes in flow rate).

[0104] As an alternative implementation, the adaptive formula for the dynamic clustering threshold of particle impact is:

[0105]

[0106] where is the dynamic amplitude threshold corresponding to the current timestamp (unit: m / s²), which is used to adjust the clustering boundary in real time; is the current timestamp (unit: s), which is used to mark the real-time update moment of the dynamic threshold to ensure that the window sliding is synchronized with the working condition change; is the basic threshold (unit: m / s²), which is the median value calculated based on 7 days of normal data; is the fluctuation amplitude of the particle medium concentration (unit: %), which is the deviation between the current moment and the historical mean; is the concentration reference value (unit: %), which takes the mean value of the past 3 days; is the current timestamp is the standard deviation of the impact characteristic peak amplitude corresponding to the current timestamp (unit: m / s²), that is, taking the current timestamp as the calculation base point, the standard deviation (unit: m / s²) calculated from the peak amplitudes of the impact signals within a certain sliding window before the moment characterizes the fluctuation dispersion degree of the impact intensity within this period, and is updated in real time with to reflect the dynamic stability of the impact signal. The clustering radius expansion amplitude of the threshold is adjusted through the coefficient ; is the concentration fluctuation influence factor (for example, it can take values such as 0.2, 0.4, etc.), which controls the adjustment intensity of the concentration change on the threshold; is the standard deviation expansion coefficient (for example, it can take values such as 1.2, 1.5, etc.), which determines the expansion ratio of the clustering radius with the amplitude fluctuation. Based on the above formula, through the dual adjustment of the concentration fluctuation

[0107] and the amplitude standard deviation , the problem of threshold drift caused by sudden changes in particle flow is solved; the expansion coefficient ensures that the clustering radius adaptsively expands when the medium flow suddenly changes, reducing the misjudgment rate.

[0108] ​Based on the embodiments provided in this application, in response to the problem of misjudgment in monitoring caused by sudden changes in the concentration of particulate medium, the dynamic adjustment rules of the sliding window density clustering algorithm are optimized in this embodiment. Specifically, when the medium concentration fluctuation exceeds the threshold, the window length is automatically shortened to a preset short period to ensure that when the particulate impact frequency changes drastically (such as the flow impact caused by sudden changes in valve opening), the characteristic statistical benchmark is updated in real time, avoiding the "pollution" of historical data caused by too long a window; the clustering radius is dynamically expanded based on the standard deviation of the impact peak amplitude, enabling the algorithm to adapt to the signal amplitude fluctuations caused by sudden changes in flow rate, suppressing false impact characteristics (such as pipeline vibration coupling signals) generated by the passing of transient high-concentration particles through the valve cavity, and improving the recognition accuracy of real impact events in a low signal-to-noise ratio environment, thus solving the problem of insufficient adaptability of traditional fixed windows to non-steady operating conditions.

[0109] Further, the frequency domain correlation screening includes:

[0110] Perform multi-scale wavelet decomposition on the valve cavity pressure pulsation signal to extract high-frequency sub-band components related to cavitation bubble collapse and particulate impact;

[0111] Construct a mutual information entropy matrix between the phase mutation points and the high-frequency sub-band components, and calculate the correlation strength between the energy distribution of each sub-band component and the phase mutation frequency;

[0112] When the correlation strength is lower than the second preset threshold, it is determined as fluid noise interference, and an anti-interference weight coefficient is generated to correct the fusion calculation result;

[0113] Optionally, in this embodiment, the second preset threshold may include but is not limited to 0.6, 0.7, etc. For example, when the correlation strength between the high-frequency sub-band component and the phase mutation point is lower than 0.6 (such as the mutual information entropy between cavitation noise and vibration signal < 0.6 bit), it is determined as fluid noise interference. Combining the corrected fusion index with the secondary verification of the torque slope weight, a confidence score for seal surface lubrication failure is generated, and an early warning is triggered when the score exceeds the preset risk threshold.

[0114] Optionally, in this embodiment, the preset risk threshold may include but is not limited to 70 points, 80 points, etc. For example, an early warning is triggered when the confidence score for seal surface lubrication failure exceeds 70 points (full score 100 points, the critical value of the machine learning model trained based on historical fault data).

[0115] As an optional implementation manner, the formula for the multi-weight fusion index of seal failure is:

[0116]

[0117] Among them, is the fusion index for seal failure early warning;

[0118] is the torque rise rate (unit: N·m / s), which is the torque change rate during the start-stop phase; is the operating torque (unit: N·m), which is obtained based on the real-time torque signal of the valve stem transmission chain;

[0119] is the high-frequency sub-band energy (unit: J), which is the energy of the i-th sub-band after multi-scale wavelet decomposition;

[0120] is the mutual information entropy (dimensionless) between the phase mutation point and the pressure pulsation frequency, which quantifies the frequency-domain correlation; is the phase angle (unit: rad) of the i-th phase mutation point, which is the instantaneous phase extracted by Hilbert transform; is the main frequency of the valve cavity pressure pulsation (unit: Hz), which is the characteristic frequency of fluid disturbance determined by FFT analysis; is the number of high-frequency sub-bands, which is the total number of sub-bands selected after multi-scale wavelet decomposition (in this embodiment, n = 5); is the total energy (unit: J), which is the sum of the energies of all sub-bands; is the maximum phase mutation rate (unit: rad / s), which is the stiffness mutation intensity extracted by Hilbert transform; , , are the weight coefficients, which are dynamically allocated according to the temperature gradient; for example, , , ; , , ; , , .

[0121] Based on the above formula, the phase mutations related to cavitation bubbles are screened by the mutual information entropy MIMI to exclude the interference of fluid noise; the weight wPwP introduces the pressure pulsation signal at high temperatures to improve the early warning accuracy.

[0122] Based on the embodiments provided in this application, for the interference of fluid noise on the lubrication failure diagnosis of the sealing surface, multi-scale wavelet decomposition and mutual information entropy matrix analysis are introduced in this embodiment. Specifically, multi-scale wavelet decomposition is performed on the valve cavity pressure pulsation signal to extract high-frequency subbands related to cavitation bubble collapse (100 - 200 kHz) and particle impact (200 - 500 kHz), and separate the flow field characteristics strongly related to the sealing surface damage; construct the mutual information entropy matrix of the phase mutation points and the high-frequency subband components to quantify the correlation strength between the vibration signal abnormality and the flow field excitation, and eliminate the pseudo-mutation points with low coupling degree with fluid noise (such as vibrations caused by pipeline resonance), to avoid misjudging non-sealing surface problems as lubrication failure; correct the fusion index through the anti-interference weight coefficient, and perform secondary verification in combination with the torque slope to generate a warning signal with confidence, realizing highly reliable diagnosis of lubrication failure.

[0123] Furthermore, the analysis of time series correlation includes:

[0124] Perform multi-resolution dynamic time warping on the harmonic attenuation event and the abnormal fluctuation segment to align the time series offsets at different time scales;

[0125] Construct a causal graph based on the fault propagation model to analyze the causal strength and time delay between the starting moment of harmonic attenuation and the abnormal fluctuation segment;

[0126] If the causal strength exceeds the preset correlation threshold and the delay time is within the preset window, it is determined as an associated event of the same transmission chain fault;

[0127] According to the propagation order of the gear meshing path in the causal graph, locate the fault source as the first-stage gear or the second-stage gear.

[0128] Optionally, in this embodiment, the preset correlation threshold may include but is not limited to 0.7, 0.8, etc. For example, when the causal strength between harmonic attenuation and the abnormal fluctuation segment exceeds 0.7 (the causal coefficient calculated by the fault propagation model ≥ 0.7), it is determined as an associated event of the same transmission chain fault. The preset window may include but is not limited to 50 ms, 60 ms, etc. For example, when the time delay is within 50 ms (in line with the mechanical transmission time delay characteristics of the first-stage gear driving the second-stage gear), the fault correlation is confirmed.

[0129] As an alternative implementation, the determination of the harmonic attenuation rate of gear wear is based on the following formula:

[0130]

[0131] Where, is the harmonic attenuation rate (unit: %), quantifying the degree of gear wear; is the harmonic amplitude of the current k-th sampling (unit: m / s²); is the mean value within the sliding window (24 hours) of the historical average amplitude (unit: m / s²); is the absolute value of the time series alignment offset (unit: s), the time difference after dynamic time warping; is the associated time window (unit: s), taking 0.2 seconds; is the number of consecutive sampling periods (for example, it can be taken 5 times).

[0132] Based on the above formula, through penalize events with excessive time series offsets, excluding non-associated noise interference; through sliding window normalization, suppress the influence of instantaneous fluctuations and reduce the false alarm rate.

[0133] Based on the embodiments provided in this application, aiming at the problem of tracing the wear of multi-stage gears in the transmission chain, this embodiment proposes multi-resolution dynamic time warping and causal graph analysis. Specifically, perform multi-resolution time series alignment on harmonic attenuation events and abnormal angular acceleration fluctuation segments to eliminate the time offset caused by the sampling rate differences of different sensors, and accurately capture the propagation time series of fault signals (such as the wear of the first-stage gear appears abnormal before the second-stage gear); construct a causal graph based on the mechanical structure of the transmission chain, analyze the causal strength and time delay between harmonic attenuation and abnormal fluctuations, and use the physical laws of the gear meshing path (such as the first-stage gear driving the second-stage gear) to judge the hierarchical relationship of the fault sources. Through the dual constraints of the causal strength threshold and the time delay window, exclude irrelevant vibration interference (such as pipeline mechanical vibration), realize the fault tracing of the multi-stage gear system, and significantly improve the maintenance efficiency of the transmission chain.

[0134] Furthermore, dynamically adjust the output torque of the actuator, including:

[0135] Based on the real-time reduction of the sealing surface contact stiffness index, generate an initial compensation torque through a fuzzy controller;

[0136] Introduce an incremental online learning module, and dynamically adjust the membership function parameters in the fuzzy rules according to the historical stiffness recovery rate after compensation and the current working condition parameters;

[0137] When it is detected that the stiffness index has not recovered to the preset safe range for multiple consecutive periods, activate the reinforcement learning module to optimize the torque gradient increment with the goal of maximizing the compensation effect;

[0138] Write the optimized compensation torque into the actuator control protocol, and synchronously update the stiffness and torque conversion coefficient database.

[0139] Optionally, in this embodiment, the preset safe range may include but is not limited to that the sealing surface contact stiffness index needs to be maintained above 80% of the initial value (when it is lower than this value, the sealing performance drops significantly, such as from 100 N / mm to 80 N / mm).

[0140] The preset sampling period may include, but is not limited to, 24 hours, 48 hours, etc. For example, the parameter adjustment period of the online learning module is set to 24 hours (covering the stiffness change period caused by the day-night temperature difference of the valve).

[0141] Based on the embodiments provided in the present application, for the problem of contact stiffness degradation caused by seal surface wear, this embodiment designs a torque compensation strategy combining fuzzy control and incremental learning. Specifically, based on the real-time reduction rate of the seal surface contact stiffness index, an initial compensation torque is generated through a fuzzy controller, taking into account the non-linear characteristics of stiffness change and the safety boundary of the torque output of the actuator; an incremental online learning module is introduced to dynamically adjust the fuzzy rule parameters according to the historical compensation effect, enabling the system to adapt to the gradual change of the seal surface material characteristics during the valve aging process (such as the stiffness response change caused by the decrease in hardness). When the conventional compensation fails, the reinforcement learning module is activated to optimize the torque gradient with the goal of maximizing stiffness recovery, forming a "monitoring-compensation-optimization" closed loop, realizing the dynamic adaptive adjustment of the seal surface gap, extending the effective seal life of the valve, and providing an intelligent solution for the predictive maintenance of old valves.

[0142] Furthermore, the sampling rate of the vibration signal is not less than 2000 Hz, and the frequency response range of the valve cavity pressure pulsation signal covers the cavitation bubble collapse frequency band.

[0143] Optionally, in this embodiment, the cavitation bubble collapse frequency band: the frequency response range of the valve cavity pressure pulsation signal covers 100 - 300 kHz (corresponding to the collapse frequency when the cavitation bubble diameter is 0.1 - 1 mm, and the typical energy distribution interval verified by simulation).

[0144] Based on the embodiments provided in the present application, for the high-frequency characteristics of cavitation bubble collapse and particle impact signals, this embodiment clarifies the technical requirements for the sampling rate of vibration signals and the frequency response range of pressure pulsations. Specifically, the sampling rate of the vibration signal is not less than 2000 Hz to ensure the complete capture of the particle impact characteristic peaks below 500 Hz (required by the Nyquist sampling theorem), avoiding frequency aliasing caused by insufficient sampling rate (such as signal distortion above 100 Hz), and solving the "undersampling and missed detection" problem of traditional low-frequency sampling for high-frequency impact signals. The valve cavity pressure pulsation signal covers the cavitation bubble collapse frequency band (usually 100 - 300 kHz), and the transient pressure wave generated by the bubble collapse is monitored in real time through a high-frequency pressure sensor, forming a two-way verification of "flow-induced damage" with the vibration signal (abnormal pressure pulsation precedes visible damage to the seal surface), providing key data support for early cavitation erosion monitoring.

[0145] Optionally, as Figure 3 shown, the present application provides a usage monitoring system for a knife gate valve, including:

[0146] The multi-source signal acquisition module 301 is used to integrate a multi-modal sensor array at the root of the valve stem, in the adjacent area of the sealing surface, and in the valve cavity flow channel of the knife gate valve to collect multi-source signals in real time; the multi-source signals include vibration signals, operating torque signals, valve cavity pressure pulsation signals, valve stem temperature signals, and angular acceleration signals;

[0147] The feature extraction module 302 is used to perform dynamic fusion processing on the multi-source signals, determine the combination of feature extraction algorithms based on the current valve stem temperature and historical working condition data, and extract the sealing surface damage feature, transmission chain wear feature, and particle impact feature;

[0148] The decoupling diagnosis module 303 is used to construct a dynamic anomaly determination model, update the clustering boundary of the vibration and torque feature spaces in real time, and perform decoupling diagnosis on the dry friction risk of the sealing surface, the expansion of the thread clearance, and the gear wear by combining the harmonic attenuation rate of the gear meshing frequency and the time-frequency domain fluctuation analysis of the angular acceleration;

[0149] The adjustment and optimization module 304 is used to generate a three-level maintenance strategy according to the diagnosis results, dynamically adjust the output torque of the actuator, and optimize the valve opening and closing frequency.

[0150] It should be noted that in this application, the embodiments implemented on the side of the use monitoring system of the knife gate valve can be mutually referred to the embodiments implemented on the side of the use monitoring method of the knife gate valve, and this application will not elaborate one by one.

[0151] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be included in the patent protection scope of the present invention by the same token.

Claims

1. A method for monitoring the use of a knife gate valve, characterized in that include: A multi-modal sensor array is integrated at the valve stem root, the adjacent area of ​​the sealing surface and the flow channel of the valve cavity of the knife gate valve to collect multi-source signals in real time; The multi-source signals include vibration signals, operating torque signals, valve chamber pressure pulsation signals, valve stem temperature signals and angular acceleration signals; Dynamically fusion processing is performed on the multi-source signals, and a feature extraction algorithm combination is determined based on the current valve stem temperature and historical operating condition data to extract sealing surface damage features, transmission chain wear features, and particle impact features; Build a dynamic anomaly judgment model, update the clustering boundaries of the vibration and torque feature space in real time, and combine the gear meshing frequency harmonic attenuation rate and angular acceleration time-frequency domain fluctuation analysis to perform decoupling diagnosis of sealing surface dry friction risk, thread clearance expansion and gear wear; A three-level maintenance strategy is generated based on the diagnosis results to dynamically adjust the actuator output torque and optimize the valve switching frequency.

2. The usage monitoring method of the knife gate valve according to claim 1, characterized in that, The extraction of the particle impact characteristics comprises: Performing envelope spectrum analysis on the vibration signal to extract the impact characteristic peak amplitude and distribution density within a preset frequency band; Based on the sliding window density clustering algorithm, a dynamic amplitude threshold interval is generated with historical normal operating condition data of a preset period, and when the impact characteristic peak amplitude exceeds the dynamic amplitude threshold interval and the distribution density reaches a preset clustering density, it is determined that there is a dent on the sealing surface that exceeds a preset damage size; When the cumulative deviation of the dynamic amplitude threshold interval exceeds a preset ratio within a preset monitoring period, a warning of wear of the sealing surface wear-resistant coating is triggered; The window length of the sliding window density clustering algorithm is dynamically adjusted according to the fluctuation amplitude of the particle medium concentration, and when the concentration fluctuation exceeds a first preset fluctuation threshold, the window length is shortened.

3. The method for monitoring the use of the knife gate valve according to claim 1, characterized in that, The decoupling diagnosis of the sealing surface dry friction risk includes: When the valve stem temperature exceeds the critical temperature of the sealing grease, the vibration signal is subjected to Hilbert transformation to extract the phase mutation point of the stiffness characteristic, and the slope change of the operating torque rising rate in the start-stop stage is simultaneously calculated; Based on the dynamic allocation of weight coefficients of the temperature gradient, the phase mutation point and the torque slope are fused and calculated, wherein when the temperature is in the first interval, the torque slope weight ratio is a first preset value, and when the temperature is in the second interval, the fluctuation entropy value of the valve chamber pressure pulsation signal is introduced as a weight item; The fusion calculation results are screened for frequency domain correlation, and the mutation points whose frequency coupling degree with the valve chamber pressure pulsation signal is lower than the first preset threshold are eliminated to generate a sealing surface lubrication failure warning.

4. The method for monitoring the use of the knife gate valve according to claim 1, characterized in that, The decoupling diagnosis of gear wear includes: Decomposing the valve stem angular acceleration signal in the uniform speed stage in the time-frequency domain, calculating the standard deviation and comparing it with the second preset fluctuation threshold to mark the abnormal fluctuation segment; Extracting a preset subharmonic component of the actuator gear meshing frequency, calculating its amplitude attenuation rate and comparing it with a preset attenuation threshold; Locating the wear of the primary or secondary gear according to the temporal correlation between the occurrence time of the abnormal fluctuation segment and the attenuation of the harmonics; Among them, the calculation of the harmonic amplitude attenuation rate includes band-pass filtering and sliding window normalization processing of the preset sub-harmonic components. When the ratio of the current amplitude to the historical average amplitude exceeds the preset sampling period and the cumulative decline exceeds the preset attenuation threshold, gear wear is determined.

5. The method for monitoring the use of the knife gate valve according to claim 2, wherein, The dynamic adjustment rule of the sliding window density clustering algorithm includes: When the fluctuation amplitude of the particle medium concentration exceeds the first preset fluctuation threshold, the window length is shortened to a preset short period, and the clustering radius is dynamically expanded based on the standard deviation of the impact characteristic peak amplitude; The expansion ratio of the clustering radius is positively correlated with the amplitude standard deviation to suppress misjudgment caused by sudden changes in the medium flow rate.

6. The method for monitoring the use of the knife gate valve according to claim 3, wherein, The frequency-domain correlation screening includes: Performing multi-scale wavelet decomposition on the valve cavity pressure pulsation signal to extract high-frequency sub-band components related to cavitation bubble collapse and particle impact; Constructing a mutual information entropy matrix between the phase mutation point and the high-frequency sub-band components, and calculating the correlation strength between the energy distribution of each sub-band component and the phase mutation frequency; When the correlation strength is lower than the second preset threshold, it is determined as fluid noise interference, and an anti-interference weight coefficient is generated to correct the fusion calculation result; Combining the corrected fusion index with the quadratic verification of the torque slope weight to generate a confidence score for seal surface lubrication failure, and triggering an alarm when the score exceeds the preset risk threshold.

7. The usage monitoring method of the knife gate valve according to claim 4, characterized in that, The analysis of the time series correlation includes: Performing multi-resolution dynamic time warping on the harmonic attenuation event and the abnormal fluctuation segment to align the time series offsets at different time scales; Constructing a causal graph based on the fault propagation model to analyze the causal strength and time delay between the starting moment of harmonic attenuation and the abnormal fluctuation segment; If the causal strength exceeds the preset correlation threshold and the delay time is within the preset window, it is determined as an associated event of the same transmission chain fault; According to the propagation order of the gear meshing path in the causal graph, the fault source is located as the first-stage gear or the second-stage gear.

8. The method for monitoring the use of the knife gate valve according to claim 1, characterized in that, The dynamic adjustment of the output torque of the actuator includes: Based on the real-time decline of the seal surface contact stiffness index, generating an initial compensation torque through a fuzzy controller; Introducing an incremental online learning module to dynamically adjust the membership function parameters in the fuzzy rules according to the historical stiffness recovery rate after compensation and the current working condition parameters; When it is detected that the stiffness index has not recovered to the preset safety range for multiple consecutive periods, the reinforcement learning module is activated to optimize the torque gradient increment with the goal of maximizing the compensation effect; Writing the optimized compensation torque into the actuator control protocol and synchronously updating the stiffness and torque conversion coefficient database.

9. The usage monitoring method of the knife gate valve according to claim 1, characterized in that, The sampling rate of the vibration signal is not less than 2000 Hz, and the frequency response range of the valve cavity pressure pulsation signal covers the cavitation bubble collapse frequency band.

10. A usage monitoring system for a knife gate valve, characterized in that, Including: A multi-source signal acquisition module for integrating a multi-modal sensor array at the root of the valve stem, in the adjacent area of the seal surface, and in the valve cavity flow channel of the knife gate valve to collect multi-source signals in real time; The multi-source signals include vibration signals, operating torque signals, valve cavity pressure pulsation signals, valve stem temperature signals, and angular acceleration signals; A feature extraction module, which is used to perform dynamic fusion processing on the multi-source signals, determine a combination of feature extraction algorithms based on the current stem temperature and historical operating condition data, so as to extract the sealing surface damage feature, the transmission chain wear feature and the particle impact feature; A decoupling diagnosis module, which is used to construct a dynamic anomaly determination model, update the clustering boundary of the vibration and torque feature spaces in real time, and combine the harmonic attenuation rate of the gear meshing frequency and the time-frequency domain fluctuation analysis of the angular acceleration to perform decoupling diagnosis on the dry friction risk of the sealing surface, the expansion of the thread clearance and the gear wear; An adjustment and optimization module, which is used to generate a three-level maintenance strategy according to the diagnosis results, dynamically adjust the output torque of the actuator and optimize the valve opening and closing frequency.

Images