A fault assessment method of running a fusion device with timing inspection data

Abstract

The application belongs to the technical field of fault assessment, and discloses a fault assessment method fusing equipment operation and time sequence inspection data, which acquires the structured equipment archives, equipment operation data and time sequence inspection data of railway power equipment, establishes object association according to the equipment unique code and topological relationship, carries out time alignment according to the collection time, inspection time and inspection time interval, generates operation abnormal characteristics, time sequence inspection abnormal characteristics, fault event characteristics and history health characteristics, then fuses the characteristics according to the weight to obtain the equipment level fault risk score, and converts the score into a comprehensive health score and a state category. The method makes the scattered history, inspection results, fault positioning and on-site inspection records form the same traceable data chain, can review the missed inspection and inspection changes, reduces the influence of the missed inspection of artificial inspection on the assessment results, and provides consistent basis for tripping disposal, maintenance plan and risk warning.

Description

Technical Field

[0001] This invention relates to the field of fault assessment technology, and in particular to a fault assessment method that integrates equipment operation and time-series inspection data. Background Technology

[0002] The field of fault assessment technology involves collecting, organizing, analyzing, and outputting equipment health status or risk level based on equipment operating status, historical records, inspection records, and fault events. One traditional fault assessment method, which integrates equipment operation and time-series inspection data, relies on equipment ledgers, paper drawings, on-site inspection records, dispatch monitoring screens, and human experience in railway power operation and maintenance scenarios to record, query, and judge operational anomalies, inspection completion status, tripping handling records, and maintenance needs of high-voltage towers, cable lines, power distribution equipment, switchgear, ring main units, transformers, remote control systems, and automatic control equipment.

[0003] In the current railway power operation and maintenance, equipment data is scattered, history updates are lagging, paper drawings are out of touch with the actual situation, cable path tracing is difficult, manual inspection results rely on experience and are difficult to verify, SCADA monitoring data volume leads to missed inspections by manual inspections, intelligent monitoring subsystems are isolated and form information islands, operational anomalies, inspection quality, fault location, scoring and evaluation lack a unified time sequence correlation, fault assessment results are unclear, risk levels are difficult to trace, and maintenance decisions lack a consistent basis. Summary of the Invention

[0004] To address the technical problems in railway power operation and maintenance, such as scattered equipment data, difficulty in verifying inspection results, easy omissions during manual inspections, and unclear sources of fault assessment results due to the lack of unified correlation between operational anomalies and on-site inspection records, this invention provides a fault assessment method that integrates equipment operation and time-series inspection data. The method includes the following steps: acquiring structured equipment files, equipment operation data, and time-series inspection data for railway power equipment. The structured equipment files are based on unique equipment codes, location coordinates, topological relationships, and real-time history. The equipment operation data is based on telemetry data, remote signaling data, inspection results, trip records, and fault location data. The time-series inspection data is based on inspection trajectories, inspection work tickets, on-site photos, defect records, and scoring and evaluation records for on-site verification. According to the... The unique equipment code associates the equipment operation data and the time-series inspection data with the structured equipment file, and forms equipment association objects based on the topological relationship. The data corresponding to the equipment association objects are time-aligned according to the acquisition time, inspection time, fault trigger time, and inspection time interval to generate time alignment results. Based on the time alignment results, operational anomaly features, time-series inspection anomaly features, fault event features, and historical health features are generated. The operational anomaly features, time-series inspection anomaly features, fault event features, and historical health features are weighted and fused to generate an equipment-level fault risk score. The equipment-level fault risk score is converted into a comprehensive health score, and a status category of normal, attention, abnormal, or severe is output based on the comprehensive health score.

[0005] Furthermore, operational anomaly characteristics are formed through telemetry limit exceeding judgment, remote signaling expected state judgment, and adjacent station patrol change judgment. Normal lower and upper limits are configured for telemetry data. Normal results are output when telemetry data falls between these limits, and abnormal results are output when data falls below or exceeds the upper limit. Expected states are configured for remote signaling data. Normal results are output when data matches the expected state, and abnormal results are output when data does not match. Change results are generated when two adjacent station patrol results are inconsistent. Thus, abnormal objects from automatic, manual, and historical station patrol queries can be uniformly transformed into operational anomaly characteristics that can participate in the fusion evaluation.

[0006] As a progressive scheme, the operational anomaly feature also receives dual-power load monitoring results. A near-zero load threshold is configured for the first and second load telemetry values ​​of the dual-power load. When both load telemetry values ​​are not greater than the near-zero load threshold, a dual-stop alarm result is generated. When a dual-stop alarm result exists and the corresponding equipment is not in a maintenance setting state, the dual-stop alarm result is included in the operational anomaly feature. When a dual-stop alarm result exists and the corresponding equipment is in a maintenance setting state, alarms caused by maintenance work are masked while the maintenance mark is retained, thereby reducing the interference of the maintenance state on fault assessment.

[0007] As an optional limitation, the time-series inspection anomaly features are formed through the coverage relationship between the inspection trajectory and the equipment location. The minimum distance between the position coordinates and the location coordinates of the inspection trajectory within the inspection time interval is used as the coverage judgment distance. When the coverage judgment distance is not greater than the coverage threshold, a visited coverage result is generated; when the coverage judgment distance is greater than the coverage threshold, a missed inspection result is generated. The visited coverage result, the missed inspection result, the inspection work order binding relationship, the on-site photo evidence result, the defect record, and the deduction item are all extracted with numerical features and normalized according to preset weights to form the time-series inspection anomaly features.

[0008] As an optional limitation, fault event characteristics are formed using fault location data and tripping handling data. The fault location data extracts the fault location acquisition device code, location status, fault trigger time (with nanoseconds), original waveform recording file, channel peak value, and device status data. The traveling wave propagation speed is determined based on the line inductance and capacitance per unit length, and the fault distance is determined based on the single-end wavefront time difference or double-end wavefront arrival time difference. Fault event characteristics are generated based on the fault trigger time, original waveform recording file, channel peak value, device status data, and fault distance. Furthermore, the wavefront sample point sequence number is determined based on the waveform recording start frame time, wavefront arrival time, and sampling frequency, and the wavefront position is located in the original waveform recording file to align the fault time, waveform data, and channel peak value. When the location status is abnormal, device status data is missing, or the original waveform recording file fails to read, a record to be reviewed is generated, and the reliability of the fault event characteristics is reduced.

[0009] As a preferred approach, the health characteristics of the service history are formed from real-time service history and scoring evaluation records. Based on the real-time service history, the years of service already served, the designed service life, the weight of each evaluation item, and the item with the largest deduction are extracted. The ratio of the years of service already served to the designed service life is mapped to a standard score as the basic service score. The item with the largest deduction is then subtracted from the basic service score to obtain the component score. When all component scores are higher than a preset shortcoming threshold, the component scores are weighted and summed to obtain the equipment health score. When any component score is not higher than the preset shortcoming threshold, the lowest component score is taken as the equipment health score, and the equipment health score is converted into a service history health characteristic.

[0010] As an additional constraint, the weighted fusion process first converts historical health features into health risk items where lower scores indicate higher risk. Then, non-negative fusion weights are assigned to health risk items, operational anomaly features, time-series inspection anomaly features, fault event features, and knowledge matching features, respectively, ensuring the sum of all fusion weights is one. Finally, the features are synthesized according to their fusion weights to form an equipment-level fault risk score. Knowledge matching features are obtained by normalizing the matching results output by knowledge graphs, typical fault case libraries, or diagnostic models for equipment-related objects.

[0011] Furthermore, the status category is formed through an interval mapping of the comprehensive health score. The equipment-level fault risk score is converted into a percentage-based comprehensive health score. When the comprehensive health score is higher than the lower limit of normal, the normal category is output; when the comprehensive health score is not higher than the lower limit of normal but higher than the lower limit of attention, the attention category is output; when the comprehensive health score is not higher than the lower limit of attention but higher than the upper limit of severity, the abnormal category is output; and when the comprehensive health score is not higher than the upper limit of severity, the severity category is output. Combining the comprehensive health score, status category, and corresponding preceding abnormal characteristics, equipment risk level, maintenance suggestions, early warning information, inspection abnormality report, and missed inspection record are generated.

[0012] As an anomaly handling solution, when equipment operation data, time-series inspection data, or structured equipment files are missing, traceable alternative data is retrieved according to the equipment's unique code, topological relationship, and adjacent collection time periods. When the alternative data cannot meet the time-series consistency requirements of the same equipment-related objects, a data anomaly marker is generated and the corresponding data is prevented from directly participating in the equipment-level fault risk scoring. When the equipment operation data is restored, the time-series inspection data is retransmitted, or the structured equipment files are updated, the time alignment results are regenerated and the comprehensive health score and status category are updated.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, equipment operation data, time-series inspection data, and structured equipment files are bound together through unique equipment codes, topological relationships, and time alignment results. This allows dispersed records, SCADA patrols, fault location, trip handling, and on-site inspection records to enter the same object chain, solving the problem of dispersed assessment sources. By using telemetry limit exceedance, telemetry deviation, dual-stop alarm, adjacent patrol changes, and maintenance setting shielding rules, operational anomalies are distinguished from operational status, reducing missed detections and misjudgments during manual patrols. By binding the patrol trajectory coverage determination, missed patrol results, and patrol work tickets, the quality of on-site inspections is verifiable. By weighting and fusing the health characteristics of the records, operational anomaly characteristics, time-series inspection anomaly characteristics, and fault event characteristics, a traceable fault risk score, comprehensive health score, and status category are formed, providing a consistent basis for early warning and maintenance decisions. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the main process of equipment fault assessment in this invention, showing the main steps of integrating structured equipment files, equipment operation data and time-series inspection data into the assessment. Figure 2 This is a diagram showing the association between the unique device code and topological relationship data of the present invention, illustrating the relationship between the device operation data and the time-series inspection data collected to the associated device objects; Figure 3This is a data flow diagram for the inspection trajectory coverage determination of the present invention, showing the data connection between the inspection trajectory, positioning coordinates, coverage determination distance, visited coverage results and missed inspection results; Figure 4 This is a single-ended traveling wave fault event feature generation diagram of the present invention, illustrating the generation relationship between single-ended fault distance, wavefront position, and fault event features in Embodiment 1; Figure 5 This is a processing diagram for the generation difference of the double-ended traveling wave fault distance in the present invention, showing the processing relationship between the arrival time of the wavefront at the M end of the line, the arrival time of the wavefront at the N end of the line, the line length, and the fault distance in Embodiment 2. Figure 6 This invention integrates a scoring and status output relationship diagram, which shows the relationship between health risk items, normalized values ​​of operational anomalies, normalized values ​​of inspection anomalies, normalized values ​​of fault events, knowledge matching features, equipment-level fault risk scores, comprehensive health scores, and status categories. Detailed Implementation

[0015] The technical solutions described in this invention are illustrated below with reference to embodiments. These embodiments are based on a railway power equipment operation and maintenance scenario, employing unified data item names to describe structured equipment files, equipment operation data, time-series inspection data, time alignment results, operational anomaly characteristics, time-series inspection anomaly characteristics, fault event characteristics, historical health characteristics, equipment-level fault risk scores, comprehensive health scores, and status categories. The original disclosure did not assign numerical labels to physical components; subsequent embodiments only use step numbers and data item names, without adding component numerical labels. The step numbers in the embodiments are used to describe the execution relationship of the method flow and do not indicate parameter limitations other than the number of execution objects or the execution order, nor do they change the technical boundaries defined by the protection scope text.

[0016] Example 1 Please see Figure 1 , Figure 2 , Figure 3 , Figure 4 and Figure 6 This embodiment provides a fault assessment method that integrates equipment operation and time-series inspection data. In this embodiment, the method uses the unique equipment code as the object's main thread, topological relationships as the association constraints, and the acquisition time, inspection time, fault trigger time, and inspection time interval as the time-series benchmark. It aggregates equipment operation data and time-series inspection data onto the same associated equipment object and generates equipment-level fault risk scores, comprehensive health scores, and status categories through multi-feature fusion. In this embodiment, fault event features are determined using single-end traveling wave ranging to form fault distance.

[0017] S1 acquires structured equipment files, equipment operation data, and time-series inspection data for railway power equipment.

[0018] S111, Establish a structured equipment file. The structured equipment file uses the equipment's unique code as the primary key and records location coordinates, Geographic Information System (GIS) map location, equipment attributes, equipment photos, drawings, 3D model index, topology relationships, real-time history, annual history, autumn review report, rating and evaluation records, maintenance records, trip records, inspection work orders, inspection trajectory index, operation records, protection settings, and Supervisory Control and Data Acquisition (SCADA) acquisition index. The equipment's unique code is used to identify the same equipment object in subsequent steps; the location coordinates are used for subsequent inspection trajectory coverage determination; the topology relationships are used to determine the lines, sites, devices, circuits, and inspection object affiliation between equipment; and the real-time history is used to generate history health characteristics.

[0019] S112, Acquire equipment operation data. Equipment operation data comes from the railway power equipment operation monitoring link and includes telemetry data, remote signaling data, device online data, equipment operation mode, protection settings, switch operation records, dual power supply load status, alarm data, maintenance setting status, screen lock status, trip records, fault handling records, and fault location data. Telemetry data must include at least the telemetry object identifier, acquisition time, telemetry value, lower normal limit, and upper normal limit; remote signaling data must include at least the remote signaling object identifier, acquisition time, actual status value, and expected status; station inspection results include the station identifier, inspection time, inspection object identifier, inspection object result, marking status, and remarks; fault location data includes the fault location acquisition device code, location status, fault trigger time, original waveform file, channel list, channel peak value, and equipment status data.

[0020] S113, Acquire Time-Series Inspection Data. Time-series inspection data comes from the mobile terminal's on-site inspection records, including inspection trajectory, inspection work order, on-site photos, defect records, inspection notes, deductions, and scoring evaluation records. The inspection trajectory is saved in the order of trajectory point collection time, with each trajectory point recording its location coordinates and collection time. The inspection work order records the inspection task, inspection equipment range, inspection start time, inspection end time, and the unique code of the bound equipment. On-site photos, defect records, and deductions are associated with the inspection work order through the unique equipment code. In weak or no network conditions, the inspection trajectory is first written to the local cache, and then retransmitted according to the trajectory point collection time after communication is restored, maintaining the original time sequence after retransmission.

[0021] S2 associates equipment operation data and time-series inspection data with structured equipment files according to the equipment's unique code, and forms equipment association objects based on topological relationships.

[0022] S211, Perform unique equipment code matching. This involves matching the unique codes of telemetry object identifiers, remote signaling object identifiers, fault location acquisition equipment codes, circuit identifiers in trip records, and the inspected equipment range, on-site photos, defect records, and deduction items in time-series inspection data. When a match is successful, the corresponding data is written to the same structured equipment file item, forming the basic dataset of the associated equipment objects. When a match fails, an abnormal coding status is recorded, and the failed-match data is not directly incorporated into the equipment-level fault risk score.

[0023] S212, Perform topology relationship verification. For data that has passed the unique device code matching, further verify the topology attribution relationships between lines, stations, devices, loops, and patrol objects. When the data source object matches the topology relationship in the structured device file, retain the data as valid input for the device association object; when the data source object conflicts with the topology relationship, generate a topology conflict state and transfer the data to the pending review queue. The pending review queue is output to the time alignment step as the basis for subsequent abnormal boundary identification, but not as normal fusion input.

[0024] S213, forming a device association object. The device association object includes at least a structured device file item, a device operation data set, a time-series inspection data set, a topology adjacency object set, and a data validity status. The device operation data set receives data output from S112, the time-series inspection data set receives data output from S113, the topology adjacency object set receives the line, station, device, and loop relationships verified in S212, and the data validity status is used for subsequent time alignment, feature generation, and anomaly handling.

[0025] S3 aligns the data corresponding to the device-associated objects according to the acquisition time, inspection time, fault trigger time, and inspection time interval, and generates time alignment results.

[0026] S311, Unified Time Base. This process unifies the time base for data acquisition times, patrol times, and fault trigger times in the equipment operation data set, and for trajectory point acquisition times, patrol start times, and patrol end times in the time-series inspection data set, ensuring that data from different sources can be integrated into the same time-series coordinates. When the fault trigger time contains nanoseconds, the nanosecond-level time is retained as the time input for the fault event characteristics; ordinary patrol results and patrol records are entered into a time-sorted queue according to their original acquisition times.

[0027] S312, Establish Time Windows. For each device-related object, establish an operational patrol window based on the patrol time, a field patrol window based on the patrol start and end times, and a fault event window based on the fault trigger time. The operational patrol window receives telemetry data, teleindication data, dual-power load status, and adjacent patrol results; the field patrol window receives patrol tracks, field photos, defect records, and deductions; the fault event window receives trip records, fault location data, original waveform files, and equipment status data.

[0028] S313 generates time alignment results. Data from the same associated device falling within the same time window or adjacent traceable time ranges are grouped into a time alignment result. The time alignment result records the device's unique code, time window type, window start and end times, participating data items, data items awaiting review, and data validity status. The time alignment result serves as the common input for generating various features in S4, enabling operational anomalies, on-site inspections, fault events, and history scores to participate in the evaluation with the same device object and the same temporal relationship.

[0029] S4 generates operational anomaly features, time-series inspection anomaly features, fault event features, and history health features based on the time alignment results.

[0030] S411, Generate operational anomaly characteristics. Perform limit exceedance judgment on telemetry data in the time alignment results: output normal telemetry result when the telemetry value is between the normal lower limit and the normal upper limit; output abnormal telemetry result when the telemetry value is below the normal lower limit or above the normal upper limit. Perform expected state judgment on telemetry data: output normal telemetry result when the actual state value matches the expected state; output abnormal telemetry result when the actual state value does not match the expected state. Perform change judgment on two adjacent station inspection results: output station inspection change result when the inspection result of the same inspection object in the current inspection window is inconsistent with the inspection result in the previous inspection window. Perform dual-power supply load status dual-stop alarm judgment: output dual-stop alarm result when both the first and second load telemetry values ​​are not greater than the near-zero load threshold. If the equipment associated with the dual-stop alarm result is in maintenance setting state, write the alarm to the maintenance shielding record; if it is not in maintenance setting state, merge the alarm into the operational anomaly characteristics. The verification method for abnormal operation characteristics is as follows: the time alignment results or equipment operation data of the same device-related objects can be used to query telemetry abnormal results, remote signaling abnormal results and dual stop alarm results and their corresponding data acquisition time or alarm generation time, and the patrol change results and their corresponding patrol time can be queried in the patrol object results of adjacent patrol windows.

[0031] S412, Generate Time-Series Inspection Anomaly Features. The inspection trajectories within the on-site inspection window are sorted according to the time of trajectory point acquisition. The positioning coordinates in the structured equipment file are read, and the distance between the position coordinates of each trajectory point and the positioning coordinates within the inspection time interval is calculated. The minimum distance is taken as the coverage judgment distance. When the coverage judgment distance is not greater than the coverage threshold, a visited coverage result is generated; when the coverage judgment distance is greater than the coverage threshold, a missed inspection result is generated. Subsequently, the visited coverage result, missed inspection result, inspection work order binding relationship, on-site photo evidence, defect records, and deduction items are normalized into time-series inspection anomaly features. The verification method for time-series inspection anomaly features is: within the inspection time interval corresponding to the inspection work order, the inspection trajectory can be replayed, and the coverage judgment distance, visited coverage result, and missed inspection result for each equipment point can be queried.

[0032] S413, Generate Fault Event Features. Read the fault location acquisition equipment code, location status, fault trigger time, original waveform file, channel peak value, and equipment status data from the fault location data. Determine the traveling wave propagation speed based on the inductance and capacitance per unit length of the line; this speed serves as the fundamental parameter for fault distance generation. When using single-end traveling wave ranging, read the time difference between the first received traveling wave and the reflected traveling wave at the measurement point, and convert the round-trip propagation distance formed by this time difference and the traveling wave propagation speed into a one-way fault distance. Determine the wavehead sample point sequence number based on the waveform start frame time, wavehead arrival time, and sampling frequency. Then, locate the wavehead position in the original waveform file according to the wavehead sample point sequence number, aligning the fault trigger time, waveform data, and channel peak value. The fault event features are obtained by normalizing the fault trigger time, wavehead position, channel peak value, equipment status data, fault distance, and trip handling records. If the location status is abnormal, equipment status data is missing, or the original waveform file reading fails, a record to be reviewed is generated, and the reliability of the fault event features is reduced. The verification method for fault event characteristics is: being able to query the fault trigger time, original waveform file, waveform head position, fault distance, and pending review status under the same device associated object.

[0033] S414, Generate Service History Health Features. Read the service life, design service life, evaluation item weights, and maximum deduction item from the real-time service history. Map the ratio of service life to design service life to a standard score as the basic service score. Then, subtract the maximum deduction item from the basic service score to obtain the component score. When all component scores are higher than a preset shortcoming threshold, the component scores are weighted and summarized according to the evaluation item weights to obtain the equipment health score; when any component score is not higher than the preset shortcoming threshold, the lowest component score is taken as the equipment health score. The equipment health score is then converted into service history health features. The verification method for service history health features is: the ability to trace the service life, design service life, evaluation item weights, maximum deduction item, component score, and equipment health score from the real-time service history.

[0034] S5 generates an equipment-level fault risk score by weighting and fusing operational anomaly characteristics, time-series inspection anomaly characteristics, fault event characteristics, and historical health characteristics.

[0035] S511, Execution Feature Normalization. Operational anomaly characteristics are converted into normalized values ​​based on the number of anomalies, duration of anomalies, and maintenance shielding records of telemetry anomaly results, remote signaling anomaly results, station patrol change results, and dual-stop alarm results. Time-series inspection anomaly characteristics are converted into normalized values ​​based on missed inspection results, defect records, deduction items, and on-site photographic evidence. Fault event characteristics are converted into normalized values ​​based on fault distance, channel peak value, fault trigger time, equipment status data, trip handling records, and reliability. Historical health characteristics are converted into health risk items based on equipment health scores; the lower the equipment health score, the higher the health risk item.

[0036] S512, Perform knowledge matching. Read the matching results from the knowledge graph, typical fault case library, or diagnostic model, and limit the matching results to the correspondence between equipment type, topological location, fault phenomenon, patrol anomaly type, and defect record. If multiple candidate results exist, prioritize those consistent with the equipment's unique code, topological relationship, and fault trigger time; if candidate results conflict in topological relationship, output a matching conflict status and reduce the credibility of the knowledge matching feature. When the knowledge matching feature is used as fusion input, its input fields and matching conditions can be traced.

[0037] S513, Generate Equipment-Level Fault Risk Score. Assign non-negative fusion weights to health risk items, normalized values ​​of operational anomalies, normalized values ​​of inspection anomalies, normalized values ​​of fault events, and knowledge matching features, ensuring the sum of all fusion weights is one. Weight the above features according to the fusion weights to obtain the equipment-level fault risk score. If the same equipment-related object simultaneously possesses high-confidence fault event features and missed inspection results, retain the fusion contribution of each type of feature, without one type overriding the other; if there are data anomaly markers, prevent abnormal data from directly entering the fusion process and retain the abnormal data in the pending review record.

[0038] S6 converts the equipment-level failure risk score into a comprehensive health score, and outputs a status category of normal, attention, abnormal or severe based on the comprehensive health score.

[0039] S611, Generate a comprehensive health score. Convert the equipment-level failure risk score into a percentage-based comprehensive health score; the higher the risk, the lower the comprehensive health score. The comprehensive health score is written into the structured equipment file and saved along with the equipment's unique code, time window type, participation characteristics, fusion weight, and data validity status, forming a traceable assessment record.

[0040] S612, Output Status Category. A "Normal" category is output when the overall health score is above the lower limit of normal; a "Caution" category is output when the overall health score is not above the lower limit of normal but above the lower limit of caution; an "Abnormal" category is output when the overall health score is not above the lower limit of caution but above the upper limit of severity; and a "Severe" category is output when the overall health score is not above the upper limit of severity. Status categories are used to generate equipment risk levels, maintenance recommendations, early warning information, inspection anomaly reports, and missed inspection records. The verification method for status categories is: the ability to trace the overall health score, status category, equipment-level fault risk score, participating characteristics, and corresponding data sources from the assessment records.

[0041] S613, execute an anomaly rollback. If equipment operation data, time-series inspection data, or structured equipment files are missing, traceable alternative data is retrieved according to the equipment's unique code, topology relationship, and adjacent acquisition time periods. If the alternative data cannot meet the time-series consistency requirements of the same equipment's associated objects, a data anomaly flag is generated, and the corresponding data is prevented from directly participating in the equipment-level fault risk scoring. If equipment operation data is recovered, time-series inspection data is retransmitted, or structured equipment files are updated, the time alignment results are regenerated, and the comprehensive health score and status category are updated.

[0042] Example 2 Please see Figure 1 , Figure 2 , Figure 3 , Figure 5 and Figure 6 This embodiment provides a fault assessment method that integrates equipment operation and time-series inspection data. The difference between this embodiment and Embodiment 1 is that the generation basis for fault distance in the fault event characteristics is changed from the round-trip time difference of a single-end wavefront to the arrival time difference of a double-end wavefront; the difference is located at S413, and the remaining structured equipment file acquisition, equipment association, time alignment, operation anomaly characteristics, time-series inspection anomaly characteristics, history health characteristics, fusion scoring, and status output all follow the common implementation link of Embodiment 1.

[0043] S1. Acquire structured equipment files, equipment operation data, and time-series inspection data for railway power equipment. The structured equipment files primarily consist of unique equipment codes, location coordinates, topological relationships, and real-time history. Equipment operation data still includes telemetry data, remote signaling data, inspection results, trip records, and fault location data. Time-series inspection data still includes inspection trajectories, inspection work orders, on-site photos, defect records, and scoring and evaluation records. Unlike Embodiment 1, this embodiment, when acquiring fault location data, reads the fault trigger time, location status, and equipment status data sent by the distance measuring devices at line M and line N, respectively, and uses the line length from line M to line N as the input data item for generating the fault distance.

[0044] S2, based on the unique equipment code, associates equipment operation data and time-series inspection data with the structured equipment file, and forms equipment association objects according to the topology relationship. The unique equipment codes corresponding to the M-end ranging device and the N-end ranging device of the line are entered into the same line topology relationship verification. If both endpoints can be located in the same line object in the topology relationship, the fault location data of the two endpoints are merged into the same equipment association object; if only one endpoint can pass the topology relationship verification, an endpoint missing status is generated, and the fault distance between the two ends is prevented from directly participating in the equipment-level fault risk scoring.

[0045] S3: Time alignment is performed on the data corresponding to the equipment-related objects according to the acquisition time, patrol time, fault trigger time, and patrol time interval, generating a time alignment result. When dual-end fault location data enters the fault event window, the arrival time of the wavefront at line M and line N are retained respectively, and the equipment status data at both endpoints are checked against the same window. If the fault trigger times at the two endpoints cannot fall into the same fault event window, a timestamp misalignment state is generated; if the location status of both endpoints is valid, a time alignment result that can participate in the dual-end fault distance calculation is generated.

[0046] S4 generates operational anomaly features, time-series inspection anomaly features, fault event features, and historical health features based on the time alignment results. Operational anomaly features are generated according to telemetry limit violations, telemetry deviations, dual-stop alarms, maintenance setting shielding, and changes in adjacent inspection stations; time-series inspection anomaly features are generated according to inspection trajectory coverage determination, missed inspection results, inspection work order binding relationships, on-site photo evidence results, defect records, and deduction items; historical health features are generated according to years of service, design service years, evaluation item weights, maximum deduction item, component scores, and equipment health scores.

[0047] S413, Generate Fault Event Features. Read the arrival times of the wavefront at line M and line N, the line length from line M to line N, and the traveling wave propagation speed. First, compare the arrival times of the wavefronts at line M and line N, then combine this with the line length to determine the location of the fault point relative to both ends of the line. If the arrival time of the wavefront at line M is later than that at line N, the fault point is relatively closer to line N; if the arrival time of the wavefront at line M is earlier than that at line N, the fault point is relatively closer to line M. Then, generate the fault distance based on the arrival time difference between the two ends, the line length, and the traveling wave propagation speed. Normalize the fault distance, the location status of the two endpoints, equipment status data, channel peak values, and tripping handling records into fault event features. If the location status of either endpoint is abnormal, equipment status data is missing, or the fault trigger time is misaligned, output a double-end verification record and reduce the reliability of the fault event features. The verification method for this difference technology section is as follows: under the same line topology, it is possible to query the arrival time of the wavefront at the M end of the line, the arrival time of the wavefront at the N end of the line, the line length, the fault distance, and the records to be verified at both ends.

[0048] S5, based on operational anomaly characteristics, time-series inspection anomaly characteristics, fault event characteristics, and historical health characteristics, performs weighted fusion to generate an equipment-level fault risk score. Before participating in the fusion of dual-end fault event characteristics, it is first verified whether the fault distance is compatible with the line topology; if the fault distance exceeds the locatable range corresponding to the line length, a fault distance out-of-bounds state is generated, and the fault distance is prevented from directly entering the fusion. After verification, the fault event characteristics, operational anomaly characteristics, time-series inspection anomaly characteristics, historical health characteristics, and knowledge matching characteristics are combined according to the fusion weights to form the equipment-level fault risk score.

[0049] S6 converts the equipment-level fault risk score into a comprehensive health score, and outputs a status category of normal, attention, abnormal, or severe based on the comprehensive health score. The output includes equipment risk level, maintenance recommendations, early warning information, patrol anomaly reports, missed patrol records, and fault location reports. Unlike Example 1, the fault location report output in this example simultaneously includes the arrival time of the wavefront at line M end, the arrival time of the wavefront at line N end, line length, fault distance, and the status of both ends pending verification, in order to verify the two-end ranging link. The verification method for the comprehensive health score and status category is: the contribution of the two-end fault distance to the equipment-level fault risk score and the corresponding original fault location data can be traced from the evaluation records.

[0050] The above embodiments integrate equipment operation data, time-series inspection data, fault location data, and history scoring data into the same evaluation link through unified unique equipment coding, topology relationships, and time alignment results. Embodiment 1 uses the round-trip time difference of a single-end waveform head to form fault event characteristics, while Embodiment 2 uses the arrival time difference of a double-end waveform head. Both maintain the same technical axis at the levels of structured equipment files, equipment-related objects, feature fusion, and status output. The specific processes, field states, processing order, judgment conditions, parameter sources, and data collaboration relationships described above are used to explain the possible implementations of this invention and should not limit this invention to the specific embodiments listed. Without departing from the technical solutions described in this invention and the scope of the original disclosure, any equivalent substitutions, equivalent modifications, equivalent combinations, order adjustments, equivalent transformations of field names, equivalent acceptance of the executing entity, or equivalent changes in the carrier form that can be conceived by those skilled in the art should fall within the scope of protection of this patent; however, they should not be extended to unclaimed topics, nor should the substantive correspondence of technical objects be changed through name changes.

Claims

1. A fault assessment method that integrates equipment operation and time-series inspection data, characterized in that, include: The system acquires structured equipment files, equipment operation data, and time-series inspection data for railway power equipment. The structured equipment files include unique equipment codes, location coordinates, topological relationships, and real-time history. The equipment operation data includes telemetry data, remote signaling data, inspection results, trip records, and fault location data. The time-series inspection data includes inspection trajectories, inspection work tickets, on-site photos, defect records, and scoring and evaluation records. The equipment operation data and the time-series inspection data are associated with the structured equipment file according to the unique equipment code, and equipment association objects are formed according to the topology relationship; The data corresponding to the device-associated objects are time-aligned according to the collection time, inspection time, fault trigger time, and inspection time interval to generate a time alignment result. Based on the time alignment results, operational anomaly features, time-series inspection anomaly features, fault event features, and history health features are generated respectively. Based on the weighted fusion of the operational anomaly characteristics, the time-series inspection anomaly characteristics, the fault event characteristics, and the historical health characteristics, an equipment-level fault risk score is generated. The equipment-level fault risk score is converted into a comprehensive health score, and a status category of normal, attention, abnormal or severe is output based on the comprehensive health score.

2. The fault assessment method for fusion equipment operation and time-series inspection data according to claim 1, characterized in that, The generation of the aforementioned runtime anomaly characteristics includes: Configure a normal lower limit and a normal upper limit for the telemetry data. When the telemetry data is between the normal lower limit and the normal upper limit, a normal telemetry result is generated. When the telemetry data is below the normal lower limit or above the normal upper limit, an abnormal telemetry result is generated. Configure a desired state for the remote signaling data. When the remote signaling data matches the desired state, generate a normal remote signaling result. When the remote signaling data does not match the desired state, generate an abnormal remote signaling result. The changes in the telemetry anomaly results, the remote signaling anomaly results, and the results of two adjacent station inspections are normalized to obtain the operational anomaly characteristics.

3. The fault assessment method for fusion equipment operation and time-series inspection data according to claim 2, characterized in that, Generating the aforementioned operational anomaly characteristics also includes: A near-zero load threshold is configured for the first and second load telemetry values ​​of the dual-power load. A dual-stop alarm result is generated when both the first and second load telemetry values ​​are not greater than the near-zero load threshold. When the dual-stop alarm result exists and the corresponding equipment is not in the maintenance setting state, the dual-stop alarm result will be included in the abnormal operation characteristics. When the dual-stop alarm result exists and the corresponding equipment is in the maintenance setting state, the alarm caused by the maintenance operation is masked and the maintenance mark is retained.

4. The fault assessment method for fusion equipment operation and time-series inspection data according to claim 1, characterized in that, The generation of the time-series inspection anomaly features includes: using the minimum distance between the position coordinates and the location coordinates of the inspection trajectory within the inspection time interval as the coverage determination distance; generating a visit coverage result when the coverage determination distance is not greater than the coverage threshold; generating a missed inspection result when the coverage determination distance is greater than the coverage threshold; and extracting numerical features from the visit coverage result, the missed inspection result, the inspection work ticket binding relationship, the on-site photo evidence result, the defect record, and the deduction item, and then normalizing them according to preset weights to form the time-series inspection anomaly features.

5. The fault assessment method for fusion equipment operation and time-series inspection data according to claim 1, characterized in that, Generating the fault event features includes: extracting the fault location acquisition device code, location status, fault trigger time with nanoseconds, original waveform recording file, channel peak value, and device status data from the fault location data; determining the traveling wave propagation speed based on the inductance per unit length of the line and the capacitance per unit length of the line; determining the fault distance based on the single-end wavefront time difference or the double-end wavefront arrival time difference; and generating the fault event features based on the fault trigger time, the original waveform recording file, the channel peak value, the device status data, and the fault distance.

6. The fault assessment method for fusion equipment operation and time-series inspection data according to claim 5, characterized in that, Generating the fault event features based on the fault location data further includes: The wavefront sample point number is determined based on the start frame time of the waveform recording, the wavefront arrival time, and the sampling frequency. The wavefront position is located in the original waveform recording file according to the wavefront sample point number, and the wavefront position is aligned with the channel peak value and the fault trigger time. When the positioning status is abnormal, the device status data is missing, or the original waveform file fails to be read, the corresponding fault positioning record is marked as a record to be reviewed, and the credibility of the fault event characteristics is reduced.

7. The fault assessment method for fusion equipment operation and time-series inspection data according to claim 1, characterized in that, The generation of the resume health characteristics includes: Based on the real-time history, extract the years of service already served, the designed service years, the weight of the evaluation items, and the maximum deduction item; The ratio of the service life to the designed service life is mapped to a standard score as the basic service score; The component score is obtained by subtracting the maximum deduction item from the basic service score; When all component scores are higher than the preset shortcoming threshold, the component scores are weighted and summed to obtain the equipment health score. When any component score is not higher than the preset shortcoming threshold, the lowest component score is taken as the equipment health score. The equipment health score is converted into the resume health characteristics.

8. The fault assessment method for fusion equipment operation and time-series inspection data according to claim 1, characterized in that, Weight fusion includes: The health characteristics of the resume are converted into health risk items where the lower the score, the higher the risk. Configure non-negative fusion weights for the health risk item, the operational anomaly feature, the time-series inspection anomaly feature, the fault event feature, and the knowledge matching feature, respectively, and make the sum of all fusion weights equal to one; The health risk items, the operational anomaly features, the time-series inspection anomaly features, and the fault event features are combined according to the fusion weights to form the equipment-level fault risk score; The knowledge matching features are obtained by normalizing the matching results output by the device-related objects from the knowledge graph, typical fault case library, or diagnostic model.

9. The fault assessment method for fusion equipment operation and time-series inspection data according to claim 1, characterized in that, Output status categories include Normal, Caution, Abnormal, or Critical: The equipment-level failure risk score is converted into a comprehensive health score on a percentage basis. When the overall health score is higher than the lower limit of normal, the normal category is output; when the overall health score is not higher than the lower limit of normal but higher than the lower limit of attention, the attention category is output; when the overall health score is not higher than the lower limit of attention but higher than the upper limit of severity, the abnormal category is output; when the overall health score is not higher than the upper limit of severity, the severity category is output. Based on the comprehensive health score, the status category, and the corresponding preceding abnormal characteristics, the equipment risk level, maintenance suggestions, early warning information, inspection anomaly report, and missed inspection record are generated.

10. The fault assessment method for fusion equipment operation and time-series inspection data according to claim 1, characterized in that, The generation of the time alignment result also includes: When equipment operation data, time-series inspection data, or structured equipment files are missing, traceable alternative data is retrieved according to the equipment's unique code, topological relationship, and adjacent collection time periods. When the alternative data fails to meet the timing consistency requirements of the same device-related objects, a data anomaly marker is generated and the corresponding data is prevented from directly participating in the device-level fault risk scoring. Once the device operation data is restored, the time-series inspection data is retransmitted, or the structured device file is updated, the time alignment result is regenerated and the comprehensive health score and status category are updated.

Images