A method and system for detecting defects of a support and hanger based on infrared detection

Abstract

The application discloses a kind of based on infrared detection's support hanger defect detection method and system, it is related to defect detection technical field, including, acquisition support hanger number and infrared thermal image sequence, infrared thermal image sequence is converted into temperature matrix sequence, and structure area division is carried out to support hanger, calculate structure area temperature mean and temperature gradient, identify structure area connection relationship, form heat characteristic table;Structure area in heat characteristic table is arranged as heat conduction node, structure area connection relationship is arranged as heat conduction side, and according to the structure area connection relationship corresponding to heat conduction side, the temperature mean and temperature gradient of two end structure area are matched, and topological table is formed;The application is characterized by topological table and heat contact anomaly scoring unified table structure area connection relationship and heat anomaly state, improve the accuracy of defect positioning, type annotation and grade division, and enhance the directivity of overhaul positioning.

Description

Technical Field

[0001] This invention relates to the field of defect detection technology, and in particular to a method and system for detecting defects in supports and hangers based on infrared detection. Background Technology

[0002] With the continuous construction and long-term operation of urban integrated utility tunnels, petrochemical pipelines, power transmission pipelines, heating networks, and industrial plant pipelines, pipe supports and hangers, as important components for pipeline load-bearing, limiting, vibration reduction, and attitude maintenance, are widely deployed in pipeline bends, branches, span supports, and equipment connection areas. During operation, pipe supports and hangers are affected by factors such as pipeline loads, changes in medium temperature, environmental vibration, loose connections, and structural aging, causing changes in the heat conduction and structural contact states at the connection points. Infrared detection technology can acquire the surface temperature field of pipe supports and hangers under non-contact conditions, recording changes in heat distribution on the structural surface through continuous infrared thermal images. This data can be further converted into temperature matrices, temperature time series, and thermal response characteristics. Therefore, it is increasingly being applied to scenarios such as defect detection of pipe supports and hangers, pipeline operation and maintenance inspections, connection status assessment, and identification of hidden anomalies, providing temperature field-level detection basis for analyzing the operational status of pipe supports and hangers.

[0003] However, existing infrared detection-based defect detection technologies for supports and hangers still have room for improvement: First, existing methods mostly focus on judging local temperature anomalies, local temperature difference thresholds, or manually selected areas in single-frame infrared thermal images. The detection object usually remains at the level of image regions or local hot spots, which is insufficient in expressing the division of support and hanger structural areas and the connection relationships between structural areas. Infrared thermal anomalies are difficult to accurately trace back to specific connection parts and structural contact paths, affecting the location expression of defect detection. Second, existing methods do not make sufficient use of edge temperature differences, temperature gradient changes, and anomaly persistence in continuous infrared thermal image sequences. It is difficult to reliably distinguish between short-term environmental disturbances, background heat source interference, and thermal contact anomalies caused by actual connection defects, resulting in unclear labeling of connection defect types and a lack of stable multi-dimensional scoring basis for defect level classification. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for detecting defects in supports and hangers based on infrared detection to solve the problem of difficulty in topologically locating thermal anomalies in support and hanger connection defects.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for detecting defects in supports and hangers based on infrared detection, comprising, Collect the support and hanger numbers and infrared thermal image sequences, convert the infrared thermal image sequences into temperature matrix sequences, divide the supports and hangers into structural regions, calculate the average temperature and temperature gradient of the structural regions, identify the connection relationship of the structural regions, and form a thermal feature table. The structural regions in the thermal feature table are arranged as heat conduction nodes, the connection relationships of the structural regions are arranged as heat conduction edges, and the average temperature and temperature gradient of the structural regions at both ends are matched according to the connection relationship of the structural regions corresponding to the heat conduction edges to form a topology table. Calculate the temperature difference score, gradient score, and persistence score of each heat conduction edge in the topology table, and weight them to obtain the heat contact anomaly score. Select heat conduction edges whose heat contact anomaly scores exceed the preset score threshold to form a candidate edge table. Read the heat conduction edges from the candidate edge table, match the connection relationship between the structural regions at both ends of the heat conduction edges in the thermal feature table, label the connection defect types and classify the defect levels to form a defect type table; The candidate heat conduction edges in the defect type table are compared with the heat conduction edges in the topology table to determine the structural areas at both ends. The support and hanger numbers are matched in the thermal feature table, and the thermal contact anomaly score, connection defect type and defect level are combined to form the defect detection result.

[0007] The infrared detection-based method for detecting defects in supports and hangers, as described in this invention, is characterized in that: the structural division of the supports and hangers specifically includes, Obtain the support and hanger number, and use the support and hanger number as the detection object identifier for the infrared thermal imaging sequence; Continuous infrared imaging is performed on the support brackets corresponding to the detected object, and multiple infrared thermal image frames are acquired and arranged in the order of acquisition time to form an infrared thermal image sequence. The infrared thermal image frames in the infrared thermal image sequence are numbered according to the acquisition time sequence, and the temperature values ​​in the infrared thermal image frames arranged by row and column positions are converted into a temperature matrix to form a temperature matrix sequence. Boundary extraction is performed on the continuous temperature region in the temperature matrix sequence to obtain the support and hanger outline boundary. The structural intersection position within the support and hanger outline boundary is marked as the connection part. The matrix position is divided according to the support and hanger outline boundary and the connection part to form the structural area division result.

[0008] The infrared detection-based method for detecting defects in supports and hangers, as described in this invention, is characterized in that: the formation of the thermal feature table specifically includes, The temperature values ​​of matrix positions within the same structural region are statistically analyzed in the structural region division results, and the average value of the temperature values ​​is calculated to form the average temperature of the structural region. Calculate the temperature difference between the horizontally adjacent matrix positions and the vertically adjacent matrix positions within the same structural region, and then average the absolute values ​​of the temperature differences to form a temperature gradient. The structural area boundaries in the structural area division results are judged to be connected, and structural areas with connected boundaries and shared connection parts are paired to form structural area connection relationships; Based on the support and hanger numbers in the identification of the test objects, the results of structural zone division, average temperature of structural zones, temperature gradient and connection relationship of structural zones are summarized to form a thermal characteristic table.

[0009] The infrared detection-based support defect detection method of the present invention is characterized in that: the formation of the topology table specifically includes, Based on the structural zone division results in the thermal characteristic table, the arrangement positions of the structural zones within the support and hanger outline boundary are numbered to form heat conduction nodes; Based on the structural region connection relationships in the thermal characteristic table, the two structural regions in each group of structural region connection relationships are paired and numbered to form a heat conduction edge; The structural regions at both ends of the heat conduction edge are paired with heat conduction nodes to form an edge-end node relationship; Extract the mean temperature and temperature gradient of the structural region corresponding to the edge node relationship in the thermal feature table, and associate the mean temperature and temperature gradient of the structural region with the structural regions at both ends of the heat conduction edge to form edge thermal features; Edge thermal features are indexed by the relationship between edge nodes, and these features are then configured onto heat conduction edges to form a topology table.

[0010] The infrared detection-based support defect detection method of the present invention is characterized in that: the calculation of temperature difference score, gradient score and persistence score of each heat conduction edge in the topology table specifically includes, Based on the edge node relationships in the topology table, the mean temperature and temperature gradient of the structural regions at both ends of each heat conduction edge are extracted to form an edge thermal characteristic table. Calculate the difference in mean temperature and the difference in temperature gradient between the two structural regions at the same heat conduction edge in the edge thermal characteristic table, and process the absolute values ​​of each to form temperature difference characteristic value and gradient characteristic value; The average temperature values ​​of the structural regions at both ends of the same heat conduction edge are collected in the order of collection time. Two edge temperature time series are constructed respectively, and mean-free normalization is performed to obtain the edge thermal response sequence. The edge thermal response sequence is synchronously segmented based on the acquisition time sequence and the preset window length to form a sliding time window; Within the sliding time window, the thermal response sequence at the edge is transformed in the frequency domain to extract the phase of the low-frequency component with the largest amplitude in the preset low-frequency band, and the phase difference between the two ends is calculated to form the window phase difference value. Calculate the change in window phase difference between adjacent sliding time windows, and mark the sliding time window whose change in window phase difference exceeds a preset phase threshold as a thermal response hysteresis window; Within the thermal response hysteresis window, infrared thermal images are selected where at least one of the temperature difference feature value and gradient feature value in the same heat conduction edge reaches a preset proportion of the maximum value of the same type of feature. The maximum number of consecutive infrared thermal images is counted to obtain the number of abnormal consecutive frames. Using the maximum value of the same type of feature and the total number of infrared thermal image frames as conversion benchmarks, the temperature difference feature value, gradient feature value and the number of consecutive abnormal frames are proportionally converted to form temperature difference score, gradient score and continuous score.

[0011] The infrared detection-based support defect detection method of the present invention is characterized in that: the formation of the candidate edge table specifically includes, The temperature difference score, gradient score and persistence score are weighted and fused to form a thermal contact anomaly score; The thermal contact anomaly score is compared with a preset score threshold, and thermal conduction edges whose thermal contact anomaly scores exceed the preset score threshold are selected to form candidate thermal conduction edges. A candidate edge table is formed by comparing the candidate heat conduction edges with additional temperature difference scores, gradient scores, continuity scores, and thermal contact anomaly scores.

[0012] The infrared detection-based method for detecting defects in supports and hangers, as described in this invention, is characterized in that: the formation of the defect type table specifically includes, Obtain candidate heat conduction edges from the candidate edge list, and locate the two end structural regions of the candidate heat conduction edges according to the edge end node relationship in the topology table to form candidate edge pairs; The candidate edge pairs are compared with the structural region connection relationships in the thermal feature table. The structural region connection relationships with the same structural regions at both ends are retained to form candidate connection relationships. Based on the candidate connection relationship, temperature difference score, gradient score, persistence score and thermal contact anomaly score are associated to form a defect judgment parameter group; Select the highest score in the defect judgment parameter group, and label the dominant connection defect according to the score type to which the highest score belongs, to obtain the connection defect type; The thermal contact anomaly scores are compared in different levels to classify the defect levels. The candidate thermal conduction edges, candidate connection relationships, connection defect types, defect levels, and thermal contact anomaly scores are merged to form a defect type table.

[0013] The infrared detection-based method for detecting defects in supports and hangers, as described in this invention, is characterized in that: the classification of defect levels specifically includes, According to the candidate connection relationship, the thermal contact anomaly score is matched from the defect judgment parameter group, and the thermal contact anomaly scores are concatenated in the order of the candidate connection relationship number to form a grade score sequence. Calculate the average and maximum scores of the rating sequence, and then calculate the mean of the average and maximum scores to obtain the high-level cutoff value. Based on the preset scoring threshold, average score, and high-level boundary value, the thermal contact anomaly scores of candidate connection relationships are compared in segments to classify them into first-level defects, second-level defects, and third-level defects, thus forming a defect level.

[0014] The infrared detection-based method for detecting defects in supports and hangers, as described in this invention, is characterized in that: the formation of the defect detection result specifically includes, Based on the candidate heat conduction edges in the defect type table, the connection defect type, defect level, and thermal contact anomaly score of the same candidate heat conduction edge are grouped into a defect edge record. The candidate heat conduction edges in the defect edge record are compared with the heat conduction edges in the topology table to match the edge-end node relationships of the candidate heat conduction edges and form defect edge-end relationships. The defect edge relationship is traced back to the thermal feature table to locate the structural regions at both ends of the candidate heat conduction edge, forming a defect structural region record. Based on the defective structural area record, match the support and hanger number in the thermal characteristic table to form a defect association record; Based on the defect association record, the defect edge record and defect structure area record are matched, and the support and hanger number, the structural areas at both ends, the candidate heat conduction edge, the heat contact anomaly score, the connection defect type and the defect level are combined to form the defect detection result.

[0015] Secondly, the present invention provides a support and hanger defect detection system based on infrared detection, comprising, The feature table building module is used to collect the support and hanger number and infrared thermal image sequence, convert the infrared thermal image sequence into a temperature matrix sequence, divide the support and hanger into structural areas, calculate the average temperature and temperature gradient of the structural areas, identify the connection relationship of the structural areas, and form a thermal feature table. The topology building module is used to arrange the structural regions in the thermal feature table into heat conduction nodes, arrange the connection relationships of the structural regions into heat conduction edges, and match the average temperature and temperature gradient of the structural regions at both ends according to the connection relationship of the structural regions corresponding to the heat conduction edges to form a topology table. The anomaly scoring module is used to calculate the temperature difference score, gradient score and persistence score of each heat conduction edge in the topology table, and to obtain the heat contact anomaly score by weighting them. Heat conduction edges with heat contact anomaly scores exceeding the preset scoring threshold are filtered to form a candidate edge table. The defect determination module is used to read heat conduction edges from the candidate edge table, match the connection relationship between the structural regions at both ends of the heat conduction edge in the thermal feature table, mark the connection defect type and classify the defect level to form a defect type table. The result generation module compares the candidate heat conduction edges in the defect type table with the heat conduction edges in the topology table to determine the structural areas at both ends, matches the support and hanger numbers in the thermal feature table, and combines the thermal contact anomaly score, connection defect type and defect level to form the defect detection result.

[0016] The beneficial effects of this invention are as follows: By dividing the structural area and constructing the heat conduction topology, the structural area of ​​the support and hanger, the connection parts, and the heat conduction path are incorporated into the same heat conduction relationship. This allows the temperature distribution, the connection relationship of the structural area, and the edge thermal characteristics to be expressed in a related manner within the same support and hanger structural space, enhancing the locatability of connection defects, the traceability of heat conduction paths, and the structural interpretability of defect detection results. By forming a unified judgment basis through edge thermal response sequences, window phase differences, thermal response hysteresis windows, and thermal contact anomaly scores, temperature difference anomalies, gradient anomalies, phase hysteresis, and anomaly persistence states are collaboratively involved in the judgment under candidate heat conduction edges. This improves the stability of support and hanger connection defect identification, the accuracy of grade classification, and the directionality of maintenance location. It also enhances the accuracy of defect detection location, the clarity of type labeling, the stability of grade classification, and the directionality of maintenance. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a method for detecting defects in supports and hangers based on infrared detection.

[0019] Figure 2 This is a system module diagram for infrared detection of support and hanger defects.

[0020] Figure 3 The flowchart for forming the candidate edge table.

[0021] Figure 4 A flowchart for creating a defect type table. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for detecting defects in supports and hangers based on infrared detection, including the following steps: S1: Collect the support and hanger number and infrared thermal image sequence, convert the infrared thermal image sequence into a temperature matrix sequence, divide the support and hanger into structural regions, calculate the average temperature and temperature gradient of the structural regions, identify the connection relationship of the structural regions, and form a thermal feature table.

[0026] S1.1: Obtain the support / hanger number and use it as the detection object identifier for the infrared thermal imaging sequence.

[0027] It should be noted that the support / hanger number is obtained from the support / hanger identification information, which includes the pipeline number, installation location number, and the support / hanger's own number. The support / hanger identification information is then checked for consistency with the installation location log and on-site positioning information. If the pipeline number, installation location number, and support / hanger's own number are all consistent, the consistency check is complete, ensuring that the support / hanger number points to a unique support / hanger to be inspected. The consistent support / hanger number is then associated with the infrared thermal imaging sequence, and the continuous infrared imaging range is limited to the installation area of ​​the support / hanger to be inspected. When adjacent pipelines, background heat sources, and adjacent supports / hangers are located outside the installation area of ​​the support / hanger to be inspected, they are designated as background areas and excluded from the infrared thermal imaging frame's scope, ensuring that the infrared thermal imaging frames obtained from continuous infrared imaging belong to the same support / hanger to be inspected. The support / hanger number serves as the identifier for the inspection object in the infrared thermal imaging sequence.

[0028] S1.2: Perform continuous infrared imaging on the support corresponding to the object being detected, acquire multiple infrared thermal image frames, and arrange them in the order of acquisition time to form an infrared thermal image sequence.

[0029] It should be noted that, based on the identification of the object being tested, the support for continuous infrared imaging is determined. The continuous infrared imaging range covers the overall outline of the support corresponding to the identification of the object being tested. The emissivity, reflection temperature, shooting distance, acquisition angle, and ambient temperature of the infrared imaging equipment are set. The emissivity is set according to the surface material and coating condition of the support, the reflection temperature is set according to the background heat source condition around the support, the shooting distance and acquisition angle are set according to the imaging position of the support's overall outline completely entering the infrared thermal image frame, and the ambient temperature is set according to the temperature acquisition results of the testing site. Inter-frame registration is performed on the continuously acquired infrared thermal image frames to ensure that the support outline position remains consistent at different acquisition times. Infrared thermal image frames are continuously acquired at fixed acquisition intervals within the same testing process, and the acquisition time is reserved for each infrared thermal image frame. The acquisition time is used as the basis for sorting, and multiple infrared thermal image frames are sorted from earliest to latest according to the acquisition time to maintain a temporal continuity between adjacent infrared thermal image frames. The multiple infrared thermal image frames that have been sorted are combined to form an infrared thermal image sequence.

[0030] To further explain, the fixed acquisition interval is set according to the infrared thermal image frame acquisition frequency, the thermal response change rate of the support and the detection time, in order to ensure that there is a consistent time interval between adjacent infrared thermal image frames; for example, the fixed acquisition interval is 0.5s, 1s or 2s.

[0031] S1.3: Number the infrared thermal image frames in the infrared thermal image sequence according to the acquisition time sequence, and convert the temperature values ​​in the infrared thermal image frames arranged by row and column positions into a temperature matrix to form a temperature matrix sequence.

[0032] It should be noted that, based on the infrared thermal image sequence arranged from earliest to latest according to the acquisition time, the infrared thermal image frames in the infrared thermal image sequence are sequentially assigned infrared thermal image frame numbers so that the infrared thermal image frame numbers are consistent with the acquisition time order; the temperature values ​​of each row and column position in a single infrared thermal image frame are read, and the temperature values ​​in a single infrared thermal image frame are arranged into a two-dimensional temperature matrix according to the original row and column positions of the infrared thermal image frame, while retaining the correspondence between the infrared thermal image frame numbers and the temperature matrices; the temperature matrices are arranged in the order of the infrared thermal image frame numbers to form a temperature matrix sequence.

[0033] S1.4: Extract the boundaries of the continuous temperature regions in the temperature matrix sequence to obtain the support and hanger outline boundaries. Mark the structural intersections within the support and hanger outline boundaries as connection points. Divide the matrix positions according to the support and hanger outline boundaries and connection points to form the structural area division results.

[0034] It should be noted that, based on the temperature matrix arranged in rows and columns in the temperature matrix sequence, continuous temperature regions with continuous temperature changes and spatial adjoining positions between adjacent matrix positions are identified, and the support and hanger outline boundaries are extracted along the outer edge of the continuous temperature regions; within the support and hanger outline boundaries, the locations where continuous temperature regions intersect and the outline direction changes are found, and the structural intersection locations are marked as connection points; the matrix position division range is defined by the support and hanger outline boundaries, and the connection points are used as the structural division boundaries, dividing the structural zones within the support and hanger outline boundaries, and assigning structural zone numbers according to the arrangement order from top to bottom and from left to right; the structural zone numbers are associated with the matrix position range of the structural zones to form the structural zone division results.

[0035] To further explain, continuous temperature change and spatial contiguity mean that adjacent matrix positions within the same infrared thermal image frame simultaneously satisfy both the continuous temperature value condition and the row and column connectivity condition. The continuous temperature value condition is determined by absolute temperature difference. The temperature values ​​of adjacent matrix positions are subtracted and the absolute value is taken to obtain the adjacent temperature difference. The maximum and minimum temperature values ​​within the frame are subtracted to obtain the intra-frame temperature range. 5% to 15% of the intra-frame temperature range is set as the continuous temperature difference threshold. The row and column connectivity condition adopts the four-neighbor connectivity rule. The four-neighbor connectivity rule means that, with the current matrix position as the center, matrix positions with a row number difference of 1 and the same column number, and matrix positions with a column number difference of 1 and the same row number are identified as spatially contiguous positions. Diagonal matrix positions where both row and column numbers change are not identified as spatially contiguous positions. Matrix positions with adjacent temperature differences not greater than the continuous temperature difference threshold and satisfying the four-neighbor connectivity rule are classified into the same continuous temperature region. Matrix positions with adjacent temperature differences greater than the continuous temperature difference threshold are considered as candidate boundary positions of the continuous temperature region.

[0036] The profile orientation refers to the directional change of continuous boundary points on the profile boundary of the support and hanger. Adjacent boundary points are selected according to the order of boundary point arrangement, and the difference between the row and column coordinates of adjacent boundary points is calculated to form a boundary direction vector. The angle between adjacent boundary direction vectors is calculated to obtain the profile turning angle. Boundary positions with profile turning angles greater than 45° are determined as positions where the profile orientation changes. The positions where the profile orientation changes are matched with the intersection of the continuous temperature region, the bolt connection positions, welded connection positions, clamp connection positions, hanger connection positions, and support connection positions in the support and hanger structural template. The positions that are successfully matched are marked as connection parts, and the positions that are not successfully matched are excluded as ordinary bending edges.

[0037] S1.5: Calculate the temperature values ​​of matrix positions within the same structural region in the statistical structural region division results, and calculate the average temperature value to form the average temperature of the structural region.

[0038] It should be noted that the temperature values ​​of the matrix positions within the same structural region are processed frame by frame in the temperature matrix sequence according to the infrared thermal image frame number; in a single infrared thermal image frame, the temperature values ​​of all matrix positions within the same structural region are summed, and the average value is calculated using the number of matrix positions within the same structural region as the average base, forming the average temperature of the structural region under a single infrared thermal image frame; the average temperature of the structural region is arranged according to the infrared thermal image frame number to form the average temperature of the structural region recorded by the infrared thermal image frame number.

[0039] S1.6: Calculate the temperature difference between the horizontally adjacent matrix positions and the vertically adjacent matrix positions within the same structural region, and then take the absolute value of the temperature difference and average it to form a temperature gradient.

[0040] It should be noted that, according to the infrared thermal image frame number, the horizontally adjacent matrix positions and the vertically adjacent matrix positions within the same structural region are selected frame by frame in the temperature matrix sequence; in a single infrared thermal image frame, the temperature values ​​of the horizontally adjacent matrix positions are subtracted, and the temperature values ​​of the vertically adjacent matrix positions are subtracted, and the absolute value of the temperature difference is taken; the absolute values ​​of all temperature differences within the same structural region are summed, and the average value is calculated using the number of temperature differences as the average base to form the temperature gradient under a single infrared thermal image frame; the temperature gradient is arranged according to the infrared thermal image frame number to form the temperature gradient recorded by the infrared thermal image frame number.

[0041] S1.7: Determine the connection between the boundaries of the structural zones in the structural zone division results, and pair up the structural zones with connected boundaries and shared connection parts to form structural zone connection relationships.

[0042] It should be explained that the matrix positions where the row and column positions are adjacent or the boundary positions coincide between the boundaries of different structural zones are searched. Structural zone boundaries that meet the adjacent or coincident conditions are used as candidate boundary connection relationships, while structural zone boundaries that do not meet the adjacent or coincident conditions are excluded from the connection judgment results. The connection parts are compared in the candidate boundary connection relationships, and the corresponding entity connection type and support and hanger structural template position are obtained. If the connection part belongs to the bolt connection position, welded connection position, clamp connection position, hanger connection position, or support connection position, and is consistent with the connection position in the support and hanger structural template, the two structural zones that are associated with the same connection part are retained as structural zones to be paired. If the connection part does not match the entity connection type or does not match the support and hanger structural template position, the corresponding structural zone is excluded from the structural zones to be paired. The structural zones to be paired are paired according to the structural zone number, and the two structural zone numbers, shared connection parts, and entity connection types are associated to form a structural zone connection relationship.

[0043] S1.8: Based on the support and hanger numbers in the identification of the test object, summarize the structural zone division results, the average temperature of the structural zone, the temperature gradient, and the connection relationship of the structural zone to form a thermal characteristic table.

[0044] It should be noted that the structural zone division results obtained under the same support / hanger number are used as the location basis. The average temperature and temperature gradient of the structural zone are assigned to the same structural zone, and the connection relationship of the structural zone is assigned to the same support / hanger number. The structural zone division results, average temperature, temperature gradient and connection relationship of the structural zone are concentrated in the same table through the support / hanger number to form a thermal characteristic table.

[0045] S2: Arrange the structural regions in the thermal characteristic table as heat conduction nodes, arrange the structural region connection relationships as heat conduction edges, and match the average temperature and temperature gradient of the structural regions at both ends according to the structural region connection relationships corresponding to the heat conduction edges to form a topology table.

[0046] S2.1: Based on the structural zone division results in the thermal characteristic table, the arrangement positions of the structural zones within the support and hanger outline boundary are numbered to form heat conduction nodes.

[0047] It should be noted that the structural zone division results and structural zone numbers are extracted from the thermal characteristic table, and the arrangement position of the structural zone within the support and hanger outline boundary is located according to the structural zone number; the structural zone number is mapped one-to-one with the heat conduction node number, so that the heat conduction node number takes over the structural zone position corresponding to the structural zone number, thus forming a heat conduction node.

[0048] S2.2: Based on the structural region connection relationship in the thermal characteristic table, pair up and number the two structural regions in each group of structural region connection relationships to form a heat conduction edge.

[0049] It should be noted that the structural zone connection relationships, common connection parts, and solid connection types are extracted from the thermal characteristic table. The common connection parts are then verified in the support and hanger structural template. If the common connection part matches the bolt connection position, welded connection position, clamp connection position, hanger connection position, or support connection position in the support and hanger structural template, it confirms that the two structural zones in the structural zone connection relationship have the conditions for thermal conduction contact. Structural zone connection relationships where the common connection part does not match the solid connection position are not included in the thermal conduction edge generation range. Thermal conduction edge numbers are generated according to the order of the structural zone connection relationships in the thermal characteristic table. Two structural zones with the conditions for thermal conduction contact are designated as the two ends of the same thermal conduction edge, and the two ends of the structural zones, the common connection part, and the solid connection type are listed under the same thermal conduction edge number to form a thermal conduction edge.

[0050] S2.3: Pair the structural regions at both ends of the heat conduction edge with the heat conduction nodes to form an edge-end node relationship.

[0051] It should be explained that the first and second end structural regions are extracted from the heat conduction edge; the structural regions identical to the first end structural region are found in the heat conduction nodes, and a pairing relationship between the first end structural region and the heat conduction node is established; the structural regions identical to the second end structural region are found in the heat conduction nodes, and a pairing relationship between the second end structural region and the heat conduction node is established; the heat conduction edge, the first end structural region, the second end structural region, the heat conduction nodes associated with the first end structural region, and the heat conduction nodes associated with the second end structural region are associated to form an edge-end node relationship.

[0052] S2.4: Extract the mean temperature and temperature gradient of the structural region corresponding to the edge node relationship in the thermal feature table, and associate the mean temperature and temperature gradient of the structural region with the structural regions at both ends of the heat conduction edge to form edge thermal features.

[0053] It should be noted that the first and second structural regions of the heat conduction edge are located according to the edge-end node relationship. The average temperature and temperature gradient of the structural region are matched for the first structural region and the average temperature and temperature gradient of the structural region are matched for the second structural region in the thermal feature table. The average temperature and temperature gradient of the structural region of the first structural region are associated with the first end of the heat conduction edge, and the average temperature and temperature gradient of the structural region of the second structural region are associated with the second end of the heat conduction edge, so that the heat conduction edge has the temperature characteristics of both structural regions at the two ends, forming edge-end thermal features.

[0054] S2.5: Index edge thermal features through edge node relationships, configure edge thermal features to heat conduction edges, and form a topology table.

[0055] It should be noted that the edge thermal features of the same heat conduction edge are indexed according to the edge-end node relationship, and the edge-end thermal features are assigned to the same heat conduction edge; the heat conduction nodes, heat conduction edges, edge-end node relationships and edge-end thermal features are arranged according to the heat conduction edge number to form a topology table.

[0056] S3: Calculate the temperature difference score, gradient score and persistence score of each heat conduction edge in the topology table, and weight them to obtain the heat contact anomaly score. Filter out heat conduction edges whose heat contact anomaly scores exceed the preset score threshold to form a candidate edge table.

[0057] S3.1: Extract the average temperature and temperature gradient of the structural regions at both ends of each heat conduction edge based on the edge node relationship in the topology table to form an edge thermal characteristic table.

[0058] It should be noted that the first and second end structural regions of all heat conduction edges are located according to the edge-end node relationships in the topology table; the average temperature and temperature gradient of the structural regions of the first and second end structural regions are extracted from the topology table; and the heat conduction edges, the first end structural region, the second end structural region, the average temperature and temperature gradient of the structural regions are arranged according to the heat conduction edge number to form an edge-end thermal characteristic table.

[0059] S3.2: Calculate the difference in mean temperature and the difference in temperature gradient between the two structural regions of the same heat conduction edge in the edge thermal characteristic table, and perform absolute value processing to form temperature difference characteristic value and gradient characteristic value.

[0060] It should be noted that in the edge thermal feature table, the heat conduction edge number and the infrared thermal image frame number are used as the search objects to locate the first and second end structural regions of the same heat conduction edge under the same infrared thermal image frame; the average temperature of the structural region of the first end structural region is subtracted from the average temperature of the structural region of the second end structural region, and the absolute value of the difference in the average temperature is taken to form the temperature difference feature value; the temperature gradient of the first end structural region is subtracted from the temperature gradient of the second end structural region, and the absolute value of the difference in the temperature gradient is taken to form the gradient feature value.

[0061] S3.3: Collect the average temperature values ​​of the structural regions at both ends of the same heat conduction edge according to the acquisition time sequence, construct two edge temperature time series respectively, and perform mean-free normalization processing to obtain the edge thermal response sequence.

[0062] It should be noted that the average temperatures of the structural regions at the first and second ends of the same heat conduction edge are collected in chronological order to form the first-end temperature time series and the second-end temperature time series. The average temperature values ​​of the first-end temperature time series and the second-end temperature time series are calculated respectively. The average temperature value of the structural region at the collection time point in the first-end temperature time series is subtracted from the average temperature value of the first end to obtain the first-end change. The average temperature value of the structural region at the collection time point in the second-end temperature time series is subtracted from the average temperature value of the second end to obtain the second-end change. The maximum absolute value of the first-end change and the maximum absolute value of the second-end change are used as the normalization benchmarks. When the normalization benchmark is greater than 0, the change at the end where the normalization benchmark is located is divided. When the normalization benchmark is 0, the normalized change at the end where the normalization benchmark is located is set to a zero sequence. The normalized change of the first end and the normalized change of the second end are arranged in chronological order to obtain the edge thermal response sequence.

[0063] S3.4: Based on the acquisition time sequence and preset window length, the edge thermal response sequence is synchronously segmented to form a sliding time window.

[0064] It should be noted that the edge thermal response sequences are arranged in the order of acquisition time, and the number of consecutive acquisition time points contained in a single sliding time window is determined by a preset window length; the first and second edge thermal response sequences are synchronously intercepted with the same window start and the same window end, so that the first and second edge thermal response values ​​within the same sliding time window are within the same acquisition time range; the window start point is continuously moved backward along the acquisition time sequence to form a sliding time window.

[0065] To further explain, the window length is set according to the acquisition interval of infrared thermal images, the rate of change of the thermal response of the support and the need for thermal response hysteresis identification, and is used to limit the number of consecutive acquisition time points within a single sliding time window; for example, the window length is 5 frames, 10 frames or 15 frames.

[0066] S3.5: Perform frequency domain transformation on the edge thermal response sequence within the sliding time window, extract the phase of the low-frequency component with the largest amplitude in the preset low-frequency band, calculate the phase difference between the two ends, and form the window phase difference value.

[0067] It should be noted that within a single sliding time window, the thermal response values ​​of the first and second ends are extracted in the order of acquisition time, forming a first-end window sequence and a second-end window sequence, respectively. Discrete Fourier transforms are performed on the first-end window sequence and the second-end window sequence to convert the thermal response values ​​in the time domain into amplitude and phase in the frequency domain, obtaining the first-end frequency domain component and the second-end frequency domain component. Within a preset low-frequency band, the first-end frequency domain component and the second-end frequency domain component are matched according to the same frequency point. The amplitudes of the first and second ends at the same frequency point are added, and the low-frequency component with the largest addition result is used as the phase extraction object. The phases of the first-end low-frequency component and the second-end low-frequency component at the same frequency point are extracted, and the phase difference between the phases of the first and second ends is calculated to form a window phase difference value.

[0068] To further explain, the low-frequency band is set according to the infrared thermal image frame acquisition frequency and the thermal response change period of the support and hanger. It is used to extract slow temperature fluctuations that reflect the hysteresis of heat conduction and to eliminate high-frequency fluctuations caused by infrared thermal image frame acquisition noise; for example, the low-frequency band is 0.01HZ to 0.2HZ.

[0069] S3.6: Calculate the change in window phase difference between adjacent sliding time windows, and mark the sliding time window whose change in window phase difference exceeds the preset phase threshold as a thermal response hysteresis window.

[0070] It should be noted that the sliding time windows are arranged in chronological order of acquisition time. Low-frequency components used for phase calculation are extracted from adjacent sliding time windows, and the low-frequency components are subjected to same-frequency point verification and amplitude validity verification. Same-frequency point verification requires that the low-frequency components in adjacent sliding time windows are at the same frequency point, and amplitude validity verification requires that the amplitudes of the first and second low-frequency components are not less than the preset amplitude validity threshold. Only when both same-frequency point verification and amplitude validity verification are satisfied are the phase continuity verification of the window phase difference between adjacent sliding time windows performed, and phase expansion processing is performed on the window phase difference of windows with angle jumps. The window phase difference of adjacent sliding time windows that have completed phase continuity verification is subtracted and the absolute value is processed to obtain the change in window phase difference. If the change in window phase difference exceeds the preset phase threshold, the next sliding time window in the adjacent sliding time window is marked as a thermal response hysteresis window. If the change in window phase difference does not exceed the preset phase threshold, the same-frequency point verification is not satisfied, and the amplitude validity verification is not satisfied, the sliding time window is not marked as a thermal response hysteresis window.

[0071] To further explain, the phase threshold is set based on the fluctuation range of the window phase difference between adjacent sliding time windows under normal connection conditions and the noise level of infrared thermal image frame acquisition, and is used to distinguish between normal phase fluctuations and phase abrupt changes caused by thermal response hysteresis; for example, the phase threshold is 15° to 30°.

[0072] The effective amplitude threshold is set based on the low-frequency component amplitude distribution and the infrared thermal image frame acquisition noise level under normal connection conditions. It is used to exclude phase jumps caused by low-amplitude noise and is set to 20% to 30% of the maximum low-frequency component amplitude within the same sliding time window.

[0073] S3.7: Within the thermal response hysteresis window, select infrared thermal images where at least one of the temperature difference feature value and gradient feature value of the same heat conduction edge reaches a preset ratio of the maximum value of the same type of feature, count the maximum number of consecutive infrared thermal images, and obtain the number of abnormal consecutive frames.

[0074] It should be noted that, within the thermal response hysteresis window, infrared thermal image frames containing the same heat conduction edge are located according to the acquisition time, and the temperature difference feature value and gradient feature value of the same heat conduction edge under the infrared thermal image frame are obtained; the temperature difference feature value is compared with a preset ratio of the maximum value of the temperature difference feature value, and the gradient feature value is compared with a preset ratio of the maximum value of the gradient feature value. Infrared thermal image frames whose temperature difference feature value or gradient feature value meets the comparison condition are marked as abnormal infrared thermal image frames. Infrared thermal image frames whose temperature difference feature value and gradient feature value do not meet the comparison condition are not included in the consecutive frame count. The number of consecutive abnormal infrared thermal image frames is counted according to the acquisition time sequence, and the maximum number of consecutive frames is taken to obtain the abnormal consecutive frame count.

[0075] To further explain, the preset ratio is set based on the maximum value of temperature difference characteristic value, the maximum value of gradient characteristic value, and the range of infrared thermal image frame acquisition fluctuation for all heat conduction edges under the same hanger number. It is used to filter infrared thermal image frames whose temperature difference characteristic value or gradient characteristic value is close to the maximum value of the same type of characteristic within the thermal response hysteresis window; for example, it is 60% to 80%.

[0076] S3.8: Using the maximum value of the same type of feature and the total number of infrared thermal image frames as conversion benchmarks, the temperature difference feature value, gradient feature value and the number of consecutive abnormal frames are proportionally converted to form temperature difference score, gradient score and continuous score.

[0077] It should be noted that the temperature difference feature values ​​and gradient feature values ​​of the same heat conduction edge are collected across all infrared thermal image frames according to the edge number. The maximum value of the temperature difference feature value for the same heat conduction edge is selected to obtain the temperature difference feature value used for scoring. Similarly, the maximum value of the gradient feature value for the same heat conduction edge is selected to obtain the gradient feature value used for scoring. The maximum value of the temperature difference feature value used for scoring across all heat conduction edges is used as the temperature difference conversion benchmark. The temperature difference feature value used for scoring is divided by the temperature difference conversion benchmark to form the temperature difference score. The maximum value of the gradient feature value used for scoring across all heat conduction edges is used as the gradient conversion benchmark. The gradient feature value used for scoring is divided by the gradient conversion benchmark to form the gradient score. Finally, the total number of infrared thermal image frames is used as the continuity conversion benchmark. The number of consecutive abnormal frames is divided by the total number of infrared thermal image frames to form the continuity score.

[0078] S3.9: The temperature difference score, gradient score and continuity score are weighted and fused to form a thermal contact anomaly score.

[0079] It should be explained that a preset temperature difference weight, a preset gradient weight, and a preset duration weight are set, and the sum of the three preset weights is 1; the temperature difference score is multiplied by the preset temperature difference weight, the gradient score is multiplied by the preset gradient weight, and the duration score is multiplied by the preset duration weight; the results of the three multiplication operations are added together to form a thermal contact anomaly score.

[0080] To further explain, the temperature difference weight, gradient weight, and persistence weight are set according to the degree to which temperature difference anomaly, gradient anomaly, and persistence anomaly directly affect the determination of connection defects. Temperature difference anomaly directly reflects the difference in thermal contact between the structural areas at both ends of the heat conduction edge, so the temperature difference weight is the highest. Gradient anomaly reflects the steepness of temperature change near the boundary of the structural area, so the gradient weight is lower than the temperature difference weight. Persistence anomaly reflects the temporal continuity of abnormal infrared thermal images, and is affected by the fixed acquisition interval and window length, so the persistence weight is the lowest. For example, the temperature difference weight is 0.40, the gradient weight is 0.35, and the persistence weight is 0.25, so the sum of the three weights is 1.

[0081] S3.10: Compare the thermal contact anomaly score with the preset score threshold, and filter out thermal conduction edges whose thermal contact anomaly scores exceed the preset score threshold to form candidate thermal conduction edges.

[0082] It should be noted that thermal contact anomaly scores are obtained item by item according to the thermal conduction edge number, and the thermal contact anomaly scores are compared with preset scoring thresholds. Thermal conduction edges whose thermal contact anomaly scores exceed the preset scoring thresholds are retained as candidate thermal conduction edges, while thermal conduction edges whose thermal contact anomaly scores do not exceed the preset scoring thresholds are not included in the candidate thermal conduction edge range, thus forming candidate thermal conduction edges.

[0083] To further explain, the scoring threshold is set based on the distribution of thermal contact anomaly scores and the allowable false alarm range under normal support and hanger connection conditions, and is used to distinguish between normal thermal conduction edges and candidate thermal conduction edges; for example, when the thermal contact anomaly score ranges from 0 to 1, the scoring threshold is 0.60 to 0.75.

[0084] S3.11: Compare the candidate heat conduction edges with additional temperature difference scores, gradient scores, continuity scores, and thermal contact anomaly scores to form a candidate edge table.

[0085] It should be noted that temperature difference scores, gradient scores, continuity scores, and thermal contact anomaly scores are matched according to the candidate heat conduction edge numbers; the matched temperature difference scores, gradient scores, continuity scores, and thermal contact anomaly scores are attached to the same candidate heat conduction edge, so that the candidate heat conduction edge and the scoring result are associated with the same number; the candidate heat conduction edge, temperature difference score, gradient score, continuity score, and thermal contact anomaly score are arranged according to the candidate heat conduction edge number to form a candidate edge table.

[0086] S4: Read the heat conduction edges from the candidate edge table, match the connection relationship between the structural regions at both ends of the heat conduction edges in the thermal feature table, label the connection defect type and classify the defect level to form a defect type table.

[0087] S4.1: Obtain the candidate heat conduction edges from the candidate edge table, and locate the two end structural regions of the candidate heat conduction edges according to the edge end node relationship in the topology table to form candidate edge pairs.

[0088] It should be explained that candidate heat conduction edges are obtained from the candidate edge table, and their numbers are extracted. Matching is performed on the edge-end node relationships in the topology table according to the candidate heat conduction edge numbers to locate the first and second end structure regions under the same heat conduction edge number. The candidate heat conduction edges, the first end structure regions, and the second end structure regions are combined as the same set of edge-end relationships to form candidate edge-end pairs.

[0089] S4.2: Compare the candidate edge pairs with the structural region connection relationships in the thermal feature table, retain the structural region connection relationships with the same structural regions at both ends, and form candidate connection relationships.

[0090] It should be noted that the connection relationship between the first and second structural regions of the candidate edge pair is compared with the structural region connection relationship in the thermal feature table. If the two structural regions in the structural region connection relationship are the same as the two structural regions of the candidate edge pair, they are retained as candidate connection relationships. If the two structural regions in the structural region connection relationship do not match the structural regions at both ends of the candidate edge pair, they are not included in the candidate connection relationship range and are formed as candidate connection relationships.

[0091] S4.3: Based on the candidate connection relationship, associate the temperature difference score, gradient score, continuity score and thermal contact anomaly score to form a defect judgment parameter group.

[0092] It should be noted that, taking the two end structural regions in the candidate connection relationship as the associated objects, the candidate heat conduction edge with the same two end structural regions is searched in the candidate edge table; the temperature difference score, gradient score, continuous score and thermal contact anomaly score of the candidate heat conduction edge are associated with the same candidate connection relationship; the temperature difference score, gradient score, continuous score and thermal contact anomaly score are arranged according to the candidate connection relationship to form a defect judgment parameter group.

[0093] S4.4: Select the maximum score in the defect judgment parameter group, and label the dominant connection defect according to the score type to which the maximum score belongs, to obtain the connection defect type.

[0094] It should be noted that in the defect judgment parameter group, the temperature difference score, gradient score, and continuity score are compared to determine the maximum score and the score type to which the maximum score belongs; the thermal contact anomaly score is used for defect level classification; when the score type to which the maximum score belongs is temperature difference score, the connection defect type is labeled as temperature difference-dominated connection defect; when the score type to which the maximum score belongs is gradient score, the connection defect type is labeled as gradient-dominated connection defect; when the score type to which the maximum score belongs is continuity score, the connection defect type is labeled as continuity-dominated connection defect, thus obtaining the connection defect type.

[0095] S4.5: Match thermal contact anomaly scores from the defect judgment parameter group according to the candidate connection relationship, and concatenate the thermal contact anomaly scores in the order of the candidate connection relationship numbers to form a grade score sequence.

[0096] It should be explained that, using candidate connection relationships as the matching object, thermal contact anomaly scores under the same candidate connection relationship are searched in the defect judgment parameter group; the thermal contact anomaly scores are arranged in the order of candidate connection relationship numbers so that the arrangement order of thermal contact anomaly scores is consistent with the order of candidate connection relationship numbers; the arranged thermal contact anomaly scores are then concatenated to form a graded score sequence.

[0097] S4.6: Calculate the average and maximum scores of the rating sequence, and calculate the mean of the average and maximum scores to obtain the high-level cutoff value.

[0098] It should be noted that the number of thermal contact anomaly scores in the statistical rating sequence is calculated. If the number of thermal contact anomaly scores is greater than 1, all thermal contact anomaly scores in the rating sequence are summed, and the sum is divided by the number of thermal contact anomaly scores to obtain the average score. The thermal contact anomaly score with the largest value in the rating sequence is selected to obtain the maximum score. The average score and the maximum score are added together, and the result is divided by 2 to obtain the high-level boundary value. If the number of thermal contact anomaly scores is equal to 1, the unique thermal contact anomaly score is used as the average score, the maximum score, and the high-level boundary value, and a single-item rating label is generated.

[0099] S4.7: Based on the preset scoring threshold, average score and high-level boundary value, the thermal contact anomaly scores of candidate connection relationships are compared in segments to classify them into first-level defects, second-level defects and third-level defects, thus forming a defect level.

[0100] It should be noted that thermal contact anomaly scores are obtained according to candidate connection relationships, and the scores are compared in segments with preset scoring thresholds, average scores, and high-level boundary values. When the thermal contact anomaly score is greater than the preset scoring threshold and less than or equal to the average score, the candidate connection relationship is classified as a Level 1 defect. When the thermal contact anomaly score is greater than the average score and less than the high-level boundary value, the candidate connection relationship is classified as a Level 2 defect. When the thermal contact anomaly score is greater than or equal to the high-level boundary value, the candidate connection relationship is classified as a Level 3 defect. When a single-level label exists, the segmented comparison between the average score and the high-level boundary value is not performed, and the candidate connection relationship corresponding to the single-level label is classified as a Level 1 defect, thus forming a defect level.

[0101] S4.8: Merge candidate thermal conduction edges, candidate connection relationships, connection defect types, defect levels, and thermal contact anomaly scores to form a defect type table.

[0102] It should be noted that, according to the candidate heat conduction edge number and candidate connection relationship, the candidate connection relationship, connection defect type, defect level and thermal contact anomaly score under the same candidate heat conduction edge are merged; the association between the candidate heat conduction edge and the candidate connection relationship is maintained, and the connection defect type, defect level and thermal contact anomaly score are arranged under the same candidate connection relationship; the candidate heat conduction edge, candidate connection relationship, connection defect type, defect level and thermal contact anomaly score are arranged according to the candidate heat conduction edge number to form a defect type table.

[0103] S5: Compare the candidate heat conduction edges in the defect type table with the heat conduction edges in the topology table to determine the structural areas at both ends, match the support and hanger numbers in the thermal feature table, combine the thermal contact anomaly score, connection defect type and defect level to form the defect detection result.

[0104] S5.1: Based on the candidate heat conduction edges in the defect type table, the connection defect type, defect level, and thermal contact anomaly score of the same candidate heat conduction edge are collected into a defect edge record.

[0105] It should be noted that, in the defect type table, candidate heat conduction edges are used as the aggregation objects. The connection defect type, defect level, and thermal contact anomaly score under the same candidate heat conduction edge are searched. The connection defect type, defect level, and thermal contact anomaly score are grouped into the same candidate heat conduction edge, and the association with the candidate heat conduction edge number is maintained. The candidate heat conduction edge, connection defect type, defect level, and thermal contact anomaly score are arranged according to the candidate heat conduction edge number to form a defect edge record.

[0106] S5.2: Compare the candidate heat conduction edges in the defect edge record with the heat conduction edges in the topology table, match the edge-end node relationships of the candidate heat conduction edges, and form defect edge-end relationships.

[0107] It should be noted that candidate heat conduction edge numbers are obtained from the defect edge record, and the same heat conduction edge number is searched in the heat conduction edge of the topology table; when the same heat conduction edge number is successfully matched, the edge end node relationship under the same heat conduction edge number is extracted; the candidate heat conduction edge, the first end structural area, the second end structural area, the first end heat conduction node and the second end heat conduction node are associated to form the defect edge end relationship.

[0108] S5.3: Backtrack the defect edge relationship to the thermal feature table, locate the structural regions at both ends of the candidate heat conduction edge, and form a defect structural region record.

[0109] It should be noted that, according to the first and second end structural regions in the defect edge relationship, the same structural regions are matched in the structural region division results of the thermal feature table; the matrix position range of the first and second end structural regions within the support outline boundary is located; the candidate heat conduction edge, the first end structural region, the second end structural region and the matrix position range are associated to form a defect structural region record.

[0110] S5.4: Match the support and hanger numbers in the thermal characteristic table based on the defect structure area record to form a defect association record.

[0111] It should be noted that, using the first end structural area, the second end structural area, and the matrix position range in the defect structural area record as matching objects, the support and hanger numbers containing the same structural area and matrix position range are searched in the thermal feature table; the matched support and hanger numbers are associated with the candidate heat conduction edge, the first end structural area, and the second end structural area to form a defect association record.

[0112] S5.5: Based on the defect association record, match the defect edge record and the defect structure area record, combine the support and hanger number, the structure areas at both ends, the candidate heat conduction edge, the heat contact anomaly score, the connection defect type and the defect level to form the defect detection result.

[0113] It should be noted that, based on the support and hanger number and candidate heat conduction edge in the defect association record, the thermal contact anomaly score, connection defect type and defect level of the same candidate heat conduction edge are matched in the defect edge record, and the first end structure area and the second end structure area of ​​the same candidate heat conduction edge are matched in the defect structure area record; the support and hanger number, the first end structure area, the second end structure area, the candidate heat conduction edge, the thermal contact anomaly score, the connection defect type and the defect level are combined to form the defect detection result.

[0114] This embodiment also provides a support and hanger defect detection system based on infrared detection, including: The feature table building module is used to collect the support and hanger number and infrared thermal image sequence, convert the infrared thermal image sequence into a temperature matrix sequence, divide the support and hanger into structural areas, calculate the average temperature and temperature gradient of the structural areas, identify the connection relationship of the structural areas, and form a thermal feature table. The topology building module is used to arrange the structural regions in the thermal feature table into heat conduction nodes, arrange the connection relationships of the structural regions into heat conduction edges, and match the average temperature and temperature gradient of the structural regions at both ends according to the connection relationship of the structural regions corresponding to the heat conduction edges to form a topology table. The anomaly scoring module is used to calculate the temperature difference score, gradient score and persistence score of each heat conduction edge in the topology table, and to obtain the heat contact anomaly score by weighting them. Heat conduction edges with heat contact anomaly scores exceeding the preset scoring threshold are filtered to form a candidate edge table. The defect determination module is used to read heat conduction edges from the candidate edge table, match the connection relationship between the structural regions at both ends of the heat conduction edge in the thermal feature table, mark the connection defect type and classify the defect level to form a defect type table. The result generation module compares the candidate heat conduction edges in the defect type table with the heat conduction edges in the topology table to determine the structural areas at both ends, matches the support and hanger numbers in the thermal feature table, and combines the thermal contact anomaly score, connection defect type and defect level to form the defect detection result.

[0115] In summary, this invention, through structural area division and heat conduction topology construction, incorporates the support and hanger structural area, connection parts, and heat conduction path into the same heat conduction relationship. This allows temperature distribution, structural area connection relationships, and edge thermal characteristics to be expressed in a correlated manner within the same support and hanger structural space, enhancing the locatability of connection defects, the traceability of heat conduction paths, and the structural interpretability of defect detection results. Furthermore, by establishing a unified judgment criterion through edge thermal response sequences, window phase differences, thermal response hysteresis windows, and thermal contact anomaly scores, temperature difference anomalies, gradient anomalies, phase hysteresis, and anomaly persistence states are collaboratively involved in the judgment under candidate heat conduction edges. This improves the stability of support and hanger connection defect identification, the accuracy of grade classification, and the directionality of maintenance location. It also enhances the comprehensive utilization capability of infrared thermal imaging sequences, improving the accuracy of defect detection and location, the clarity of type labeling, the stability of grade classification, and the directionality of maintenance.

[0116] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for detecting defects in supports and hangers based on infrared detection, characterized in that: include, Collect the support and hanger numbers and infrared thermal image sequences, convert the infrared thermal image sequences into temperature matrix sequences, divide the supports and hangers into structural regions, calculate the average temperature and temperature gradient of the structural regions, identify the connection relationship of the structural regions, and form a thermal feature table. The structural regions in the thermal feature table are arranged as heat conduction nodes, the connection relationships of the structural regions are arranged as heat conduction edges, and the average temperature and temperature gradient of the structural regions at both ends are matched according to the connection relationship of the structural regions corresponding to the heat conduction edges to form a topology table. Calculate the temperature difference score, gradient score, and persistence score of each heat conduction edge in the topology table, and weight them to obtain the heat contact anomaly score. Select heat conduction edges whose heat contact anomaly scores exceed the preset score threshold to form a candidate edge table. Read the heat conduction edges from the candidate edge table, match the connection relationship between the structural regions at both ends of the heat conduction edges in the thermal feature table, label the connection defect types and classify the defect levels to form a defect type table; The candidate heat conduction edges in the defect type table are compared with the heat conduction edges in the topology table to determine the structural areas at both ends. The support and hanger numbers are matched in the thermal feature table, and the thermal contact anomaly score, connection defect type and defect level are combined to form the defect detection result.

2. The method for detecting defects in supports and hangers based on infrared detection as described in claim 1, characterized in that: The structural zone division of the supports and hangers specifically includes, Obtain the support and hanger number, and use the support and hanger number as the detection object identifier for the infrared thermal imaging sequence; Continuous infrared imaging is performed on the support brackets corresponding to the detected object, and multiple infrared thermal image frames are acquired and arranged in the order of acquisition time to form an infrared thermal image sequence. The infrared thermal image frames in the infrared thermal image sequence are numbered according to the acquisition time sequence, and the temperature values ​​in the infrared thermal image frames arranged by row and column positions are converted into a temperature matrix to form a temperature matrix sequence. Boundary extraction is performed on the continuous temperature region in the temperature matrix sequence to obtain the support and hanger outline boundary. The structural intersection position within the support and hanger outline boundary is marked as the connection part. The matrix position is divided according to the support and hanger outline boundary and the connection part to form the structural area division result.

3. The method for detecting defects in supports and hangers based on infrared detection as described in claim 2, characterized in that: The formation of the thermal characteristic table specifically includes, The temperature values ​​of matrix positions within the same structural region are statistically analyzed in the structural region division results, and the average value of the temperature values ​​is calculated to form the average temperature of the structural region. Calculate the temperature difference between the horizontally adjacent matrix positions and the vertically adjacent matrix positions within the same structural region, and then average the absolute values ​​of the temperature differences to form a temperature gradient. The structural area boundaries in the structural area division results are judged to be connected, and structural areas with connected boundaries and shared connection parts are paired to form structural area connection relationships; Based on the support and hanger numbers in the identification of the test objects, the results of structural zone division, average temperature of structural zones, temperature gradient and connection relationship of structural zones are summarized to form a thermal characteristic table.

4. The method for detecting defects in supports and hangers based on infrared detection as described in claim 3, characterized in that: The formation of the topology table specifically includes, Based on the structural zone division results in the thermal characteristic table, the arrangement positions of the structural zones within the support and hanger outline boundary are numbered to form heat conduction nodes; Based on the structural region connection relationships in the thermal characteristic table, the two structural regions in each group of structural region connection relationships are paired and numbered to form a heat conduction edge; The structural regions at both ends of the heat conduction edge are paired with heat conduction nodes to form an edge-end node relationship; Extract the mean temperature and temperature gradient of the structural region corresponding to the edge node relationship in the thermal feature table, and associate the mean temperature and temperature gradient of the structural region with the structural regions at both ends of the heat conduction edge to form edge thermal features; Edge thermal features are indexed by the relationship between edge nodes, and these features are then configured onto heat conduction edges to form a topology table.

5. The method for detecting defects in supports and hangers based on infrared detection as described in claim 4, characterized in that: The calculation of temperature difference score, gradient score, and persistence score for each heat conduction edge in the topology table specifically includes: Based on the edge node relationships in the topology table, the mean temperature and temperature gradient of the structural regions at both ends of each heat conduction edge are extracted to form an edge thermal characteristic table. Calculate the difference in mean temperature and the difference in temperature gradient between the two structural regions at the same heat conduction edge in the edge thermal characteristic table, and process the absolute values ​​of each to form temperature difference characteristic value and gradient characteristic value; The average temperature values ​​of the structural regions at both ends of the same heat conduction edge are collected in the order of collection time. Two edge temperature time series are constructed respectively, and mean-free normalization is performed to obtain the edge thermal response sequence. The edge thermal response sequence is synchronously segmented based on the acquisition time sequence and the preset window length to form a sliding time window; Within the sliding time window, the thermal response sequence at the edge is transformed in the frequency domain to extract the phase of the low-frequency component with the largest amplitude in the preset low-frequency band, and the phase difference between the two ends is calculated to form the window phase difference value. Calculate the change in window phase difference between adjacent sliding time windows, and mark the sliding time window whose change in window phase difference exceeds a preset phase threshold as a thermal response hysteresis window; Within the thermal response hysteresis window, infrared thermal images are selected where at least one of the temperature difference feature value and gradient feature value in the same heat conduction edge reaches a preset proportion of the maximum value of the same type of feature. The maximum number of consecutive infrared thermal images is counted to obtain the number of abnormal consecutive frames. Using the maximum value of the same type of feature and the total number of infrared thermal image frames as conversion benchmarks, the temperature difference feature value, gradient feature value and the number of consecutive abnormal frames are proportionally converted to form temperature difference score, gradient score and continuous score.

6. The method for detecting defects in supports and hangers based on infrared detection as described in claim 5, characterized in that: The formation of the candidate edge table specifically includes, The temperature difference score, gradient score and persistence score are weighted and fused to form a thermal contact anomaly score; The thermal contact anomaly score is compared with a preset score threshold, and thermal conduction edges whose thermal contact anomaly scores exceed the preset score threshold are selected to form candidate thermal conduction edges. A candidate edge table is formed by comparing the candidate heat conduction edges with additional temperature difference scores, gradient scores, continuity scores, and thermal contact anomaly scores.

7. The method for detecting defects in supports and hangers based on infrared detection as described in claim 6, characterized in that: The formation of the defect type table specifically includes, Obtain candidate heat conduction edges from the candidate edge list, and locate the two end structural regions of the candidate heat conduction edges according to the edge end node relationship in the topology table to form candidate edge pairs; The candidate edge pairs are compared with the structural region connection relationships in the thermal feature table. The structural region connection relationships with the same structural regions at both ends are retained to form candidate connection relationships. Based on the candidate connection relationship, temperature difference score, gradient score, persistence score and thermal contact anomaly score are associated to form a defect judgment parameter group; Select the highest score in the defect judgment parameter group, and label the dominant connection defect according to the score type to which the highest score belongs, to obtain the connection defect type; The thermal contact anomaly scores are compared in different levels to classify the defect levels. The candidate thermal conduction edges, candidate connection relationships, connection defect types, defect levels, and thermal contact anomaly scores are merged to form a defect type table.

8. The method for detecting defects in supports and hangers based on infrared detection as described in claim 7, characterized in that: The classification of defect levels specifically includes, According to the candidate connection relationship, the thermal contact anomaly score is matched from the defect judgment parameter group, and the thermal contact anomaly scores are concatenated in the order of the candidate connection relationship number to form a grade score sequence. Calculate the average and maximum scores of the rating sequence, and then calculate the mean of the average and maximum scores to obtain the high-level cutoff value. Based on the preset scoring threshold, average score, and high-level boundary value, the thermal contact anomaly scores of candidate connection relationships are compared in segments to classify them into first-level defects, second-level defects, and third-level defects, thus forming a defect level.

9. The method for detecting defects in supports and hangers based on infrared detection as described in claim 8, characterized in that: The formation of defect detection results specifically includes, Based on the candidate heat conduction edges in the defect type table, the connection defect type, defect level, and thermal contact anomaly score of the same candidate heat conduction edge are grouped into a defect edge record. The candidate heat conduction edges in the defect edge record are compared with the heat conduction edges in the topology table to match the edge-end node relationships of the candidate heat conduction edges and form defect edge-end relationships. The defect edge relationship is traced back to the thermal feature table to locate the structural regions at both ends of the candidate heat conduction edge, forming a defect structural region record. Based on the defective structural area record, match the support and hanger number in the thermal characteristic table to form a defect association record; Based on the defect association record, the defect edge record and defect structure area record are matched, and the support and hanger number, the structural areas at both ends, the candidate heat conduction edge, the heat contact anomaly score, the connection defect type and the defect level are combined to form the defect detection result.

10. A support and hanger defect detection system based on infrared detection, based on the support and hanger defect detection method based on infrared detection according to any one of claims 1 to 9, characterized in that: include, The feature table building module is used to collect the support and hanger number and infrared thermal image sequence, convert the infrared thermal image sequence into a temperature matrix sequence, divide the support and hanger into structural areas, calculate the average temperature and temperature gradient of the structural areas, identify the connection relationship of the structural areas, and form a thermal feature table. The topology building module is used to arrange the structural regions in the thermal feature table into heat conduction nodes, arrange the connection relationships of the structural regions into heat conduction edges, and match the average temperature and temperature gradient of the structural regions at both ends according to the connection relationship of the structural regions corresponding to the heat conduction edges to form a topology table. The anomaly scoring module is used to calculate the temperature difference score, gradient score and persistence score of each heat conduction edge in the topology table, and to obtain the heat contact anomaly score by weighting them. Heat conduction edges with heat contact anomaly scores exceeding the preset scoring threshold are filtered to form a candidate edge table. The defect determination module is used to read heat conduction edges from the candidate edge table, match the connection relationship between the structural regions at both ends of the heat conduction edge in the thermal feature table, mark the connection defect type and classify the defect level to form a defect type table. The result generation module compares the candidate heat conduction edges in the defect type table with the heat conduction edges in the topology table to determine the structural areas at both ends, matches the support and hanger numbers in the thermal feature table, and combines the thermal contact anomaly score, connection defect type and defect level to form the defect detection result.

Images