A remote operation and maintenance and fault processing method for a supply system

Abstract

The present application relates to the technical field of industrial feeding system process control and fault monitoring, and particularly relates to a feeding system remote operation and maintenance and fault processing method, comprising the following steps: collecting visual data of key equipment of a feeding system and corresponding PLC / DCS working condition data, identifying abnormalities by an edge model, generating a fault event in combination with visual and working condition abnormalities within a time window; a fault processing unit forms a hierarchical control request, which is executed or rejected after PLC / DCS interlocking judgment, and the event state is updated according to the returned result, and the record is uploaded to a remote operation and maintenance platform. The present application generates a feeding fault event by associating visual abnormalities of key equipment and PLC / DCS working condition abnormalities within a preset time window, reduces dust shielding and single frame misidentification and mistriggering; the fault processing unit generates a hierarchical control request, which is executed or rejected after PLC / DCS interlocking decision, and the event state is updated according to the change of visual and working condition before and after disposal, and the record is uploaded to a remote operation and maintenance platform, forming a complete process of identification, decision, feedback, verification, record and optimization.

Description

Technical Field

[0001] This invention relates to the field of process control and fault monitoring technology for industrial material supply systems, specifically a method for remote operation and maintenance and fault handling of material supply systems. Background Technology

[0002] In the feeding system of powder, granular, lumpy materials and mixed materials, abnormalities such as bridging of silos, interruption of feeding, blockage of pipelines, valves, belt conveyors and safety gates are prone to occur.

[0003] In the existing technology, there are solutions that detect material flow status, belt misalignment, abnormal material dropping, or local blockage by means of cameras or image recognition. These solutions mainly rely on visual results for alarms and are easily affected by dust, light, camera dirt, short-term splashes, or obstruction, which may lead to false alarms or malfunctions.

[0004] In existing technologies, there are also solutions that collect equipment operating status, pressure, flow, current, valve position and alarm signals through PLC / DCS or industrial gateways and upload them to a remote platform for monitoring. However, PLC / DCS operating data is difficult to directly reflect the actual flow pattern of materials, blockage areas, material spillage status or personnel entry status. It is also difficult to accurately distinguish material shortage, bridging, flow interruption, blockage, valve jamming or sensor malfunction based solely on operating data.

[0005] Furthermore, if the remote platform directly triggers control actions based on the diagnostic results without the PLC / DCS performing interlock judgment according to the target equipment and control level, inappropriate actions may be performed during maintenance, emergency stop, safety door opening, personnel entry, or downstream material receiving not being permitted. Existing systems also typically lack post-action effect verification and status update mechanisms based on the same fault event.

[0006] Therefore, a remote operation and maintenance and fault handling method for the material supply system is needed. This method would correlate visual anomalies with PLC / DCS operating anomalies, perform graded handling under PLC / DCS interlock constraints, and then determine the handling result based on changes in visual and operating data before and after handling, thus forming a complete fault handling process. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a method for remote operation and maintenance and fault handling of a material supply system.

[0008] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a method for remote operation and maintenance and fault handling of a material feeding system, comprising the following steps: S1: Collect visual data of key equipment locations in the feeding system and obtain PLC / DCS operating data corresponding to the key equipment locations; S2: The visual data is identified using a material supply fault identification model deployed on an edge computing device to obtain visual identification results; S3: The fault handling unit generates a material supply fault event based on the correspondence between the visual anomaly features represented by the visual recognition results and the working condition anomalies represented by the PLC / DCS working condition data within a preset time window. S4: The fault handling unit generates a graded control request based on the material supply fault event and sends the graded control request to the PLC / DCS; S5: The PLC / DCS determines whether the set of interlocking conditions corresponding to the hierarchical control request meets the execution conditions, and executes the hierarchical control request or refuses to execute the hierarchical control request according to the judgment result, and returns the PLC / DCS return result to the fault handling unit. S6: The fault handling unit updates the event status of the material feeding fault event based on the result returned by the PLC / DCS; When the PLC / DCS returns a result indicating that a fault handling action has been performed, the fault handling unit determines the event handling result based on the changes in visual data and PLC / DCS operating data before and after the fault handling action, and updates the event status to one of the following based on the event handling result: closed, continued monitoring, pending escalation, pending manual confirmation, or pending shutdown control. S7: Upload the processing data of the material supply failure event to the remote operation and maintenance platform, and the remote operation and maintenance platform will form an event record; The processing data includes at least one of the following: material supply failure events, hierarchical control requests, PLC / DCS return results, event processing results, and event status.

[0009] Furthermore, the key equipment locations include at least one of the following: silo location, feeding equipment location, conveying pipeline location, valve location, belt conveyor location, personnel entry monitoring location, equipment external protection location, or safety door monitoring location; The PLC / DCS operating data includes at least one of the following: material supply status data, conveying status data, actuator status data, safety interlock status data, and upstream and downstream coordination status data. The material supply status data includes material level data, weighing flow rate data, or loss-in-weight scale flow rate data. The conveying status data includes the operating status of the feeding equipment, the current of the feeding equipment, the rotation speed of the feeding equipment, the torque of the feeding equipment, the conveying pressure, the conveying pressure difference, the air volume or the belt speed; The actuator status data includes valve opening data, air hammer status, vibrator status, purge valve status, or arch-breaking device status. The safety interlock status data includes automatic operation status, maintenance status, emergency stop status, safety door status, safety light curtain status, personnel entry status, automatic control freeze status, or PLC / DCS prohibition signal; The upstream and downstream collaborative status data includes the operating status of upstream equipment, the downstream material receiving status, and the availability status of backup branch lines or backup pipelines.

[0010] Furthermore, before collecting visual data, establish the correspondence between cameras, key equipment locations, regions of interest, PLC / DCS variable addresses, and fault types; Based on the correspondence, the edge computing device calls the corresponding region of interest for visual data of different key device locations, and calls the recognition model corresponding to the fault type and the corresponding recognition threshold or post-processing rules.

[0011] Furthermore, in step S3, generating a material supply failure event includes: Determine visual confidence level based on visual recognition results; Determine temporal consistency based on visual recognition results of multiple consecutive frames; The operating condition matching result is determined based on the matching relationship between PLC / DCS operating condition data and corresponding fault types; Based on the visual confidence score, time consistency, and working condition matching results, a comprehensive fault confidence score is generated according to a preset fusion rule. When the overall fault confidence level meets the preset triggering conditions, the material supply fault event is generated.

[0012] Furthermore, the material supply failure events include at least one of the following: hopper bridging event, rat hole flow event, material supply interruption event, conveying pipeline blockage event, valve jamming event, belt misalignment event, material spillage event, and personnel entering a hazardous area event; The material supply failure event is determined by visual anomalies and corresponding PLC / DCS operating anomalies within the same preset time window.

[0013] Furthermore, the fault handling unit is deployed on the edge computing device or the remote operation and maintenance platform; The hierarchical control request includes at least one of the following: fine-tuning level control request, switching level control request, load reduction level control request, and shutdown level control request; The fine-tuning level control request is used to trigger air hammer, vibration, purging, low-speed pulse, or feed rate adjustment; The switching level control request is used to switch backup branches, backup pipelines, backup valves, or backup feeding equipment; The deload control request is used to reduce the feeding speed, reduce the conveying load, or suspend upstream material supply. The shutdown-level control request is used to stop the relevant feeding equipment, conveying equipment, or trigger a safety shutdown.

[0014] Furthermore, the interlock condition set is a set of interlock conditions pre-configured for the target device based on the control level of the hierarchical control request; The interlock condition set includes one or more of the following: automatic operation state, maintenance state, emergency stop state, safety door state, safety light curtain state, personnel entry state, downstream material receiving state, target actuator state, and PLC / DCS prohibition signal; The PLC / DCS return results include execution feedback or reasons for rejection; When the set of interlocking conditions meets the execution conditions, the PLC / DCS executes the corresponding fault handling action and returns the execution feedback to the fault handling unit. When the set of interlocking conditions does not meet the execution conditions, the PLC / DCS refuses to execute the hierarchical control request and returns the reason for the refusal to the fault handling unit.

[0015] Furthermore, in step S6, the event handling result is determined by comparing and contrasting the changes in the fault area, the changes in the material flow state, the changes in the feedback from the actuator, and the changes in the PLC / DCS operating data before and after the fault handling action. When the event processing result meets the recovery conditions, the fault processing unit updates the event status to closed; When the event handling result does not meet the recovery conditions and does not reach the shutdown trigger condition, the fault handling unit updates the event status to continue monitoring or pending upgrade handling. When the event status is updated to pending escalation, the fault handling unit regenerates the corresponding level of hierarchical control request and sends it to the PLC / DCS; When the event processing result meets the shutdown triggering condition, the fault processing unit updates the event status to pending shutdown control and generates a shutdown-level control request to be sent to the PLC / DCS. When the PLC / DCS returns a rejection reason, the fault handling unit updates the event status to pending manual confirmation or continued monitoring.

[0016] Furthermore, the material supply fault identification model is obtained in the following way: Obtain real fault samples, normal operating condition samples, and manually confirmed samples of the material feeding system; Using the real fault samples, and combining them with simulation samples or enhanced samples, a visual fault sample library is constructed. Based on the aforementioned visual fault sample library, the pre-trained visual model is trained, fine-tuned, or its parameters are adjusted to obtain a material feeding fault identification model for identifying material feeding faults. The material supply fault identification model is deployed to an edge computing device.

[0017] Furthermore, the remote operation and maintenance platform generates a feedback sample based on the event records; The returned samples include at least one of the following: samples where the visual recognition results are inconsistent with the manual confirmation results; samples that have not been recovered after fault handling actions; samples where the same key equipment location is repeatedly generated within a preset time range and has been manually confirmed not to be faulty; samples with a confidence level lower than a preset confidence threshold; and paired samples before and after fault handling actions. When the event status of the material supply failure event is pending manual confirmation, the remote operation and maintenance platform receives the manual confirmation result and sends the manual confirmation result to the fault handling unit, so that the fault handling unit closes the material supply failure event, continues monitoring, or regenerates the hierarchical control request based on the manual confirmation result. The remote operation and maintenance platform updates the material supply fault identification model or fault handling strategy based on the returned samples, and sends the updated material supply fault identification model or fault handling strategy to the edge computing device. The fault handling strategy includes at least one of the following: material supply fault event triggering conditions, fault level classification rules, hierarchical control request generation rules, action number limit, action cooldown time, manual confirmation conditions, and shutdown triggering conditions.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention correlates visual anomalies at key equipment locations with corresponding PLC / DCS operational anomalies within a preset time window. As a result, material supply failure events are not directly triggered by a single image frame, a single visual threshold, or a single alarm signal. Instead, they require simultaneous fulfillment of correlation conditions on both the visual and operational sides. This reduces false triggering caused by dust obstruction, changes in lighting, short-term material splashing, camera contamination, or misidentification of a single frame.

[0019] This invention introduces visual data from key equipment locations and uses a material supply fault identification model to identify visual anomalies such as material flow status, blockage areas, spillage areas, valve operation status, and personnel entering dangerous areas. As a result, flow rate decreases, differential pressure increases, current changes, or valve position anomalies in PLC / DCS operating data can be correlated with the on-site visual status, making it easier to distinguish between different causes such as material shortage, bridging, flow interruption, blockage, valve jamming, and sensor malfunction.

[0020] This invention generates hierarchical control requests through a fault handling unit, and the PLC / DCS determines whether the execution conditions are met based on the control level and the set of interlock conditions corresponding to the target equipment. Thus, actions such as air hammer, vibration, purging, switching, unloading, or shutdown are all under the safety interlock constraints of the PLC / DCS, which can prevent inappropriate actions from being performed in maintenance, emergency stop, when safety doors are open, personnel enter, downstream material receiving is not allowed, or when the PLC / DCS prohibition signal is valid.

[0021] After the PLC / DCS returns that the fault handling action has been executed, this invention continues to compare the changes in visual data and PLC / DCS operating data before and after the action, and judges the event handling result accordingly. Thus, the event status can be updated to closed, continue monitoring, pending escalation handling, pending manual confirmation, or pending shutdown control according to the actual recovery situation, so that the fault handling forms a complete system of identification, request, interlock decision, action feedback, effect verification, and status update.

[0022] This invention takes material supply failure events as a unified object and records visual anomaly features, abnormal operating conditions, hierarchical control requests, PLC / DCS return results, event handling results, and event status in association. As a result, the remote operation and maintenance platform can trace the cause of the failure, the handling actions, the reasons for interlock rejection, the recovery status after the action, and the subsequent sample return status based on the same event. It also provides a sample basis for subsequent correction of identification thresholds, updating of material supply failure identification models, and adjustment of failure handling strategies.

[0023] Furthermore, when image quality is abnormal, the present invention does not directly generate automatic handling graded control requests, but generates image quality abnormality prompts, continued monitoring status, or pending manual confirmation status, thereby reducing erroneous handling actions caused by camera contamination, dust obstruction, strong reflection, or screen occlusion. Attached Figure Description

[0024] Figure 1 This is a flowchart of the method steps of the present invention. Detailed Implementation

[0025] 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.

[0026] This embodiment focuses on a powder feeding system, which includes a silo, feeding equipment, loss-in-weight scale, conveying pipeline, valves, belt conveyor, air hammer, vibrator, purge valve, safety door, and personnel access monitoring area. A PLC / DCS is used to collect data on feeding status, conveying status, actuator status, safety interlock status, and upstream and downstream coordination status. An edge computing device is used to receive visual data and run a feeding fault identification model. A fault handling unit is deployed on the edge computing device and communicates with the PLC / DCS and remote operation and maintenance platform.

[0027] In this embodiment, the personnel entry monitoring point is only used for the safety protection of the feeding equipment. The edge computing device only determines whether a personnel target has entered the preset danger zone; it does not perform facial recognition, extract personnel identity information, or make differentiated control decisions based on personnel identity.

[0028] Before collecting visual data, establish the correspondence between cameras, key equipment locations, regions of interest, PLC / DCS variable addresses, and fault types.

[0029] Specifically, a first camera is installed at the material outlet of the silo, a second camera at the observation section of the conveyor pipeline, a third camera at the valve actuator, a fourth camera at the material drop point of the belt conveyor, and a fifth camera at the safety gate and personnel entry hazardous area. Each camera corresponds to a critical equipment location, and each critical equipment location has at least one region of interest. Each region of interest is bound to a corresponding fault type and PLC / DCS variable address.

[0030] For example, the region of interest at the silo outlet corresponds to silo bridging events and feed interruption events, and the bound PLC / DCS variables include material level, loss-in-weight scale flow rate, feeder operating status, feeder speed, air hammer status, and vibrator status. The region of interest in the observation section of the conveying pipeline corresponds to the conveying pipeline blockage event, and the bound PLC / DCS variables include conveying pressure, conveying differential pressure, air volume, purge valve status, and downstream allowable material receiving status. The valve location corresponds to the valve jamming event, and is bound to the valve opening degree, switch position signal, and PLC / DCS valve command. The belt conveyor location corresponds to the belt misalignment event and spillage event, and is bound to the belt speed, belt operating status, and misalignment switch status. The personnel entry monitoring location corresponds to the personnel entering the danger zone event, and is bound to the automatic operation status, safety door status, safety light curtain status, personnel entry status, emergency stop status, and PLC / DCS prohibition signal.

[0031] After the edge computing device loads the above correspondence, it determines the region of interest to be processed based on the camera number, key equipment location number, and region of interest number. Then, it calls the corresponding material supply fault identification model, identification threshold, and post-processing rules according to the fault type bound to the region of interest. The post-processing rules include continuous frame consistency judgment, abnormal area filtering, image quality filtering, and suppression of repeated triggering of the same event.

[0032] Please see Figure 1 This invention provides a method for remote operation and maintenance and fault handling of a material supply system, comprising the following steps: S1: Collect visual data of key equipment locations in the feeding system and obtain PLC / DCS operating data corresponding to the key equipment locations.

[0033] The edge computing device receives video streams from each camera and extracts region of interest (ROI) images from the video streams. Each frame of the image records the acquisition time, camera number, key device location number, and ROI number.

[0034] PLC / DCS operating data is read through the industrial communication interface. The read data includes material level data, loss-in-weight scale flow data, feeding equipment operating status, feeding equipment current, feeding equipment speed, conveying pressure, conveying pressure differential, air volume, valve opening data, air hammer status, vibrator status, purge valve status, automatic operation status, maintenance status, emergency stop status, safety door status, safety light curtain status, personnel entry status, PLC / DCS prohibition signals, downstream material receiving status, and backup pipeline availability status.

[0035] To ensure that visual data corresponds with PLC / DCS operating data, the edge computing device and PLC / DCS use the same timestamp, and both visual data and PLC / DCS operating data are entered into the fault handling unit according to the acquisition time.

[0036] S2: Use the material supply fault identification model deployed on the edge computing device to identify visual data and obtain visual recognition results.

[0037] The material feeding fault identification model takes multiple consecutive frames of region of interest images as input and outputs one or more of the following: fault type, visual confidence, abnormal area location, abnormal area area ratio, material flow status, target entry status, and image quality status.

[0038] In this embodiment, the material supply fault identification model includes an image preprocessing module, a feature extraction module, a target detection module, a semantic segmentation module, a time sequence judgment module, and a result output module.

[0039] The image preprocessing module is used to perform size normalization, brightness normalization, noise suppression, and image quality detection on the region of interest image. The image quality detection is used to determine whether there is dust obstruction, condensation, oil stains, strong reflection, overexposure, underexposure, or obstruction in the image.

[0040] The feature extraction module extracts material edges, textures, local motion, and equipment structural features. The target detection module identifies personnel targets, material spillage areas, valve actuators, and belt edges. The semantic segmentation module outputs masks for stationary material areas at silo outlets, discharge cavity areas, material accumulation areas in conveyor pipes, or material spillage areas outside the belt. The temporal sequence determination module determines temporal consistency based on changes in abnormal regions and material flow in multiple consecutive frames of images. The result output module outputs the visual recognition results.

[0041] For example, in the case of a silo bridging event, the model identifies whether there are multiple consecutive frames of no obvious material flow, a stable arched boundary above the discharge port, or a static material area in the silo outlet area. In the case of a conveying pipeline blockage event, the model identifies whether there is a persistent material accumulation area, an increase in the material accumulation area, or a decrease in the material flow rate in the transparent observation section. In the case of a person entering a dangerous area, the model identifies whether the person has entered a preset dangerous area.

[0042] S3: The fault handling unit generates a material feeding fault event based on the correspondence between the visual anomaly features represented by the visual recognition results and the working condition anomalies represented by the PLC / DCS working condition data within a preset time window.

[0043] In this embodiment, abnormal operating conditions can be directly provided by the PLC / DCS in the form of alarm bits, status bits, fault codes, interlock status, or variable quality flags, or they can be determined by the fault handling unit by comparing the PLC / DCS operating condition data with preset thresholds, normal ranges, trend changes, or interlock status.

[0044] For example, when the PLC / DCS directly outputs alarms such as high differential pressure, valve not in position, safety door open, emergency stop, or downstream material receiving not allowed, the fault handling unit can use these as abnormal operating conditions for judgment.

[0045] When the PLC / DCS does not directly output a fault flag, the fault handling unit determines the abnormal operating condition based on the PLC / DCS operating condition data. For example, in a hopper bridging event, if the material level is higher than the low material level threshold, the feeding equipment is in operation, and the loss-in-weight scale flow rate is lower than 30% to 50% of the set feeding amount and continues for more than a preset time, then an abnormal operating condition corresponding to hopper bridging is determined. In a conveying pipeline blockage event, if the conveying pressure difference exceeds the blockage alarm threshold, the conveying pressure increases, and the air volume decreases, then an abnormal operating condition corresponding to conveying pipeline blockage is determined. In a valve jamming event, if the PLC / DCS has issued a valve action command, but the valve opening degree or switch position signal does not reach the target state within a preset time, then an abnormal operating condition corresponding to valve jamming is determined.

[0046] The correspondence between visual anomalies and operational anomalies within a preset time window includes correspondences based on equipment location, fault type, time, and variable logic.

[0047] Equipment location correspondence means that visual data comes from a key equipment location, and PLC / DCS operating condition data also comes from that key equipment location or related upstream and downstream equipment. Fault type correspondence means that visual anomalies and operating condition anomalies both match the same fault type. Time correspondence means that the occurrence time of visual anomalies and the occurrence time of operating condition anomalies fall within the same preset time window. Variable logic correspondence means that PLC / DCS operating condition anomalies and visual anomalies satisfy a preset logical relationship.

[0048] In this embodiment, the fault handling unit generates a comprehensive fault confidence score based on visual confidence score, time consistency, and operating condition matching results.

[0049] Visual confidence is denoted as Output from the material supply failure identification model, time consistency is denoted as This is used to indicate whether the visual recognition results of multiple consecutive frames are stable; the working condition matching result is denoted as... This is used to represent the degree of matching between PLC / DCS operating condition data and corresponding fault types. The overall fault confidence level is denoted as... And calculate according to the following formula: , in, , , These are the weighting coefficients, and + + =1, for example, Take 0.45, Take 0.25, Take 0.30.

[0050] When the visual confidence level is not less than the visual threshold, the time consistency is not less than the time consistency threshold, the working condition matching result is not less than the working condition matching threshold, and the comprehensive fault confidence level is not less than the comprehensive triggering threshold, the fault handling unit generates a material supply fault event.

[0051] In this embodiment, the preset time window, preset fusion rules, preset trigger conditions, and various thresholds are all pre-configured based on key equipment locations, fault types, material characteristics, equipment action risks, PLC / DCS operating condition variables, and on-site safety requirements, and can be corrected through on-site calibration samples, manually confirmed samples, or return samples.

[0052] Material feeding failure events include event number, critical equipment location, failure type, occurrence time, visual recognition results, associated PLC / DCS operating data, overall failure confidence level, recommended control level, and event status.

[0053] In this embodiment, the material supply failure events that can be generated include hopper bridging events, rat hole flow events, material supply interruption events, conveyor pipe blockage events, valve jamming events, belt misalignment events, material spillage events, and personnel entering dangerous areas events.

[0054] S4: The fault handling unit generates a graded control request based on the material feeding fault event and sends the graded control request to the PLC / DCS.

[0055] The hierarchical control request includes the event number, critical equipment location number, fault type, control level, target equipment number, action type, action parameters, request time, action duration, and action count number.

[0056] The hierarchical control requests include fine-tuning level control requests, switching level control requests, load reduction level control requests, and shutdown level control requests.

[0057] Fine-tuning level control requests are used to trigger air hammers, vibrations, purging, low-speed pulses, or feed rate adjustments. For example, in the initial stage of hopper bridging, the fault handling unit generates an air hammer action request, with an action duration of 3 seconds and an upper limit of 3 actions. When there is slight material accumulation in the conveying pipeline, the fault handling unit generates a short-term purging request.

[0058] Switching-level control requests are used to switch backup branches, backup pipelines, backup valves, or backup feeding equipment. For example, when the current conveying pipeline is blocked and the backup pipeline is available, the fault handling unit generates a backup pipeline switching request.

[0059] The load reduction control request is used to reduce the feed rate, reduce the conveying load, or suspend upstream material supply. For example, when the conveying differential pressure increases but does not reach the shutdown trigger condition, the fault handling unit generates a request to reduce the feed rate.

[0060] A shutdown-level control request is used to stop related feeding equipment, conveying equipment, or trigger a safety shutdown. For example, when personnel enter a hazardous area and the equipment is still in automatic operation, the fault handling unit generates a shutdown-level control request.

[0061] S5: The PLC / DCS determines whether the set of interlocking conditions corresponding to the hierarchical control request meets the execution conditions, and executes or refuses to execute the hierarchical control request based on the judgment result, and returns the result from the PLC / DCS to the fault handling unit.

[0062] The PLC / DCS invokes a pre-configured set of interlocking conditions based on the control level and target device in the hierarchical control request.

[0063] For fine-tuning level control requests, the interlock condition set includes automatic operation status, non-maintenance status, no emergency stop status, safety door closed, safety light curtain not triggered, personnel entry status, target actuator without fault, air source pressure meeting requirements, and PLC / DCS prohibition signal.

[0064] For a switching-level control request, the interlock condition set includes: backup branch available, backup valve fault-free, downstream material receiving permitted, current pipeline closure permitted, and normal feedback from the target actuator.

[0065] For load reduction control requests, the interlock condition set includes the target feeder being controllable, the automatic operation status being valid, the target actuator being able to receive settings, and the upstream and downstream coordination status allowing load reduction.

[0066] For shutdown-level control requests, the interlock condition set is configured according to the safe shutdown logic. Personnel entry, safety door opening, or safety light curtain triggering can be used as reasons to trigger shutdown, rather than reasons to refuse shutdown. The PLC / DCS still judges the validity of the shutdown link, the disconnectable status of the target equipment, and the safety status of upstream and downstream after shutdown.

[0067] When the interlock condition set meets the execution conditions, the PLC / DCS executes the corresponding fault handling action and returns execution feedback to the fault handling unit. The execution feedback includes whether the action has started, the action has been completed, the action has failed, or the action has timed out.

[0068] When the interlock condition set does not meet the execution conditions, the PLC / DCS refuses to execute the hierarchical control request and returns the reason for refusal to the fault handling unit. The reasons for refusal include maintenance status, emergency stop status, safety door open, safety light curtain triggered, personnel entering, downstream material receiving not allowed, target actuator failure, insufficient air supply pressure, or PLC / DCS prohibition signal being valid.

[0069] After receiving the rejection reason from the PLC / DCS, the fault handling unit does not overwrite or modify the interlock logic on the PLC / DCS side, nor does it repeatedly send hierarchical control requests with the same action parameters. Instead, it updates the event status according to the rejection reason.

[0070] S6: The fault handling unit updates the event status of the material feeding fault event based on the results returned by the PLC / DCS.

[0071] When the PLC / DCS returns a result indicating that the fault handling action has been executed, the fault handling unit judges the event handling result based on the changes in visual data and PLC / DCS operating data before and after the fault handling action.

[0072] The event handling result is determined based on changes in the fault area, changes in material flow status, changes in actuator feedback, and changes in PLC / DCS operating data. Changes in the fault area can be determined by the percentage of abnormal area, changes in the boundary of abnormal area, or the pixel ratio of abnormal area. Changes in material flow status can be determined by the continuous frame texture change rate, optical flow velocity, or inter-frame difference of the feeding area. Changes in actuator feedback can be determined by valve opening, air hammer status, vibrator status, or purge valve status. Changes in PLC / DCS operating data can be determined by loss-in-weight scale flow rate, weighing flow rate, conveying pressure, conveying differential pressure, air volume, current, rotational speed, or belt speed.

[0073] In this embodiment, the recovery conditions are objective judgment conditions pre-configured based on the fault type of the material supply failure event, the location of the critical equipment, the type of fault handling action, and the corresponding PLC / DCS operating condition variables. The recovery conditions include the change of the fault area reaching the preset recovery range, the material flow state returning to the preset normal state, the actuator feedback reaching the target state, the PLC / DCS operating condition data returning to the preset normal range, the safety interlock state returning to the safe state, or a combination of the above conditions.

[0074] For example, the recovery conditions for a hopper bridging event are: the area of ​​the abnormal region at the discharge port decreases by no less than 60% after the action, the confidence level of normal visual flow is no less than 0.75, and the loss-in-weight scale flow rate reaches more than 80% of the set feed rate and lasts for no less than 20 seconds.

[0075] The recovery conditions for a blockage event in the conveying pipeline are: the conveying pressure differential drops to below 90% of the blockage alarm threshold, the area of ​​the material accumulation zone decreases by no less than 60%, and the air volume recovers to above the normal lower limit.

[0076] The recovery condition for a belt misalignment event is: the offset of the belt edge relative to the baseline is less than the misalignment alarm distance threshold, and this offset is maintained for at least 10 seconds.

[0077] The recovery conditions for personnel entering a dangerous area are: the personnel target leaves the preset dangerous area, the safety door or safety light curtain status returns to normal, and the PLC / DCS confirms that the relevant equipment is in a safe state.

[0078] When the event handling result meets the recovery conditions, the fault handling unit updates the event status to closed.

[0079] When the event handling result does not meet the recovery conditions but the fault trend weakens, the fault handling unit updates the event status to continue monitoring.

[0080] When the event handling result does not meet the recovery conditions and the fault trend does not improve significantly, but the shutdown trigger condition is not met, the fault handling unit updates the event status to pending escalation and regenerates the corresponding level of hierarchical control request to send to the PLC / DCS.

[0081] When the event processing result meets the shutdown triggering conditions, the fault processing unit updates the event status to pending shutdown control and generates a shutdown-level control request to be sent to the PLC / DCS. The shutdown triggering conditions include the comprehensive fault confidence level continuously exceeding the shutdown confidence threshold, the same event fine-tuning level action reaching the upper limit of the number of times it has not recovered, the conveyor pressure difference exceeding the safety limit, personnel entering the dangerous area while the equipment is still running, the belt seriously deviating beyond the shutdown distance threshold, or the PLC / DCS returning a critical actuator failure.

[0082] When the PLC / DCS returns a rejection reason, the fault handling unit updates the event status to pending manual confirmation or continued monitoring based on the rejection reason. For example, if the rejection reason is maintenance status, safety door open, personnel entering, or target actuator malfunction, the event status is updated to pending manual confirmation.

[0083] S7: Upload the processing data of the material supply failure event to the remote operation and maintenance platform, and the remote operation and maintenance platform will form an event record.

[0084] The processing data includes at least one of the following: material supply failure events, hierarchical control requests, PLC / DCS return results, event processing results, and event status. The event records generated by the remote operation and maintenance platform include event number, occurrence location, fault type, trigger time, visual recognition results, associated PLC / DCS operating data, hierarchical control requests, PLC / DCS execution feedback or rejection reasons, image segments before and after the action, operating curves before and after the action, event processing results, event status changes, and manual confirmation results.

[0085] When the event status of a material supply failure is pending manual confirmation, the remote operation and maintenance platform displays the visual recognition results, before and after image clips, PLC / DCS operating data trends, hierarchical control requests, PLC / DCS return results, and current event status to remote operation and maintenance personnel or on-site personnel.

[0086] The remote operation and maintenance platform receives the manual confirmation results and sends them to the fault handling unit. The manual confirmation results include confirming a fault, confirming a non-fault, correcting the fault type, recommending continued monitoring, recommending escalation of the solution, or recommending system shutdown.

[0087] When the manual confirmation result is that it is not a fault, the fault handling unit closes the material feeding fault event and records the event as a misidentified sample or a low-confidence sample. When the manual confirmation result is that it is a confirmed fault and it is recommended to continue monitoring, the fault handling unit continues to collect visual data and PLC / DCS operating data of the corresponding key equipment locations, and re-judges within the subsequent preset observation window. When the manual confirmation result is to correct the fault type, the fault handling unit re-matches the corresponding PLC / DCS variable address, fault level classification rules and hierarchical control request generation rules according to the corrected fault type. When the manual confirmation result is to recommend escalation or shutdown control, the fault handling unit regenerates the corresponding level hierarchical control request and sends it to the PLC / DCS.

[0088] The remote operation and maintenance platform generates feedback samples based on event records. Feedback samples include samples where the visual recognition results are inconsistent with the manual confirmation results, samples that have not been recovered after fault handling actions, samples where the same key equipment location is repeatedly generated within a preset time range and has been manually confirmed not to be a fault, samples with a confidence level lower than the preset confidence threshold, samples with abnormal image quality that lead to misidentification or require manual confirmation, and paired samples before and after fault handling actions.

[0089] The remote operation and maintenance platform updates the material feeding fault identification model or fault handling strategy based on the feedback samples. The fault handling strategy includes material feeding fault event triggering conditions, fault level classification rules, hierarchical control request generation rules, action limit, action cooldown time, manual confirmation conditions, and shutdown triggering conditions.

[0090] Before updating the fault handling strategy, the remote operation and maintenance platform uses historical event records for offline playback verification. Offline playback verification is used to compare the differences between the old and new strategies in terms of false alarm rate, false alarm rate, number of actions, number of downtimes, number of manual confirmations, and number of PLC / DCS refusal to execute. After the new strategy passes the verification, it is sent to the edge computing device, which saves the strategy version number, effective time, and rollback version.

[0091] The material feeding fault identification model is trained or fine-tuned through real fault samples, normal operating condition samples, and manually confirmed samples. Real fault samples include on-site historical fault videos, fault reproduction test videos, and fault segments in the event records of the remote operation and maintenance platform. Normal operating condition samples include normal operation videos under different material batches, different material levels, different feeding speeds, different lighting conditions, different dust concentrations, different equipment loads, and different camera perspectives. Manually confirmed samples include confirmations of faults, confirmations of non-faults, fault type corrections, location corrections, and processing result corrections made by remote operation and maintenance personnel or on-site personnel to the event records.

[0092] For fault types where there are insufficient real fault samples, a visual fault sample library can be constructed by combining simulation samples or enhanced samples. Based on the visual fault sample library, the pre-trained visual model can be trained, fine-tuned, or its parameters adjusted to obtain a feeding fault identification model for identifying feeding faults. The feeding fault identification model can then be deployed to an edge computing device.

[0093] Through the above embodiments, key equipment locations in the feeding system can be associated with cameras, regions of interest, PLC / DCS variable addresses, and fault types. This ensures that visual recognition results are no longer isolated alarms, but can be correlated with specific equipment locations and corresponding operating condition data. Therefore, the accuracy of fault location can be improved, and misjudgments caused by erroneous correlations between different equipment locations can be reduced.

[0094] By deploying a material supply fault identification model on an edge computing device, anomalies such as hopper bridging, rodent hole flow, material flow interruption, pipeline blockage, valve jamming, belt misalignment, material spillage, and personnel entering hazardous areas can be identified on-site. This approach reduces reliance on long-distance transmission of raw video, improves fault identification response speed, and reduces the impact of remote network fluctuations on on-site fault handling.

[0095] By correlating visual anomalies with PLC / DCS operational anomalies within a preset time window, fault events can be avoided by relying solely on a single visual recognition result or a single PLC / DCS variable alarm. For example, the absence of flow at the silo outlet requires consideration of material level, loss-in-weight scale flow rate, and the operating status of the feeding equipment; material accumulation in the conveying pipeline requires consideration of conveying differential pressure, pressure, and airflow. This reduces false alarms caused by dust obstruction, changes in lighting, short-term material fluctuations, or momentary sensor malfunctions.

[0096] By explicitly stating in step S3 that the operating condition anomaly is directly provided by the PLC / DCS, or that the fault handling unit determines the operating condition anomaly based on the PLC / DCS operating condition data, the source and formation method of the operating condition anomaly become clearer. This ensures that the mutual verification between visual anomaly features and operating condition anomalies has a clear data basis and judgment logic.

[0097] The fault handling unit generates fine-tuning, switching, de-loading, and shutdown control requests based on feeding fault events. This allows fault handling actions to be matched with the severity of the fault. Minor faults are handled with priority by less impactful methods such as air hammer, vibration, purging, low-speed pulse, or feed rate adjustment. Switching, de-loading, or shutdown controls are then executed when the fault worsens. This reduces unnecessary downtime, minimizes equipment impact, and improves the continuous operation capability of the feeding system.

[0098] By having the PLC / DCS determine whether the set of interlocking conditions corresponding to the hierarchical control request meets the execution conditions, edge computing devices or remote operation and maintenance platforms can be prevented from bypassing the on-site safety control logic to directly drive the actuator. Thus, inappropriate actions can be prevented and the operational safety of the material supply system can be improved when the system is under maintenance, in an emergency stop, when the safety door is open, when personnel enter, when downstream material receiving is not allowed, or when the PLC / DCS prohibition signal is valid.

[0099] By judging the event handling result based on the changes in visual data and PLC / DCS operating data before and after the fault handling action, it is possible to avoid directly equating the issuance of control requests or the execution of actions with the fault being recovered. Only when the fault area changes, the material flow status, the actuator feedback, and the operating data meet the recovery conditions will the event status be updated to closed. Thus, a complete processing process from identification and handling to effect verification can be formed.

[0100] By setting event states such as closed, continuing monitoring, pending escalation, pending manual confirmation, and pending shutdown control, events can be classified and processed according to the degree of fault recovery, safety status, image quality, and PLC / DCS return results. This can avoid prematurely closing events when the fault has not been recovered, and also avoid repeatedly performing the same action for the same fault.

[0101] Event logs are generated through the remote operation and maintenance platform, and feedback samples are generated based on the event logs. Samples that are inconsistent between visual recognition results and manual confirmation results, samples that have not been recovered after an action, samples of repeated false alarms, low confidence samples, samples with abnormal image quality, and paired samples before and after an action can be used for subsequent model or strategy updates. As a result, the material supply fault identification model and fault handling strategy can be continuously optimized according to the on-site operation conditions, improving the system's adaptability to changes in material batches, equipment wear, changes in ambient light, and camera contamination.

[0102] In summary, this embodiment can realize visual recognition of material supply failures, mutual verification of operating conditions, hierarchical handling, PLC / DCS interlocking decision-making, post-action effect verification, and remote feedback optimization, which can improve the reliability of material supply failure event generation, the safety of failure handling, and the traceability of remote operation and maintenance.

[0103] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for remote operation and maintenance and fault handling of a material feeding system, characterized in that, Includes the following steps: S1: Collect visual data of key equipment locations in the feeding system and obtain PLC / DCS operating data corresponding to the key equipment locations; S2: The visual data is identified using a material supply fault identification model deployed on an edge computing device to obtain visual identification results; S3: The fault handling unit generates a material supply fault event based on the correspondence between the visual anomaly features represented by the visual recognition results and the working condition anomalies represented by the PLC / DCS working condition data within a preset time window. S4: The fault handling unit generates a graded control request based on the material supply fault event and sends the graded control request to the PLC / DCS; S5: The PLC / DCS determines whether the set of interlocking conditions corresponding to the hierarchical control request meets the execution conditions, and executes the hierarchical control request or refuses to execute the hierarchical control request according to the judgment result, and returns the PLC / DCS return result to the fault handling unit. S6: The fault handling unit updates the event status of the material feeding fault event based on the result returned by the PLC / DCS; When the PLC / DCS returns a result indicating that a fault handling action has been performed, the fault handling unit determines the event handling result based on the changes in visual data and PLC / DCS operating data before and after the fault handling action, and updates the event status to one of the following based on the event handling result: closed, continued monitoring, pending escalation, pending manual confirmation, or pending shutdown control. S7: Upload the processing data of the material supply failure event to the remote operation and maintenance platform, and the remote operation and maintenance platform will form an event record; The processing data includes at least one of the following: material supply failure events, hierarchical control requests, PLC / DCS return results, event processing results, and event status.

2. The remote operation and maintenance and fault handling method for the material feeding system according to claim 1, characterized in that, The key equipment locations include at least one of the following: silo location, feeding equipment location, conveying pipeline location, valve location, belt conveyor location, personnel entry monitoring location, equipment external protection location, or safety door monitoring location; The PLC / DCS operating data includes at least one of the following: material supply status data, conveying status data, actuator status data, safety interlock status data, and upstream and downstream coordination status data. The material supply status data includes material level data, weighing flow rate data, or loss-in-weight scale flow rate data. The conveying status data includes the operating status of the feeding equipment, the current of the feeding equipment, the rotation speed of the feeding equipment, the torque of the feeding equipment, the conveying pressure, the conveying pressure difference, the air volume or the belt speed; The actuator status data includes valve opening data, air hammer status, vibrator status, purge valve status, or arch-breaking device status. The safety interlock status data includes automatic operation status, maintenance status, emergency stop status, safety door status, safety light curtain status, personnel entry status, automatic control freeze status, or PLC / DCS prohibition signal; The upstream and downstream collaborative status data includes the operating status of upstream equipment, the downstream material receiving status, and the availability status of backup branch lines or backup pipelines.

3. The remote operation and maintenance and fault handling method for the material feeding system according to claim 1, characterized in that, Before collecting visual data, establish the correspondence between cameras, key equipment locations, regions of interest, PLC / DCS variable addresses, and fault types. Based on the correspondence, the edge computing device calls the corresponding region of interest for visual data of different key device locations, and calls the recognition model corresponding to the fault type and the corresponding recognition threshold or post-processing rules.

4. The remote operation and maintenance and fault handling method for the material feeding system according to claim 1, characterized in that, In step S3, generating a material supply failure event includes: Determine visual confidence level based on visual recognition results; Determine temporal consistency based on visual recognition results of multiple consecutive frames; The operating condition matching result is determined based on the matching relationship between PLC / DCS operating condition data and corresponding fault types; Based on the visual confidence score, time consistency, and working condition matching results, a comprehensive fault confidence score is generated according to a preset fusion rule. When the overall fault confidence level meets the preset triggering conditions, the material supply fault event is generated.

5. The remote operation and maintenance and fault handling method for the material feeding system according to claim 1, characterized in that, The material supply failure events include at least one of the following: hopper bridging event, rat hole flow event, material flow interruption event, conveyor pipeline blockage event, valve jamming event, belt misalignment event, material spillage event, and personnel entering a hazardous area event; The material supply failure event is determined by visual anomalies and corresponding PLC / DCS operating anomalies within the same preset time window.

6. The remote operation and maintenance and fault handling method for the material feeding system according to claim 1, characterized in that, The fault handling unit is deployed on the edge computing device or the remote operation and maintenance platform; The hierarchical control request includes at least one of the following: fine-tuning level control request, switching level control request, load reduction level control request, and shutdown level control request; The fine-tuning level control request is used to trigger air hammer, vibration, purging, low-speed pulse, or feed rate adjustment; The switching level control request is used to switch backup branches, backup pipelines, backup valves, or backup feeding equipment; The deload control request is used to reduce the feeding speed, reduce the conveying load, or suspend upstream material supply. The shutdown-level control request is used to stop the relevant feeding equipment, conveying equipment, or trigger a safety shutdown.

7. The remote operation and maintenance and fault handling method for the material feeding system according to claim 1, characterized in that, The interlock condition set is a set of interlock conditions pre-configured for the target device based on the control level of the hierarchical control request. The interlock condition set includes one or more of the following: automatic operation state, maintenance state, emergency stop state, safety door state, safety light curtain state, personnel entry state, downstream material receiving state, target actuator state, and PLC / DCS prohibition signal; The PLC / DCS return results include execution feedback or reasons for rejection; When the set of interlocking conditions meets the execution conditions, the PLC / DCS executes the corresponding fault handling action and returns the execution feedback to the fault handling unit. When the set of interlocking conditions does not meet the execution conditions, the PLC / DCS refuses to execute the hierarchical control request and returns the reason for the refusal to the fault handling unit.

8. The remote operation and maintenance and fault handling method for the material feeding system according to claim 1, characterized in that, In step S6, the event handling result is determined by comparing the changes in the fault area, the changes in the material flow status, the changes in the feedback of the actuator, and the changes in the PLC / DCS operating data before and after the fault handling action. When the event processing result meets the recovery conditions, the fault processing unit updates the event status to closed; When the event handling result does not meet the recovery conditions and does not reach the shutdown trigger condition, the fault handling unit updates the event status to continue monitoring or pending upgrade handling. When the event status is updated to pending escalation, the fault handling unit regenerates the corresponding level of hierarchical control request and sends it to the PLC / DCS; When the event processing result meets the shutdown triggering condition, the fault processing unit updates the event status to pending shutdown control and generates a shutdown-level control request to be sent to the PLC / DCS. When the PLC / DCS returns a rejection reason, the fault handling unit updates the event status to pending manual confirmation or continued monitoring.

9. The remote operation and maintenance and fault handling method for the material feeding system according to claim 1, characterized in that, The material supply fault identification model is obtained through the following method: Obtain real fault samples, normal operating condition samples, and manually confirmed samples of the material feeding system; Using the real fault samples, and combining them with simulation samples or enhanced samples, a visual fault sample library is constructed. Based on the aforementioned visual fault sample library, the pre-trained visual model is trained, fine-tuned, or its parameters are adjusted to obtain a material feeding fault identification model for identifying material feeding faults. The material supply fault identification model is deployed to an edge computing device.

10. The remote operation and maintenance and fault handling method for the material feeding system according to claim 1, characterized in that, The remote operation and maintenance platform generates a return sample based on the event records; The returned samples include at least one of the following: samples where the visual recognition results are inconsistent with the manual confirmation results; samples that have not been recovered after fault handling actions; samples where the same key equipment location is repeatedly generated within a preset time range and has been manually confirmed not to be faulty; samples with a confidence level lower than a preset confidence threshold; and paired samples before and after fault handling actions. When the event status of the material supply failure event is pending manual confirmation, the remote operation and maintenance platform receives the manual confirmation result and sends the manual confirmation result to the fault handling unit, so that the fault handling unit closes the material supply failure event, continues monitoring, or regenerates the hierarchical control request based on the manual confirmation result. The remote operation and maintenance platform updates the material supply fault identification model or fault handling strategy based on the returned samples, and sends the updated material supply fault identification model or fault handling strategy to the edge computing device. The fault handling strategy includes at least one of the following: material supply fault event triggering conditions, fault level classification rules, hierarchical control request generation rules, action number limit, action cooldown time, manual confirmation conditions, and shutdown triggering conditions.

Images