Remote intelligent operation and maintenance early warning system of dust removal equipment based on industrial internet of things

Abstract

The application discloses a dust removal equipment remote intelligent operation and maintenance early warning system based on an industrial Internet of Things, relates to the technical field of industrial Internet of Things operation and maintenance, and comprises the following steps: an edge gateway synchronously collects bag leak monitoring signals, ash removal trigger signals, compressed air main pipe pressure, dust remover pressure difference and temperature and humidity, identifies ash removal events, and aligns and slices ash removal signature windows to form ash removal event records and stability and effectiveness labels; ash removal signature features are extracted based on the event records, a template library is established according to chambers or valve groups and working condition clusters, and the template library is controlled and updated through a trusted update access control; the features are compared with the templates to obtain a deviation degree score, and an alarm type, a level and a suspected chamber are classified and output in combination with time continuity and spatial distribution patterns; when the alarm is uncertain, a diagnosis pulse sequence is triggered, and re-measurement is performed after a maintenance action, and an evidence package is generated. The application can inhibit false alarms caused by ash removal transients and wet interference, improve leakage positioning accuracy and remote closed-loop disposal efficiency.

Description

Technical Field

[0001] This invention relates to the field of industrial Internet of Things (IoT) operation and maintenance technology, specifically to a remote intelligent operation and maintenance early warning system for dust removal equipment based on the Industrial Internet of Things. Background Technology

[0002] In industry, baghouse or cartridge dust collectors are used in steel, cement, coal / biomass boilers, waste incineration, and chemical industries to continuously purify dust-laden flue gas. Because these devices use compartmentalized offline or online pulse-jet cleaning, there are numerous points of contact and limited maintenance access points in the factory. Increasingly, companies are leveraging the Industrial Internet of Things (IIoT) to connect signals such as differential pressure, compressed air mains pressure, valve operation, and outlet dust (or bag leak detection) to edge gateways and transmit them to a platform for remote monitoring, alarms, and maintenance. Anomaly warnings on the platform side are primarily displayed on alarm bandwidths determined by threshold exceeding limits, rule triggering, or experience, alerting maintenance personnel.

[0003] In baghouse dust collectors, the relative load of outlet dust or the bag leakage detection signal is coupled with the cleaning process. Pulse jet cleaning causes dust to fall off the filter bag surface, resulting in short-term concentration fluctuations and spikes in the sensor output. The cleaning frequency varies with the operating load, dust properties, and differential pressure control method, and the amplitude and shape of the "normal spikes" exhibit time-varying characteristics. High humidity or dew point conditions cause condensation on the probe surface, altering the charge / conductivity of particles and causing abnormal responses. During long-term operation, dust accumulation, moisture, or conductive bridging on the probe or insulation parts can cause zero-point drift, decreased sensitivity, or intermittent large signals, with the baseline slowly shifting over time. Under these circumstances, a single threshold or rule is insufficient to reliably distinguish between "transient fluctuations in cleaning, ambient humidity, and sensor contamination drift" and "actual filter bag damage / leakage leading to abnormal emissions," resulting in frequent alarms that are difficult to verify and early leaks being masked. Relying on manual adjustment of thresholds or temporary alarm suppression in the long term will lead to increased costs of unplanned unpacking inspections and shutdown troubleshooting, and will result in consequences such as emission exceeding limits, increased filter bag failure, and delayed operation and maintenance decisions after the reliability of alarms decreases.

[0004] If remote monitoring is conducted under conditions such as changes in pulse cleaning operation, humidity approaching the dew point, and probe contamination drift, how can we determine whether abnormal changes in bag leak detection or outlet dust signals correspond to actual filter bag leakage / emissions caused by transient changes in cleaning or sensor status? Summary of the Invention

[0005] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a remote intelligent operation and maintenance early warning system for dust removal equipment based on the Industrial Internet of Things (IIoT). It identifies dust removal events and aligns slices according to the dust removal signature window to form dust removal event records and stability and validity labels. Based on the event records, it extracts dust removal signature features, establishes a template library by compartment or valve group and operating condition cluster, and updates them in a controlled manner through a trusted update access control system. It compares the features with the templates to obtain a deviation score, and classifies and outputs alarm types, levels, and suspected compartments based on temporal persistence and spatial distribution. When an alarm is uncertain, it triggers a diagnostic pulse sequence and retests it after maintenance actions to generate an evidence package. This system can suppress false alarms caused by dust removal transients and moisture interference, improve the accuracy of leak location and the efficiency of remote closed-loop handling, and solves the technical problems described in the background art.

[0006] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: The remote intelligent operation and maintenance early warning system for dust removal equipment based on the Industrial Internet of Things includes: synchronously collecting bag leakage monitoring signals, dust removal trigger signals, compressed air main pipe pressure, dust collector differential pressure, temperature and humidity by the edge gateway; establishing a dust removal signature window according to the dust removal trigger signal and slicing it to form a dust removal event record, and assigning stability and validity labels. Extract the dust removal signature features from the dust removal event records, establish a template library according to the compartment or valve group and operating condition cluster, and update the template library through the trusted access control when the stability and validity labels are satisfied and the condensation risk is satisfied. The deviation score is obtained by comparing the dust removal signature features with the corresponding template. Combined with the time persistence determination and spatial distribution pattern classification, alarm type, alarm level and suspected room location suggestions are output. When the alarm type is uncertain due to environmental or sensor abnormalities, a diagnostic pulse sequence is executed for the selected compartment or valve group, triggering maintenance actions, followed by retesting and generating an evidence package.

[0007] Furthermore, the edge gateway provides unified time synchronization and timestamp annotation for bag leak monitoring signals, dust cleaning trigger signals, compressed air main pipe pressure, dust collector differential pressure, temperature and humidity, and marks missing points and abnormal points and caches the original data, and writes the timestamps and markings into the dust cleaning event record.

[0008] Furthermore, the dust removal trigger signal originates from the pulse controller and is differentiated by compartment. The edge gateway performs de-jitter processing on the dust removal trigger signal. The de-jitter processing includes removing pulses with insufficient width and merging repeated pulses within a short period of time. A dust removal event record is established for each valid pulse, and the event number, occurrence time, and corresponding compartment number are recorded.

[0009] Furthermore, the dust removal signature window covers the stable segment before dust removal and the peak and attenuation segment after dust removal. The length of the dust removal signature window is set by the preset window length parameter and corrected according to the historical peak arrival time. The stability and validity labels are jointly determined by the fluctuation of the dust collector differential pressure within the window, whether the compressed air main pipe pressure is lower than the safety lower limit, and the equipment start-up and shutdown status. Unstable events do not participate in the template library update.

[0010] Furthermore, the dust removal signature features are composed of the peak amplitude of dust removal, the peak area of ​​dust removal, the peak arrival time, the peak decay rate, the peak decay duration, the waveform similarity, and the high-frequency noise intensity. The dust removal intensity-related features are normalized by the compressed air main pipe pressure, and the dust removal signature features are compared in segments by combining the dust collector differential pressure level and the dust collector differential pressure change trend.

[0011] Furthermore, the template library adopts a two-layer structure. The first layer establishes basic templates according to the room number, and the second layer establishes multiple templates according to the working condition clusters. The working condition clusters are determined by the dust collector pressure difference range, production load range, and temperature and humidity corresponding ranges. During operation, the working condition clusters are determined first and then the corresponding multiple templates are called, and the working condition clusters and calling conditions are written into the template library.

[0012] Furthermore, the access control system updates based on the following conditions: stability and validity labels, condensation risk meeting preset thresholds, no current leakage alarms, and sensor health status meeting preset thresholds. The template library update is frozen when a suspected contamination bridging is detected, and the template library update is frozen when an abnormal sensor health status is detected. The template update process is also managed and backtracked.

[0013] Furthermore, the duration determination uses the most recent preset number of dust removal event records as the observation window, and triggers an alarm when the proportion of deviation score exceeding the limit reaches a preset proportion threshold. When the deviation score exceeds the limit for a preset number of consecutive times, the alarm level is upgraded step by step; the spatial distribution pattern classification is determined based on the synchronization range and continuous abnormal range of the deviation score between rooms.

[0014] Furthermore, the spatial distribution pattern classification includes local anomaly patterns, global synchronous anomaly patterns, and dust removal system anomaly patterns. Among them, the local anomaly pattern corresponds to the proportion of compartments with continuously abnormal deviation scores being lower than the first threshold and outputting a suspected compartment ranking. The global synchronous anomaly pattern corresponds to the proportion of compartments with synchronously increasing deviation scores being higher than the second threshold and synchronous with condensation risk. The dust removal system anomaly pattern corresponds to the consistent change in dust removal response pattern synchronous with the decrease in compressed air main pipe pressure.

[0015] Furthermore, when the alarm type is uncertain due to environmental or sensor abnormalities, a diagnostic pulse sequence is executed in a preset order for the selected compartment and the corresponding dust removal signature characteristics are recorded. After triggering a short-term purging of the probe area, a retest diagnostic pulse sequence is immediately executed and an evidence package is generated. The evidence package includes the dust removal event sequence before and after the alarm, deviation change records, condensation risk status, maintenance action records, and retest results.

[0016] (III) Beneficial Effects This invention provides a remote intelligent operation and maintenance early warning system for dust removal equipment based on the Industrial Internet of Things, which has the following beneficial effects: By establishing a dust removal signature window with the dust removal trigger signal as the time anchor point, and simultaneously generating dust removal event records by slicing the bag leak monitoring signal, compressed air main pipe pressure, dust collector differential pressure, temperature and humidity, and assigning stability and validity labels to them, bag leak monitoring is transformed from a continuous wave into an event-based process. The dust removal transient is incorporated into an interpretable response process, avoiding the misjudgment of directly defining the dust removal peak as a leak, and providing a consistent input base for the next step with traceable boundaries.

[0017] By extracting dust removal signature features characterized by amplitude, accumulation, hysteresis, recovery, and disturbance, and normalizing or conditionalizing these features using compressed air header pressure, dust collector differential pressure, and temperature and humidity, the dust removal responses of the same compartment or valve group under different loads, resistances, and humidity conditions can be compared on the same scale. This improves the stability of the template reference and avoids alarm drift caused by threshold failure due to changes in air source or operating conditions.

[0018] By establishing a template library based on compartments, valve groups, and operating condition clusters, and introducing a reliable update access control system, stability and effectiveness labels, condensation risk, leakage alarm status, and sensor health status are incorporated into the template update conditions. This enables the template library to adapt to long-term changes in operating conditions while avoiding contamination by start-up / shutdown, bypass, insufficient gas supply, or moisture contamination events. This creates a constraint loop between "updatable" and "controllable update," ensuring that subsequent deviation scores are always based on a reliable reference.

[0019] By comparing the dust removal signature features with the corresponding template to obtain a deviation score, and combining the time persistence judgment to distinguish between occasional disturbances and trend anomalies, the system further classifies local anomaly patterns, global synchronous anomaly patterns, and dust removal system anomaly patterns by spatial distribution. This enables the system to not only output alarm levels, but also alarm types and suspected room location suggestions, thereby binding anomaly identification with actionable investigation objects and reducing the consumption of operation and maintenance resources caused by ineffective unpacking and blind investigation. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the structure of the remote intelligent operation and maintenance early warning system for dust removal equipment based on the Industrial Internet of Things of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 This invention provides a remote intelligent operation and maintenance early warning system for dust removal equipment based on the Industrial Internet of Things, including: Step 1: Using the cleaning action as the anchor point, the continuous bag leakage signals are organized into cleaning event records with uniform timestamps and boundaries, thus providing directly callable input for subsequent calculation of cleaning signature features and template library.

[0023] Pulse jet cleaning can cause dust layer shedding and airflow disturbance, resulting in bag leakage signals. The appearance of spikes near the dust removal area is a normal response. Fluctuations in main pipe pressure, compartment switching, start-up and shutdown conditions, and humid environments near the dew point can cause the amplitude and shape of these spikes to change over time. If an alarm is triggered solely by continuous waveform exceeding limits, waveform changes caused by dust removal transients, insufficient air supply, or probe contamination can easily be interpreted as signs of leakage. To ensure that subsequent judgments focus on the response process after the dust removal action is triggered, the dust removal trigger edge is identified as a candidate event. The response segments of these candidate events on the bag leak signal, main pipe pressure, and differential pressure are aligned and sliced ​​to establish a correspondence between the dust removal action and the response process.

[0024] Edge gateway records gateway time The controller side receives the control timing. To eliminate cumulative drift, a proportional parameter is used. With offset parameter Construct a mapping and suppress outliers using log-hyperbolic cosine cost: ; Where: proportional parameter :for Used to scale the gateway timescale to align with the reference timescale; offset parameter : Used to compensate for fixed timescale offsets; gateway time : Used to trigger along the time anchor point; controls the timing. : Used as a reference time anchor; sample size : Used to map constraint size; steepness coefficient The range of values ​​is This is used to limit the influence of abnormal residuals on parameters; Mapping yields mapping parameters After that, the Sub-room Gateway time of the next candidate trigger edge Transformation into dust removal time It serves as the sole time anchor point for the signature window, fixing the time boundary before comparing the response patterns; the boundaries of multi-source signal slices are consistent, avoiding alignment errors from affecting subsequent features.

[0025] A glitch in the switching input may cause a trigger edge but no effective blowing. This is to trigger the residual. Gas source residual With signal residual Calculate trigger credibility The three types of residuals correspond to pulse duration deviation, post-trigger pressure drop gap, and inconsistent response morphology after baseline subtraction, respectively: ; Where: Trigger credibility : Used to quantify the reliability of effective spraying; trigger residuals : Used to filter pulse spikes and vibrations; air source residual : Used to constrain the existence of the main pipe pressure response; signal residual : Used to constrain the interpretability of bag leak response; weighting coefficients , , The range of values ​​is This is used to adjust the proportion of influence of the three types of residuals. The weighting coefficients are... , , It can be set based on field experience and can be configured on the gateway.

[0026] When credibility is triggered When the threshold is below, the candidate trigger still preserves the original slice and residual field, but the stable and valid label remains. If not set to 1, the trigger edge is confirmed twice using the gas source and signal response; invalid blow events are not included in subsequent reference references.

[0027] Define the header pressure reference in the pre-segment of the dust removal signature window. In the short period of time after triggering Minimum value Define the pressure drop of the main pipe. , with the most recent The median of a stable and valid event is used as a reference drop. ,definition: ; Gas source residual : Its function is to confirm that there is an observable injection gas source response after triggering; the main pipe pressure pre-reference. : Its function is to serve as a reference for the gas source status before the event occurs; minimum pressure : Its function is to characterize the pressure drop after triggering; Pressure response time : Its function is to limit the observation window for pressure drop; reference drop : It serves as a reference for the gas source response in stable events; pressure trace : Its function is to stabilize values; reference window length : A positive integer, used to determine the sample size of the reference drop.

[0028] Define the pre-leakage benchmark in the pre-section. (Signal for bag leakage) exist (median amount), in a short period of time after triggering Calculate response energy Calculate the disturbance energy in the pre-processing stage. : ; And define: ; Response Energy : Its function is to confirm the existence of a recognizable response to the bag leak signal after triggering; disturbance energy : Its function is to characterize the noise level before triggering; signal residual : Its function is to increase when the response is weaker than the noise, thus suppressing unresponsive events from proceeding; response observation duration : Its function is to limit the observation window of the response energy; the signal is minute. : Its function is to stabilize values; Furthermore, regarding the dust removal time Set the pre-set duration as the center. Cover the background before and after dust removal; restore the endpoint. The intensity of the bag leak signal change first falls below the threshold. The timing is determined, the intensity of change is obtained using a sliding integral, and the derivative is obtained by fitting a local polynomial: ; In the formula: the endpoint of recovery : Used to determine the endpoint of the signature window; dust removal time. : Used for window time anchor points; sliding length : Used to smooth the time scale of changes in intensity; bag leak signal : Used for observing objects in the dust removal response; derivative operator : Used to characterize the instantaneous rate of change for integration; change threshold : The threshold used to determine the return to the background; the infimum operator. : This is used to return the earliest time when the condition is met; The starting point of the window is denoted as the window start point. Thus, the signature window is obtained. This ensures that the window boundaries change self-consistently with the recovery process, and that semantically consistent response fragments are obtained under different cleaning cycles.

[0029] Within the signature window, for the bag leak signal Main pipe pressure With pressure difference Synchronous slicing and resampling to a uniform step size are performed, and monotonically piecewise linear interpolation is used for interpolation. To identify unstable periods caused by start-up, shutdown, bypass, or sudden load changes, the total differential pressure variation is calculated. Together with the lower limit of the main pipe pressure and the condensation margin threshold, a stable and effective label is generated. : ; Where: Total pressure difference : Used to characterize the cumulative change in pressure difference within a window; discrete pressure difference sequence : Used for differential pressure samples after slice resampling; sample subscript : Used to identify discrete time sequence; sample size : This is used to solidify the length of the event log and determine the accumulation range; Stable and effective label The value selection rule is: when the credibility is triggered Not less than the threshold and the total pressure difference varies The pressure should not exceed the upper limit and the main pipe pressure should not be lower than the lower limit, and the condensation margin should be... Stable and effective labels when not falling below the threshold Select 1 if the value is not specified, otherwise select 0, and write the decision criteria along with the slice sequence into the event log. This allows us to solidify available event conditions into fields that can be replayed; step two allows us to directly filter the event update template library with a stable and valid tag of 1, while retaining unstable events as evidence.

[0030] For example, maintenance personnel connect the edge gateway and valve group trigger circuit in parallel within the control cabinet, and also connect the main pipe pressure, differential pressure, condensation margin, and bag leak probe signals. After the equipment is operational, when dust removal is triggered in a certain compartment, the platform displays an event record with the corresponding compartment number. The event log includes the trigger along the three-way slice curve and displays stable and valid labels. When the gas supply is shut off on-site, causing insufficient pressure in the main pipeline, triggering the dust cleaning process again will still generate an event log. However, stable and effective labels Set to 0 and retain trigger credibility This field allows maintenance personnel to attribute abnormal events to the gas source link status without opening the box.

[0031] Specifically, mapping parameters It can be solved using a finite-memory quasi-Newton solver; when the control time cannot be directly obtained. When the pressure slice in the main pipe is at a local minimum, the time can be used as a reference sample. Derivative estimation uses a local quadratic polynomial fit with the minimum absolute deviation as the fitting criterion, and a linear programming solver can be used to solve it. Resampling interpolation can switch between monotonically piecewise linear interpolation and zero-order hold interpolation to match the signal hold characteristics. Equivalent implementations include using the valve coil current instead of the switching trigger signal as a candidate trigger source, or using the transient pressure drop in the main pipe instead of the trigger signal as a candidate trigger source, but both are based on trigger confidence. As a threshold for event solidification, reference time-scale mapping ensures event boundaries are consistent across multi-source signals; trigger credibility binds the trigger edge to the gas source response and bag leak response, preventing invalid injection events from entering the reference baseline; the intensity change endpoint ensures the window covers the peak and recovery segments while maintaining semantic consistency of the boundaries; total differential pressure variation and stable effective labels structurally isolate unstable conditions while still preserving evidence, thus forming an event record set. This allows for the direct extraction of dust removal signature features in step two and the creation of a template library.

[0032] Step 2: Translate the response process of each dust removal event into a dust removal signature feature, and accumulate a multi-channel template library under continuously changing operating conditions so that subsequent judgments are based on the same reference scale.

[0033] Step one has aligned the bag leak signal, main pipe pressure, and differential pressure within the dust removal signature window, and assigned a stable and valid label to each dust removal event record. Trigger credibility Total variation of pressure difference During field operation, the main pipe pressure changes the jet momentum, the pressure difference changes the dust layer state, and approaching the dew point changes the conductivity of the probe surface. These factors simultaneously alter the amplitude, hysteresis, and attenuation process of the dust removal spike. If a threshold is directly used to limit the continuous waveform, the meaning of the threshold will change with the shift in operating conditions. Therefore, it is necessary to construct comparable features on a per-dust-removal-event basis and create template references under different operating conditions.

[0034] Record each dust removal event Read stable and valid tags, trigger confidence, condensation margin, and total pressure difference, and solidify the permission to write to the template update chain into an update permission quantity. The updated permit quantity is used to block non-steady-state events from entering the template update and retains them as evidence fields. ; Where: Update license quantity : Values Its function is to indicate whether the event participates in the template update; indicator function : Values Its function is to map access control conditions to permitted quantities; stable and valid labels. : Values Its function is to determine the stability of an event through the inheritance step; triggering credibility. :for Its function is to limit the triggering of the jetting to be consistent with the response of the air source; Trusted threshold : Value Its function is to limit the minimum allowable value for trigger confidence; condensation margin. : A real number, used to characterize the temperature margin between the event time and the dew point; margin threshold : Real number, used to limit the minimum condensation margin for participation in template updates; Total pressure difference variation : Values Its function is to characterize the intensity of pressure difference fluctuations within the signature window; upper limit of variation. : Values Its function is to limit the maximum differential pressure fluctuation that participates in template updates; Among them, the source of template samples is written as a traceable condition with permissible quantity, so that the template will not be skewed by non-steady-state events; Updating license quantity When valid, construct the cleanup signature feature vector from the slice within the cleanup signature window. .in and The pressure of the main pipe before ash removal is normalized. Using the trigger alignment point as a reference, This was obtained by matching the exponential decay pattern with a robust tail segment. Formed by second-order difference intensities. Step 3: Based on this, compare the dust removal event response process on a unified dimension: ; Where: The feature vector of the cleaned-up signature The value is a five-dimensional real vector, serving as a representation of five comparable descriptors of the dust removal response; normalized peak amplitude. : Values Its function is to normalize the peak cleaning intensity based on the main pipe pressure reference; and to normalize the cleaning area. : Values Its function is to provide a cumulative release characteristic after normalization based on the main pipe pressure benchmark; Peak arrival time : The value is a real number, and its function is to characterize the hysteresis from dust removal triggering to the peak value; decay time constant. : Values Its function is to characterize the recovery speed after the peak; high-frequency disturbance quantity : Values Its function is to characterize the degree of jitter and peak density within the window; Specifically, the main mechanisms are covered by five types of features: amplitude, accumulation, hysteresis, recovery, and perturbation, giving the template interpretable boundaries. Each compartment is numbered. Maintenance Channel Number The corresponding template is used to construct a working condition vector by combining the main pipe pressure benchmark, the pre-averaged differential pressure, and the condensation margin. , and its relationship with the working condition anchor point vector Deviation calculation gate distance Then obtain soft gating weights The details are as follows: ; Working condition vector A three-dimensional real vector used to describe the gas source, resistance, and humidity conditions at a given moment in an event; differential pressure reference. : Real number, used as a reference for pre-event filtering resistance; pre-condensation margin benchmark. Real numbers serve as a reference for the risk of dampness before an event occurs; The working condition anchor point vector is It is recommended to use the same robust recursive update mechanism as the template center (only for updating licenses). (Event update) ; Working condition anchor point vector : A three-dimensional real vector, used to represent the operating conditions of the channel. Operating condition step size coefficient. : Its function is to control the anchor point update speed.

[0035] ; Where: gate distance : Values Its function is to measure the deviation between event conditions and channel conditions; soft-gating weights : Values The function is to determine the contribution weight of the event to the channel template update; the working condition vector. The value is a three-dimensional real number vector, used to describe the working condition location when the event occurs; Working condition anchor point vector The value is a three-dimensional real vector, used to describe the operating conditions of the channel; gating steepness coefficient. : Values Its function is to control the sensitivity of the gating distance to deviations from operating conditions; Scaling coefficient : Values Its function is to control the rate at which weights decay with distance; number of templates : Values ​​are positive integers, used to limit the number of channels; summation index : Values ​​are positive integers; their function is to iterate through channels to complete weight normalization; hyperbolic cosine function : Values Its function is to construct a symmetric and differentiable robust distance mapping; exponential function : Values Its function is to map distance to weights; the natural logarithm function : Takes a real number as its value and is used to take the natural logarithm; the first Channel distance : Values Its function is the first Channel gating distance; Specifically, robust distance and soft gating transform the continuity of operating conditions into template continuity, preventing discontinuities. Once the soft gating weights are determined, the template center vector is maintained in each channel. and template tolerance vector When a certain dust removal event record meets the update permission amount, the dust removal signature feature vector is written into the template center for robust recursion; the acceptable deviation is solidified by the quantile operator as a tolerance to avoid scale expansion when using variance-based scale description.

[0036] ; Where: Template center vector : Values ​​are five-dimensional real vectors, used to describe the center position of channel features; template tolerance vector The value is a five-dimensional real vector, used to describe the acceptable deviation of a channel from each feature; step size coefficient. : Values Its function is to control the recursive amplitude of the template center; Characteristic steepness coefficient : Values Its function is to control the compressibility of the residual in the hyperbolic tangent; the hyperbolic tangent function : Values Its function is to compress large residuals to suppress abnormal traction; quantile operator : Takes a real number as its value and is used to return the input sample. Percentage points are used as tolerance; quantile parameters : Values Its function is to set the tolerance coverage level; absolute value operator : Values Its function is to convert the residual into a non-negative deviation magnitude; update the allowable quantity. : Values Its function is to control the updating of access control templates; software gate control weights : Values Its function is to update channel weights; clear the signature feature vector. : Values ​​are five-dimensional real vectors, used for template update input; recursive assignment operator The value is a sign, and its function is to update the variable on the left. Specifically, robust recursion and quantile tolerance work together to enable the template to evolve slowly with the operating conditions while maintaining anomaly suppression boundaries.

[0037] For example: After maintenance personnel complete the maintenance and restore the dust removal cycle, they start the reference template database creation state. The edge gateway receives dust removal event records according to the sub-room number, calculates the dust removal signature feature vector for events with valid update permissions, and writes it into the channel template center. When the pressure of the main pipe drops or the condensation margin decreases on site, the event record is still retained as evidence, but the template is not updated because the update permission is invalid. After the environment and gas source are restored, the template continues to be pushed in the existing channel. Maintenance personnel can replay the event record fields before and after the freeze to confirm that the template has not been skewed by non-steady-state events.

[0038] The gating distance and weight calculations are performed in the edge gateway math library; the quantile operator is obtained from the sliding event window through a sorting selection algorithm; the main pipe pressure benchmark and the pre-average differential pressure are obtained through the pre-cleaning segment; the robust matching of the decay time constant is achieved using a one-dimensional search combined with a quasi-Newton solver; the number of channels is increased or decreased due to the gating distance distribution splitting criterion, and the changes are written to the template version field.

[0039] The updated license quantity specifies template update events to retain the source field; the normalization of the clean signature feature vector and robust matching represent the boundary mechanism; soft gating weights reflect changes in operating conditions for channel updates; robust recursion at the template center and quantile tolerance limit anomaly traction to form the usable boundary, outputting the template library. , including template center vector Template tolerance vector and working condition anchor vector This is used in step three to calculate template deviation and spatial distribution classification.

[0040] Step 3: Inherit the dust removal event record set from Step 1 and the template library in step two In the dust removal signature window, the dust removal signature features are matched and accumulated, and the solidification can trigger the alarm record package in step four. According to the template reference, the deviation of the dust removal event is transformed into continuous evidence and the abnormal form is classified in the room dimension.

[0041] Step Two Template Library Already arranged by room number With channel number Give the template center vector Template tolerance vector With soft gating weights During field operation, a single deviation during dust removal may originate from transient disturbances, while a sustained deviation is more likely to be due to localized damage to the filter bag or seal failure; a drop in main pipe pressure and approaching the dew point can cause multiple compartments to deviate synchronously. To make the conclusions actionable, single deviations are extrapolated to continuous evidence, and spatial distribution is used to characterize locality and synchronicity.

[0042] Read the first Sub-room The signature feature vector of the secondary event And call the template center vector of the channel reference. With template tolerance vector Element-wise normalized residuals are generated. Then, log-hyperbolic cosine is used to map the residuals into robust increments, with soft-gated weights. The deviation score is obtained by summing the results of each channel. The details are as follows: ; Where: Deviation score :scope ; Measurement Sub-room The overall deviation of the secondary dust removal event from the template reference; room numbering. : Positive integer, index of room and template entry; event sequence number : Positive integer, indicating the time series position for accumulation; Channel number : Positive integer, index cluster channel reference; number of channels Positive integers, limiting the range of summation; soft-gated weights. :scope And for The summation equals 1, providing the matching weights of the event for each channel reference; (Clear signature feature vector) : The range is a five-dimensional real number vector, carrying characteristics of amplitude, area, hysteresis, recovery, and perturbation; template center vector : A five-dimensional real vector, providing a reference for the channel feature center; template tolerance vector The range is a five-dimensional positive real number vector, which normalizes the residuals element by element and reflects the allowable deviation range; Element-by-element division operator : The range is defined by the operator, which performs division on the corresponding components to form a dimensionless residual; characteristic steepness coefficient :scope Adjusting the steepness of the hyperbolic function mapping; natural logarithm function : Range is real numbers, outputs the cumulative deviation; hyperbolic cosine function :scope Large residuals are compressed while maintaining monotonically increasing; Specifically, tolerance normalization and robust mapping are used to write multi-feature deviations into comparable scalars for subsequent continuous accumulation; beneficial effect: deviation scores are not sensitive to individual peaks and maintain a uniform scale; To avoid triggering non-executable actions due to a single deviation, the deviation score is recursively used to calculate the amount of continuous evidence. The forgetting coefficient is introduced recursively. By attenuating historical evidence and compressing the contribution of over-threshold deviations with hyperbolic tangent, the sustained evidence is made closer to a description of continuously occurring states. Specifically: ; Where: Continuous evidence quantity :scope , characterizing the The sub-room is in The continuous deviation state at the moment of the next event; the forgetting coefficient :scope Control the proportion of historical evidence retained; characteristic steepness coefficient :scope Adjust the steepness of the hyperbolic function mapping; Deviation score :scope Provides single deviation intensity; deviation threshold :scope Define the starting point for sustained evidence and suppress the accumulation of small deviations; hyperbolic tangent function :scope This compresses large deviations and maintains monotonicity; Specifically, by converting single deviations into persistent states, alarms rely more on continuous evidence rather than individual fluctuations. The beneficial effect is that the amount of persistent evidence suppresses one-off fluctuations while preserving the cumulative trend of continuous anomalies. Once persistent evidence is established, it's necessary to determine whether the deviation is concentrated in a few rooms or occurs simultaneously in most rooms. For the same event sequence number... The spatial concentration of continuous evidence in the compartment It uses logarithmic and exponential methods to form scale-invariant difference measures, thus maintaining usable boundaries as compartment sizes change: ; Where: Spatial concentration :scope Characterizes the degree of concentration or dispersion of persistent evidence among compartments; total number of compartments : The range is positive integers, limiting the range of summation; concentration coefficient :scope Adjusting the distinguishing strength of persistent evidence after it enters the index mapping; Sustained Evidence :scope As spatial distribution input; exponential function :scope High-evidence compartmentation; maximum operator : The range is real numbers; the maximum term is extracted to construct a measure of difference; natural logarithm function. The range is for real numbers, mapping the summation terms back to comparable nonnegative quantities; Specifically, spatial concentration separates local anomalies from global synchronization and provides spatial evidence for classification. Beneficial effects: Spatial concentration provides evidence of locality and reduces sensitivity to changes in compartment size.

[0043] During classification and solidification, spatial concentration and continuous evidence quantity are jointly written into the alarm record package. First, when the spatial concentration is below the concentration threshold and at least one compartment has a continuous evidence quantity above the trigger threshold, this moment is solidified as a local anomaly, and the compartment number with the largest continuous evidence quantity is written as the priority investigation target. Second, when the spatial concentration is above the concentration threshold and the continuous evidence quantity of most compartments simultaneously exceeds the trigger threshold, and the condensation margin field in step one remains below the access control requirement in adjacent events, this moment is solidified as a global synchronization state, and the environmental or sensor interference category is written. Third, when most compartments have zero update permission quantity and the main pipe pressure segment is below the field lower limit in adjacent events, and the continuous evidence quantity changes in the same direction, this moment is solidified as a dust removal execution anomaly state, and the gas source link or valve group action prompt field is written. The above solidification does not change the access control logic; instead, it uses the access control fields as classification evidence, making the conclusion interpretable.

[0044] For example, during the operation of a compartment-based pulse-jet baghouse dust collector, the remote interface stores a local anomaly alarm record package for a specific compartment, displaying the compartment number and event sequence number. On-site personnel observe the valve assembly operation and check the compartment door seals at the valve box according to the compartment number. Subsequently, the compartment triggers cleaning again, and the edge gateway recalculates the continuous evidence quantity and no longer stores the local anomaly alarm. In another operation, most compartments were stored as having an abnormal cleaning execution state. On-site personnel confirmed a drop in the main pipe pressure at the air manifold pressure gauge and restored air supply. Subsequent event records were restored to a stable and valid tag of one.

[0045] Deviation score and persistence evidence quantity are calculated on an event-triggered basis at the edge gateway; the maximum compartment number for the compartment persistence evidence quantity is maintained through a max-heap, thus following the event sequence number. Updates are performed without repeated full sorting. Spatial concentration employs a stable process of first extracting the maximum item, then summing, and finally taking the logarithm to avoid exponential overflow. Continuous evidence is stored in a fixed-point format and cached along with event records, allowing for evidence chain replay even during network outages. Concentration and trigger thresholds are bound to the template version number and updated with the template version. Alternative implementations include replacing the recursion of continuous evidence with a combination of median and tail percentile values ​​within a sliding window, or replacing spatial concentration input with deviation scores, while maintaining the constraint that temporal and spatial evidence jointly drive classification.

[0046] The template center and template tolerance normalized residuals of the deviation score are aggregated by soft gating, so that the dust removal signature can be compared under different conditions in each channel; the forgetting recursion of continuous evidence solidifies continuous evidence, so that alarms can be continuous and not single; the spatial concentration characterizes the local area, and works with the access control field in step one, so that local anomalies, global synchronization and dust removal execution anomalies can be distinguished in the same evidence, and the final output is an alarm record package that directly triggers the diagnostic pulse sequence and maintenance action retest judgment in step four.

[0047] Step 4: Based on the abnormal morphology and compartment number solidified in Step 3, organize the diagnostic pulse sequence and cooperate with maintenance actions to complete the retest and judgment, so that the alarm conclusion has a replayable chain of evidence.

[0048] Step 3: Output Deviation Score Sustained amount of evidence Spatial Concentration The alarm log package records local anomalies, global synchronization anomalies, and dust removal execution anomalies. During on-site handling, local anomalies often require verification of the filter bags, door frame seals, and ash hopper status in the corresponding compartment; global synchronization anomalies are often related to condensation margins. Reduce or reduce mother tube pressure slices Changes occur in the same direction; abnormal cleaning behavior is more similar to problems with the air supply chain and valve assembly. Because moisture film, probe contamination bridging, and sensitivity attenuation can alter the cleaning response process of bag leak signals, directly triggering an unpacking inspection based on alarms would increase the frequency of operations; if alarms are blocked, the actual leak handling window would be delayed.

[0049] Therefore, the statistical evidence in step three is transformed into a closed-loop process of proactive triggering, maintenance actions, and retesting confirmation, so that the conclusion is transformed from calculable deviation evidence into on-site actions and retesting results.

[0050] Read the sub-room number from the alarm log packet Event number Deviation score Sustained amount of evidence Spatial concentration Simultaneously read the condensation allowance retained in step one. Update license quantity Pressure slices from the mother tube Then, a diagnostic priority system is constructed. This compresses deviation intensity, synchronization degree, and condensation risk to a uniform scale, thereby avoiding frequent changes in the operation window, as detailed below: ; Where: Diagnostic priority : Value A unified metric used to diagnose operational needs; room numbering. : Value is a positive integer, used to locate the compartment of the diagnostic object; Event number : Value is a positive integer, used to correlate evidence of events before and after the alarm; Deviation score : Value Used to characterize the degree of deviation relative to the template reference; deviation threshold : Value , used to limit the deviation from the reference point for entering the compression mapping; Spatial concentration : Value This is used to characterize the degree of concentration or synchronization of anomalies between compartments; concentration threshold. : Value A benchmark used to limit the degree of synchronization entering the compression mapping; condensation margin. : Takes a real number to characterize temperature and dew point margin and to characterize the risk of dampness; margin threshold : Takes a real number value, used to limit the risk of moisture entering the compression map as a baseline; hyperbolic cosine function : Value , used for robust compression mapping and maintaining monotonicity; Specifically, multiple sources of evidence are combined into a single priority using the same compression function, making the triggering decision subject to both deviation and humidity constraints. When the diagnostic priority exceeds the trigger threshold and the pressure slice of the main pipe is not lower than the field lower limit in the pre-cleaning stage, the compartment is placed in the operation window and written into the evidence package header field. When the condensation margin is lower than the margin threshold, the maintenance action path is entered first, and the inheritance chain of the template update permission is frozen, thereby avoiding mistaking the probe water film effect for filter bag leakage under humid conditions. After the operation window is established, diagnostic pulse sequence parameters are sent to the programmable logic controller. The diagnostic pulse sequence is based on the number of pulses. With pulse interval Fixed time structure, with the first pulse time as the starting point. This serves as an anchor point, enabling the re-tested events to be reproduced in terms of time structure, and allowing the dust removal event identification and time-aligned slicing logic from step one to directly reuse for generating diagnostic event records.

[0051] ; Where: Diagnostic trigger time : Real number, used to define the first The timing anchor point of the secondary diagnostic pulse; the moment of the first pulse. : Real number, used to define the starting anchor point of the diagnostic sequence; pulse number : positive integers and Used to identify the blowing order within a sequence; number of pulses : positive integers and , used to limit the length of the diagnostic sequence; Pulse interval : Used to isolate adjacent diagnostic event windows and avoid slice overlap; compartment numbering : Positive integer, used for locating diagnostic compartments; diagnostic sequence number : A positive integer used to identify the diagnostic round and associate it with retesting; Specifically, deterministic triggering times transform diagnostic stimuli into repeatable inputs, ensuring that differences in retests primarily stem from object state rather than trigger randomness. After each diagnostic pulse-jet trigger, the edge gateway generates a diagnostic event record according to step one and extracts the dust removal signature feature vector according to step two. Then, the deviation score is calculated according to step three to form the diagnostic deviation score. This serves as a diagnostic benchmark before maintenance actions and is included in the evidence package.

[0052] When the alarm log packet indicates a global synchronization status and the condensation margin is below the margin threshold, or when the update permission quantity repeatedly reaches zero within adjacent events, the maintenance action path is initiated first. The maintenance action is completed by the controller triggering the probe purging solenoid valve, or by field personnel cleaning the probe and confirming the completion of the maintenance action at the terminal. After the maintenance is completed, the diagnostic pulse sequence is executed again and a retest event record is generated, and the retest deviation score is calculated. Subsequently, a pullback was constructed. The amount of fallback directly characterizes the deviation change before and after maintenance actions, thus reversibility is used as the criterion for distinguishing between contamination bridging and moisture interference. Specifically: ; Where: the amount of decline : A real number used to characterize the degree of deviation and regression between maintenance actions and retests; diagnostic deviation score. :for Used to maintain diagnostic benchmarks before actions are performed; deviation scores are retested. :for Used for retesting results after maintenance actions; compartment number : A positive integer used to determine the compartment to which the dropout amount belongs; Diagnostic serial number : is a positive integer used to distinguish different closed-loop cycles; Specifically, the amount of regression is used to convert the deviation of maintenance actions into calculable evidence, so that attribution does not rely on verbal judgment.

[0053] When the dropout amount exceeds the dropout threshold and the spatial concentration drops after retesting, the alarm is attributed to probe contamination bridging or a moisture effect, and the maintenance action type and retest conclusion are written into the evidence package. When the dropout amount does not meet the conditions and the local abnormal morphology persists, and the continuous evidence amount in the corresponding compartment does not drop after retesting, the alarm is attributed to continuous leakage on the compartment side, and that compartment remains the priority for investigation. To avoid misjudging sensitivity attenuation as the disappearance of leakage, the dust removal signature feature vector is also read. Normalized peak amplitude With normalized dust removal area and the template center vector The corresponding component alignment comparison; when and If the value is consistently below the template center component and the drop is close to zero across multiple diagnostic sequences, it will be classified as a sensitivity attenuation category and written into the evidence package to trigger subsequent probe maintenance procedures.

[0054] As a supplement, the maintenance action provides three equivalent implementation methods. Probe purging execution method: The controller drives the purging solenoid valve to output, and the purging duration is... Repeatable Once executed, set the maintenance completion flag. Probe heating execution method: The controller drives the heating belt output, and the continuous heating time is specified. After execution, set the maintenance completion flag. Manual cleaning execution method: The platform generates a work order, on-site personnel clean the probe and confirm on the terminal, setting a maintenance completion indicator. .

[0055] The maintenance completion indicator is uniformly abstracted as the maintenance completion indicator. : ; Maintenance Complete Sign : Values Or a positive integer, used to confirm that at least one maintenance action has been completed, thereby allowing a retest to begin.

[0056] For example, during the operation of a bag filter, the remote interface solidified the global synchronization pattern and displayed a decrease in condensation margin. The system included this period in the maintenance action path and triggered the probe purging solenoid valve. After purging, the controller triggered a pulse jet to the same compartment according to the diagnostic pulse sequence. The edge gateway generated a retest event record and calculated the fallback amount. The interface showed a deviation from the fallback and a decrease in spatial concentration. Based on this, the maintenance personnel classified the alarm as a damp or contamination bridge and completed the recording. In another operation, the interface solidified a local abnormal pattern and locked a certain compartment number. After the system executed the diagnostic pulse sequence, the fallback amount did not meet the conditions, and the continuous evidence amount of that compartment remained unchanged. On-site personnel checked the seal at the corresponding compartment door frame and replaced the damaged filter bag. Subsequently, the event record was restored to a stable and valid tag of one.

[0057] Diagnostic priority and fallback calculations are performed at the edge gateway. Trigger thresholds and fallback thresholds are bound to template version fields and can be configured on the maintenance terminal. Diagnostic pulse sequence parameters are issued through the controller's sequential function block and interlocked to limit the main pipe pressure and maintenance status. The maintenance action path switches between purging the solenoid valve, probe heating band, and manual cleaning confirmation, and is written into the evidence package with the same maintenance action field. The retest deviation score reuses the calculation paths of steps two and three to ensure comparability. Diagnostic priority compresses the deviation score, spatial concentration, and condensation margin into a unified scale, making the trigger decision subject to both deviation and humidity. The deterministic trigger time sequence makes the diagnostic excitation repeatable and consistent with the event slice boundary of step one. The fallback amount solidifies the deviation changes before and after the maintenance action and supports pollution bridging and humidity interference attribution. The alignment comparison of the normalized peak amplitude and the normalized cleaning area distinguishes the sensitivity decay and writes it into the evidence package. Finally, an evidence package containing diagnostic event records, retest event records, maintenance action fields, and attribution conclusions is formed, allowing alarm conclusions to be replayed and used for subsequent work order closure.

[0058] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0059] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0060] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0061] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0062] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A remote intelligent operation and maintenance early warning system for dust removal equipment based on the Industrial Internet of Things, characterized in that: include, The edge gateway synchronously collects bag leak monitoring signals, dust cleaning trigger signals, compressed air main pipe pressure, dust collector differential pressure, temperature and humidity, establishes a dust cleaning signature window according to the dust cleaning trigger signal, and slices it to form a dust cleaning event record, assigning stability and validity labels; Extract the dust removal signature features from the dust removal event records, establish a template library according to the compartment or valve group and operating condition cluster, and update the template library through the trusted access control when the stability and validity labels are satisfied and the condensation risk is satisfied. The deviation score is obtained by comparing the dust removal signature features with the corresponding template. Combined with the time persistence determination and spatial distribution pattern classification, alarm type, alarm level and suspected room location suggestions are output. When the alarm type is uncertain due to environmental or sensor abnormalities, a diagnostic pulse sequence is executed for the selected compartment or valve group, triggering maintenance actions, followed by retesting and generating an evidence package.

2. The remote intelligent operation and maintenance early warning system for dust removal equipment according to claim 1, characterized in that: The edge gateway provides unified time synchronization and timestamp annotation for bag leak monitoring signals, dust cleaning trigger signals, compressed air header pressure, dust collector differential pressure, temperature and humidity, and marks missing and abnormal points and caches the original data, and writes the timestamps and marks into the dust cleaning event record.

3. The remote intelligent operation and maintenance early warning system for dust removal equipment according to claim 2, characterized in that: The dust removal trigger signal originates from the pulse controller and is differentiated by compartment. The edge gateway performs de-jitter processing on the dust removal trigger signal. The de-jitter processing includes removing pulses with insufficient width and merging repeated pulses within a short period of time. A dust removal event record is established for each valid pulse, and the event number, occurrence time and corresponding compartment number are recorded.

4. The remote intelligent operation and maintenance early warning system for dust removal equipment according to claim 3, characterized in that: The dust removal signature window covers the stable section before dust removal and the peak and attenuation section after dust removal. The length of the dust removal signature window is set by the preset window length parameter and corrected according to the historical peak arrival time. The stability and validity labels are jointly determined by the fluctuation of the dust collector pressure difference within the window, whether the compressed air main pipe pressure is lower than the safety lower limit, and the equipment start-up and shutdown status. Unstable events do not participate in the template library update.

5. The remote intelligent operation and maintenance early warning system for dust removal equipment according to claim 4, characterized in that: The dust removal signature features consist of the peak amplitude of dust removal, the peak area of ​​dust removal, the peak arrival time, the peak decay rate, the peak decay duration, the waveform similarity, and the high-frequency noise intensity. The dust removal intensity-related features are normalized by the compressed air main pipe pressure, and the dust removal signature features are compared in segments by combining the dust collector differential pressure level and the dust collector differential pressure change trend.

6. The remote intelligent operation and maintenance early warning system for dust removal equipment according to claim 5, characterized in that: The template library adopts a two-layer structure. The first layer establishes basic templates according to the room number, and the second layer establishes multiple templates according to the working condition clusters. The working condition clusters are determined by the dust collector pressure difference range, production load range, and temperature and humidity corresponding ranges. During operation, the working condition clusters are determined first and then the corresponding multiple templates are called, and the working condition clusters and calling conditions are written into the template library.

7. The remote intelligent operation and maintenance early warning system for dust removal equipment according to claim 6, characterized in that: The access control system is updated based on the following conditions: stability and validity labels, condensation risk meeting preset thresholds, no current leakage alarms, and sensor health status meeting preset thresholds. The template library update is frozen when a suspected contamination bridging is detected, and the template library update is frozen when an abnormal sensor health status is detected. The template update process is also managed and backtracked.

8. The remote intelligent operation and maintenance early warning system for dust removal equipment according to claim 7, characterized in that: The duration determination uses the most recent preset number of dust removal event records as the observation window, and triggers an alarm when the proportion of deviation score exceeding the limit reaches a preset proportion threshold; When the deviation score exceeds the limit for a preset number of consecutive times, the alarm level is upgraded step by step; the spatial distribution pattern classification is determined based on the synchronization range and continuous abnormal range of the deviation score between rooms.

9. The remote intelligent operation and maintenance early warning system for dust removal equipment according to claim 8, characterized in that: The spatial distribution pattern classification includes local anomaly patterns, global synchronous anomaly patterns, and dust removal system anomaly patterns. Among them, the local anomaly pattern corresponds to the proportion of compartments with continuously abnormal deviation scores being lower than the first threshold and outputting a suspected compartment ranking. The global synchronous anomaly pattern corresponds to the proportion of compartments with synchronously rising deviation scores being higher than the second threshold and synchronous with condensation risk. The dust removal system anomaly pattern corresponds to the consistent change in dust removal response pattern synchronous with the decrease in compressed air main pipe pressure.

10. The remote intelligent operation and maintenance early warning system for dust removal equipment according to claim 9, characterized in that: When the alarm type is uncertain due to environmental or sensor abnormality, the diagnostic pulse sequence is executed in a preset order for the selected compartment and the corresponding dust removal signature characteristics are recorded. After triggering a short-term purging of the probe part, the retest diagnostic pulse sequence is immediately executed and an evidence package is generated. The evidence package includes the dust removal event sequence before and after the alarm, the deviation change record, the condensation risk status, the maintenance action record, and the retest results.

Images