A curtain wall welding defect early warning method and system based on big data analysis

Abstract

The application relates to the technical field of construction quality control of building curtain walls, and discloses a curtain wall welding defect early warning method and system based on big data analysis, which comprises the following steps: collecting curtain wall welding arc amplitude and energy signals and labeling mutation nodes, monitoring welding arc length changes to construct abnormal node cluster information, combining current deviation calculation and drawing a key node distribution map, performing multi-dimensional abnormal joint analysis to map a welding seam section risk, and finally screening abnormal areas and establishing a defect early warning identification table. By analyzing the node abnormality and cluster characteristics of welding amplitude, energy, length and current data, quantifying the spatial distribution and fluctuation law, accurately labeling the potential defect area and the welding seam section risk level, realizing the evaluation span from a single point to a paragraph, comprehensively evaluating the abnormal concentration area and visualizing the risk mode, thus a precise early warning mechanism is established, the construction risk is significantly reduced, and the quality management process of the whole curtain wall welding process is comprehensively optimized.

Description

Technical Field

[0001] This invention relates to the field of building curtain wall construction quality control technology, specifically to a method and system for early warning of welding defects in curtain walls based on big data analysis. Background Technology

[0002] The technical field of building curtain wall construction quality control encompasses aspects such as curtain wall material selection, curtain wall structural design, construction process specifications, welding process management, and on-site inspection and monitoring. The core content of this technical field mainly involves quality control methods for curtain wall components during installation and welding, including weld joint strength testing, weld appearance inspection, construction parameter recording, and construction environment monitoring. The overall technical field systematically covers pre-construction design verification, construction process monitoring, and post-construction quality assessment, ensuring that curtain wall construction meets design and specification requirements through material performance analysis, construction process optimization, on-site inspection, and data management.

[0003] Among them, the curtain wall welding defect early warning method and system based on big data analysis refers to a method and system for collecting, organizing, and analyzing various construction data generated during the welding process. This patent addresses potential defects that may occur during curtain wall welding, including weld joint cracks, uneven welds, and welding parameter deviations. It achieves this through establishing a welding data acquisition mechanism, data cleaning and classification, statistical analysis of defect characteristics, and setting early warning rules. Specific methods include collecting data on welding current and voltage, welding speed, and temperature changes; organizing the data according to time and welding location; statistically analyzing weld quality indicators; combining rules or thresholds to determine potential defects; and generating early warning information through the system for identification.

[0004] Existing technologies rely solely on single-parameter statistics and threshold judgments in welding construction, lacking comprehensive analysis of the correlation between amplitude, energy, length, and current. Continuous abnormal fluctuations and node cluster characteristics cannot be quantified, the potential defect range and risk level of weld segments are difficult to define, and the ranking and spatial distribution of abnormal patterns cannot be presented. This results in delayed or incomplete defect warning information, weld quality assessment relying on experience-based judgment, and welding defects may be missed or misjudged, limiting the accuracy and reliability of construction quality control. Summary of the Invention

[0005] The present invention aims to provide a method and system for early warning of welding defects in curtain walls based on big data analysis, which solves the problem of relying solely on single parameter statistics and threshold judgment in the existing technology. It realizes the joint analysis of multi-dimensional data such as amplitude, energy, length, and current during the welding process, quantifies node anomalies, cluster characteristics and weld segment risks, establishes a precise early warning mechanism, and improves the accuracy of construction quality control.

[0006] The basic solution provided by this invention is: a method for early warning of welding defects in curtain walls based on big data analysis, comprising: S1: Collect the welding arc amplitude signal and instantaneous energy signal of the welding power source of the welding robot arm at the construction site for early warning of welding defects in the curtain wall. Organize the amplitude and energy data in chronological order, compare the amplitude changes and energy deviations of adjacent nodes, mark the nodes with abrupt amplitude changes and record the energy status, call the node status to establish the amplitude-energy correspondence, and obtain the welding arc amplitude-energy status nodes. S2: Monitor the changes in welding arc length during the welding process, identify continuous abnormal nodes and arrange their spatial positions, count the duration and fluctuation range of abnormalities, analyze the relationships and abnormal states of nodes within the cluster, call spatial distribution information to form abnormal cluster labels, and obtain information on abnormal arc length node clusters. S3: Based on the welding arc amplitude energy state node and arc length abnormal node cluster information, the welding current signal is collected, the instantaneous deviation of each node is calculated and the difference before and after is compared, the number of abnormal fluctuations is counted, the relationship between key nodes and neighboring nodes is analyzed, the amplitude energy and length state are called to mark the location of abnormal nodes, and the distribution map of abnormal welding current key nodes is obtained. S4: Combining the distribution map of key nodes with abnormal welding current, cluster information of nodes with abnormal arc length, and nodes with energy status of welding arc amplitude, perform joint anomaly analysis on the weld segment, compare the node anomaly concentration area with the status of adjacent nodes, call the spatial distribution and anomaly concentration situation to form a segment risk mapping, and obtain the defect risk distribution map of the weld segment. S5: Call the defect risk distribution map of the weld segment, sort and filter the abnormal concentration areas of the weld trajectory, mark the location and risk level of key abnormal nodes of the weld segment, analyze the risk concentration pattern of each segment and establish early warning signs, and obtain the curtain wall welding defect early warning sign table.

[0007] The welding arc amplitude energy state nodes include amplitude anomaly points, energy anomaly points, and node state mappings; the arc length anomaly node cluster information includes anomaly intervals, anomaly amplitudes, and cluster spatial relationships; the welding current anomaly key node distribution map includes current deviation points, continuous anomaly points, and anomaly concentration areas; the weld segment defect risk distribution map includes anomaly dense areas, adjacent related areas, and potential defect segments; and the curtain wall welding defect early warning identification table includes key anomaly nodes, risk levels, and early warning identifications.

[0008] This invention also provides a curtain wall welding defect early warning system based on big data analysis, to execute a curtain wall welding defect early warning method based on big data analysis. The system includes: The welding amplitude acquisition module collects the welding arc amplitude signal and instantaneous energy signal of the welding power source of the welding robot arm at the construction site for early warning of welding defects in curtain walls. It organizes the amplitude and energy data in chronological order, analyzes the amplitude changes and energy deviations of adjacent nodes, marks the nodes with abrupt amplitude changes and records the energy state, establishes the correspondence between amplitude and energy nodes, and obtains the welding arc amplitude and energy state nodes. The welding length monitoring module monitors the changes in welding arc length during the welding process, identifies continuous length abnormal nodes and arranges their spatial positions, counts the duration and fluctuation range of abnormalities, analyzes the length relationship of nodes within the cluster, calls spatial distribution information to divide abnormal node clusters, and obtains information on arc length abnormal node clusters. The welding current analysis module, based on the welding arc amplitude energy state node and arc length abnormal node cluster information, collects welding current signals, calculates the instantaneous deviation of each node and compares the sampling difference before and after, counts the number of consecutive abnormal fluctuations, analyzes the abnormal relationship between key nodes and adjacent nodes, filters the abnormal concentrated area of ​​weld trajectory and marks the location of key nodes, calls amplitude energy and length state for abnormal comparison, and obtains the distribution map of abnormal welding current key nodes. The weld risk mapping module combines the distribution map of key nodes with abnormal welding current, cluster information of nodes with abnormal arc length, and nodes with energy status of welding arc amplitude to perform joint anomaly analysis on the weld segment. It compares the relationship between the node anomaly concentration area and adjacent nodes, calls the state of the abnormal node concentration to establish the defect risk mapping of the weld segment, and obtains the defect risk distribution map of the weld segment. The weld seam early warning and identification module calls the defect risk distribution map of the weld seam segment, sorts and filters abnormal areas of the weld seam trajectory, marks the high-risk weld seam segments and key abnormal node locations, establishes early warning identification, and obtains the curtain wall welding defect early warning identification table.

[0009] The working principle and advantages of this invention are as follows: This invention identifies and analyzes node-level anomalies and clusters in amplitude, energy, length, and current data during the welding process. The spatial distribution and continuous fluctuation characteristics of abnormal nodes are quantified, potential defect areas and weld segment risk levels are marked, key abnormal node rankings and risk concentration patterns are visualized, the relationship between anomaly concentration areas and adjacent nodes is comprehensively evaluated, and multi-indicator joint judgment enhances defect identification capabilities. Weld quality assessment is expanded from single-point judgment to segment-level risk analysis, and defect warnings during construction can be targeted for control, significantly reducing potential welding risks and optimizing construction quality management. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating a method for early warning of welding defects in curtain walls based on big data analysis, provided in an embodiment of the present invention. Figure 2This is a schematic diagram of a curtain wall welding defect early warning system based on big data analysis, provided in an embodiment of the present invention. Detailed Implementation

[0011] The following detailed explanation illustrates the specific implementation methods: The basic implementation examples are as follows: Figure 1 As shown: A method for early warning of welding defects in curtain walls based on big data analysis, including: S1: Collect the welding arc amplitude signal and instantaneous energy signal of the welding power source of the welding robot arm at the construction site for early warning of welding defects in the curtain wall. Extract the amplitude data and energy data in chronological order along the weld trajectory. Determine the amplitude change range and energy deviation of adjacent sampling points. Mark the amplitude change nodes and record the energy status. Establish the correspondence between amplitude and energy between nodes. Call the above status to complete the anomaly comparison between nodes and obtain the welding arc amplitude and energy status nodes.

[0012] Specifically, S1 includes: S101: Collect the arc amplitude signal and instantaneous power energy signal of the welding robot arm at the construction site for early warning of welding defects in curtain walls. Align the sampling times according to the time sequence of the weld trajectory, extract the amplitude value, energy value, trajectory position, and sampling sequence number, verify the continuity of the sampling sequence number, remove sampling points corresponding to missing sampling times, and generate a weld trajectory amplitude energy sequence; details are as follows: Arc amplitude sensors mounted on the welding robotic arm collect amplitude signals during the welding process, while an energy monitoring module inside the welding power source collects instantaneous energy signals. During data acquisition, the arc amplitude sensor continuously monitors the arc amplitude at a sampling frequency of 5000Hz, while the energy monitoring module simultaneously collects instantaneous energy data at a sampling frequency of 2500Hz. The sampling times of both are aligned using a unified timestamp to ensure data consistency across time.

[0013] In practice, from the start to the end of welding, the system extracts the amplitude value, energy value, weld trajectory position coordinates, and sampling sequence number sequentially. The continuity of the sampling sequence number is checked by comparing the difference between adjacent sampling sequence numbers. If the difference is equal to 1, it is considered continuous; if the difference is greater than 1, a missing sampling moment is identified, and the sampling point data corresponding to that missing sampling moment needs to be removed. Through the above processing, a weld trajectory amplitude-energy sequence containing amplitude value, energy value, trajectory position, and sampling sequence number is finally generated.

[0014] For example, in a curtain wall welding operation, the welding path was 2.5 meters long, the sampling time span was 125 seconds, and a total of 625,000 valid sampling points were collected. Table 1 lists some examples of data from the sampling points: Table 1: Examples of Weld Trajectory Amplitude Energy Sequences As shown in Table 1, the amplitude values ​​of sampling numbers 1001 to 1004 remained between 1.48 mm and 1.55 mm, and the energy values ​​remained between 2832 J and 2865 J, which is considered normal welding condition. At sampling number 1005, the amplitude value suddenly increased to 3.82 mm, and the energy value increased to 3520 J, indicating that there was an anomaly at this sampling point.

[0015] S102: Based on the weld trajectory amplitude energy sequence, calculate the amplitude difference between adjacent sampling points. Compare the absolute value of the amplitude difference with a preset amplitude variation limit. Determine the interval by comparing the energy value with a preset energy reference interval. Mark sampling points exceeding the preset amplitude variation limit as amplitude abrupt change nodes. Establish a node amplitude energy association table by associating amplitude abrupt change nodes, trajectory position, sampling time, energy-high state, energy-stable state, and energy-low state. Details are as follows: Anomaly node labeling is performed based on the amplitude energy sequence of the weld trajectory. First, the amplitude difference between adjacent sampling points is calculated, i.e., the amplitude value of the current sampling point is subtracted from the amplitude value of the previous sampling point. The absolute value of the amplitude difference is compared with a preset amplitude variation limit, which is determined based on statistical analysis of a large number of normal welding samples. The 95th percentile of the amplitude variation under normal welding conditions is taken as the limit, and statistical calculations determine the amplitude variation limit to be 1.20 mm. Simultaneously, the energy value is compared with a preset energy reference range. The energy reference range is determined based on the distribution range of energy values ​​during normal welding, taking the 5th to 95th percentile of the energy value distribution as the reference range, which is calculated to be 2500 J to 3200 J.

[0016] When the absolute value of the amplitude difference exceeds 1.20 mm, the sampling point is marked as an amplitude abrupt change node. Next, based on the positional relationship between the energy value and the energy reference interval, the energy state is determined: an energy value greater than 3200 J is marked as an energy-high state, an energy value between 2500 J and 3200 J is marked as an energy-stable state, and an energy value less than 2500 J is marked as an energy-low state. Finally, by associating the trajectory position, sampling time, energy-high state, energy-stable state, and energy-low state of the amplitude abrupt change node, a node amplitude-energy correlation table is established.

[0017] Taking the data in Table 1 as an example, the amplitude difference between sampling numbers 1004 and 1005 is 3.82 mm minus 1.51 mm, which equals 2.31 mm. The absolute value of the amplitude difference is 2.31 mm, which is greater than the amplitude variation limit of 1.20 mm. Therefore, sampling number 1005 is marked as an amplitude abrupt change node. The energy value of sampling number 1005 is 3520 J, which is greater than the upper limit of the energy reference range of 3200 J. Therefore, the energy state of this node is marked as an upper-energy state.

[0018] S103: Call the node amplitude energy association table, read the amplitude direction, energy state, and trajectory position interval between adjacent amplitude change nodes segment by segment, compare the consistency between the amplitude direction change state and the energy deviation direction, determine the matching relationship between the number of energy state continuation points and the trajectory position interval, mark the node segment with the amplitude direction reversed and the energy deviation direction continuing in the same direction as the node anomaly, associate the anomaly mark with the node number, and obtain the welding arc amplitude energy state node; as follows: The node amplitude-energy correlation table is invoked to analyze data segment by segment between adjacent nodes with abrupt amplitude changes. Specific data read includes amplitude direction, energy state, and trajectory position interval. The amplitude direction is determined by the difference in amplitude values ​​between adjacent nodes; a positive difference indicates an upward amplitude direction, while a negative difference indicates a downward amplitude direction. The energy deviation direction is determined by comparing the energy value with the median value of the energy reference interval; a positive deviation indicates a positive energy deviation, while a negative deviation indicates a negative energy deviation.

[0019] When comparing the amplitude direction change state with the energy deviation direction consistency, if a node segment exhibits an amplitude direction reversal and an energy deviation direction continuing in the same direction, that node segment is marked as an inter-node anomaly. An amplitude direction reversal means that the amplitude direction between adjacent amplitude abrupt change nodes is opposite to the amplitude direction of the previous segment; a continuation of the same energy deviation direction means that the energy deviation direction remains consistent within the node segment. When determining the matching relationship between the number of energy state continuation points and the trajectory position interval, the number of energy state continuation points refers to the number of sampling points where the energy-biased state or energy-biased state occurs consecutively, and the trajectory position interval refers to the spatial distance between adjacent amplitude abrupt change nodes. When the ratio of the number of energy state continuation points to the trajectory position interval is greater than a preset matching threshold, it is judged as an anomaly. Finally, the anomaly marker is associated with the node sequence number to obtain the welding arc amplitude energy state node.

[0020] For example, the amplitude values ​​of three consecutive amplitude abrupt change nodes are as follows: Node A amplitude 3.82 mm (energy slightly higher), Node B amplitude 1.45 mm (energy stable), and Node C amplitude 3.95 mm (energy slightly higher). The amplitude direction from Node A to Node B is decreasing (3.82 mm to 1.45 mm), and the amplitude direction from Node B to Node C is increasing (1.45 mm to 3.95 mm), indicating a reversal of amplitude direction. In terms of energy state, Node A is energy slightly higher, Node B is energy stable, and Node C is energy slightly higher. The energy deviation direction is generally positive within the interval from Node A to Node C, and the direction is the same at both ends, showing a continuous characteristic of the same direction. Therefore, this node segment is marked as an inter-node anomaly and recorded as a welding arc amplitude energy state node.

[0021] S2: Monitor the change in welding arc length during the welding process, perform length deviation judgment for continuous sampling points, count the duration of abnormality and mark the fluctuation range, arrange abnormal nodes according to the spatial location of the weld, analyze the length change relationship and fluctuation amplitude between nodes within the cluster, compare the abnormal duration of each cluster and screen out prominent length abnormal areas, call spatial distribution information to establish cluster association, and obtain cluster information of arc length abnormal nodes.

[0022] Specifically, S2 includes: S201: Monitor the change in arc length during the welding process. Extract the sampling time, weld spatial location, and arc length value from continuous sampling points. Calculate the difference between the arc length value and the preset arc length benchmark value. Compare the absolute value of the difference with the preset length deviation limit. Mark sampling points that exceed the preset length deviation limit as length deviation nodes, obtaining a length deviation node sequence. Details are as follows: The arc length is monitored during the welding process. It is measured by an arc length sensor mounted on the welding torch head, which collects data at a sampling frequency of 1000Hz. The arc length sensor's measurement principle is based on the linear relationship between arc voltage and welding current, specifically calculated as: arc length equals arc voltage divided by the arc voltage gradient coefficient. This coefficient is determined through standard arc length calibration experiments, with a commonly used value of 3.5V / mm.

[0023] After extracting the sampling time, weld spatial location, and arc length value from continuous sampling points, the difference between the arc length value and the preset arc length benchmark value is calculated. The preset arc length benchmark value is determined based on parameters such as welding wire type, welding current, and welding voltage. For ER50-6 welding wire with a diameter of 1.2mm, under the conditions of welding current of 200A and welding voltage of 24V, the standard arc length benchmark value is 8.0mm. The absolute value of the difference is compared with the preset length deviation limit. The preset length deviation limit is determined based on the normal fluctuation range of the arc length, and 1.5 times the standard deviation of the normal fluctuation range is taken as the limit value, which is calculated to be 2.5mm. When the absolute value of the difference exceeds 2.5mm, the sampling point is marked as a length deviation node, resulting in a length deviation node sequence.

[0024] For example, the arc length at a certain sampling point is 11.2 mm, which is 3.2 mm different from the reference value of 8.0 mm. The absolute value of the difference, 3.2 mm, is greater than the length deviation limit of 2.5 mm. Therefore, this sampling point is marked as a length deviation node.

[0025] S202: Based on the length deviation node sequence, the difference in sampling time between adjacent length deviation nodes and the spatial position interval of the weld are statistically analyzed. Nodes with continuous sampling times and adjacent spatial positions are merged. The duration and length fluctuation amplitude of each segment are calculated. The duration is compared with a preset duration limit, and the length fluctuation amplitude is compared with a preset fluctuation amplitude limit. Fluctuation intervals are marked, and a table of abnormal length fluctuation intervals is generated. Details are as follows: Abnormal fluctuation intervals are labeled based on the length deviation node sequence. First, the sampling time difference and weld spatial position interval of adjacent length deviation nodes are calculated. When the sampling time of adjacent nodes is continuous (i.e., the time difference is less than or equal to 1.5 times the sampling interval) and the spatial position is adjacent (i.e., the position difference is less than or equal to 2 times the welding wire diameter), these nodes are merged to form a continuous abnormal segment.

[0026] The duration and length fluctuation amplitude of each segment are calculated. The duration is the difference between the sampling time of the last node and the sampling time of the first node within that segment. The length fluctuation amplitude is the difference between the maximum and minimum arc length within that segment. The duration is compared with a preset duration limit, which is determined based on welding process requirements and is set to 0.5 seconds. The length fluctuation amplitude is compared with a preset fluctuation amplitude limit, which is determined based on the normal arc length fluctuation amplitude and is set to 3.0 mm. When the duration exceeds 0.5 seconds and the length fluctuation amplitude exceeds 3.0 mm, it is marked as a fluctuation interval, and a table of abnormal length fluctuation intervals is generated.

[0027] For example, a continuous abnormal segment contains 5 length deviation nodes. The sampling time of the first node is 12.345 seconds, and the sampling time of the last node is 12.890 seconds, with a duration of 0.545 seconds, exceeding the duration limit of 0.5 seconds. The maximum arc length in this segment is 11.8 mm, the minimum is 6.2 mm, and the length fluctuation range is 5.6 mm, exceeding the fluctuation range limit of 3.0 mm. Therefore, this segment is marked as a length abnormal fluctuation range.

[0028] S203: Call the abnormal length fluctuation interval table, arrange abnormal nodes according to the spatial location of the weld, compare the length change direction and length difference of adjacent nodes within the same interval, associate adjacent fluctuation intervals according to spatial location intervals, compare the duration, number of nodes, and length fluctuation amplitude of each interval, filter areas where the duration exceeds the preset duration limit and the number of nodes reaches the preset node number limit, establish cluster association, and obtain arc length abnormal node cluster information; details are as follows: The system retrieves the abnormal arc length fluctuation interval table and arranges abnormal nodes according to their spatial location within the weld. It compares the direction of length change and the length difference between adjacent nodes within the same interval. The direction of length change is determined by the sign of the arc length difference between adjacent nodes, which is the absolute difference in arc length values. Adjacent fluctuation intervals are associated at spatial intervals, and the duration, number of nodes, and amplitude of length fluctuation are compared for each interval. Regions whose duration exceeds the preset duration limit (0.5 seconds) and whose number of nodes reaches the preset node number limit (10) are selected, and cluster associations are established to obtain information on the abnormal arc length node clusters.

[0029] The limit of 10 nodes is determined based on the following: Under normal welding conditions, arc length deviations from nodes are usually discretely distributed, and the number of consecutive deviation nodes generally does not exceed 5. When the number of consecutive deviation nodes reaches 10 or more, it indicates that there is a persistent arc length anomaly in the area, and cluster marking is required.

[0030] S3: Based on the welding arc amplitude energy status node and arc length abnormal node cluster information, the welding current signal is collected, the instantaneous current deviation is calculated for each sampling point and the difference between the nodes before and after is compared, the number of consecutive abnormal fluctuations is counted, the abnormal relationship between key nodes and adjacent nodes is analyzed, the abnormal concentrated area is determined and the abnormal node position is marked, the amplitude energy and length status are called to make a comprehensive judgment on the abnormality, and the distribution map of key nodes of welding current abnormality is obtained.

[0031] Specifically, S3 includes: S301: Acquire welding process current signals, match weld space positions according to sampling time, extract current values, sampling sequence numbers, adjacent preceding nodes, and adjacent following nodes, calculate the difference between the current value and the preset current reference value, compare the absolute value of the difference with the preset current deviation limit, compare the direction of change of the difference between adjacent preceding nodes and adjacent following nodes, and obtain the welding current deviation node sequence; as detailed below: The welding process current signal is acquired through a current sensor inside the welding power supply, with a sampling frequency of 5000Hz. The weld spatial position is matched according to the sampling time, and the current value, sampling sequence number, preceding node, and following node are extracted. The difference between the current value and the preset current reference value is calculated. The preset current reference value is set according to the welding process; for curtain wall welding, the commonly used welding current of 200A is set to 200A. The absolute value of the difference is compared with a preset current deviation limit, which is determined based on the normal fluctuation range of the welding current. Twice the standard deviation of the normal fluctuation range is taken as the limit, which is calculated to be 15A. The direction of change of the difference between the preceding and following nodes is compared. When the signs are the same and the absolute difference is greater than the preset change direction threshold (5A), the deviation direction is considered consistent, resulting in a welding current deviation node sequence.

[0032] For example, if the current value at a sampling point is 218A, the difference from the reference value of 200A is 18A. The absolute value of the difference, 18A, is greater than the current deviation limit of 15A. Therefore, this sampling point is marked as a current deviation node. Comparing the direction of the difference between this node and its previous deviation node (current value 215A, difference 15A), both differences are positive and their absolute values ​​are 18A and 15A respectively. The change in difference is 3A, which is less than the change direction threshold of 5A. Therefore, it is determined that the deviation direction is consistent.

[0033] S302: Based on the welding current deviation node sequence, count the number of sampling points that continuously exceed the preset current deviation limit. Compare the number of sampling points with the preset fluctuation limit to extract the spatial location interval of adjacent abnormal nodes. Determine the consistency of the deviation direction between key nodes and adjacent nodes, mark the location of abnormal nodes, and generate a table of concentrated current abnormality areas; details are as follows: Anomaly concentration areas are marked based on the welding current deviation node sequence. The number of sampling points continuously exceeding a preset current deviation limit is counted. When the number of consecutive sampling points reaches a preset fluctuation limit, it is marked as an abnormal fluctuation. The preset fluctuation limit is determined based on the duration of the current anomaly, and is set to 8 sampling points, corresponding to an anomaly duration of approximately 1.6 milliseconds. The spatial interval between adjacent anomaly nodes is extracted, and the consistency of the deviation direction between key nodes and neighboring nodes is determined, generating a table of current anomaly concentration areas.

[0034] For example, if the current deviation value of 12 consecutive sampling points all exceeds 15A, and the number of continuous sampling points is 12, which is greater than the fluctuation limit of 8, then this area is marked as a region of concentrated current anomalies. The spatial interval between nodes in this region is between 0.01m and 0.03m, and the deviation direction of adjacent nodes is positive, indicating that there is a continuous positive current deviation in this region.

[0035] S303: Call the current anomaly concentration area table, associate the welding arc amplitude energy state nodes with the arc length anomaly node cluster information, map the current anomaly concentration area, amplitude energy state nodes, and length anomaly node cluster according to the weld spatial location, compare the spatial overlap relationship and anomaly persistence state of the three types of nodes, mark the distribution location of key nodes, and obtain the distribution map of key nodes of welding current anomalies; as detailed below. The table of concentrated current anomalies is retrieved, and the welding arc amplitude energy status nodes and arc length anomaly node cluster information are associated. Current anomaly concentrated areas, amplitude energy status nodes, and length anomaly node clusters are mapped according to the spatial location of the weld, and the spatial overlap and anomaly duration of the three types of nodes are compared. When the three types of nodes overlap in spatial location (i.e., the positional interval is less than 0.05m) and the anomaly duration is similar (the duration difference is less than 0.2 seconds), they are marked as key node distribution locations, and a distribution map of key nodes for welding current anomalies is obtained.

[0036] The advantage of this step is that by spatially correlating multi-dimensional anomaly information, it is possible to accurately locate high-incidence areas of welding defects, thereby improving the accuracy of defect early warning. While single-dimensional anomaly assessment may lead to misjudgments, the combined assessment of three types of anomaly information—amplitude energy, length, and current—can effectively eliminate sporadic interference and identify genuine welding defects.

[0037] S4: Combining the distribution map of key nodes with abnormal welding current, cluster information of nodes with abnormal arc length, and nodes with welding arc amplitude and energy status, perform joint analysis of nodes with abnormal amplitude, energy, length, and current on the weld segment. Compare the node abnormal dense area and the correlation status of adjacent nodes to determine the potential defect range of the weld segment. Call the spatial distribution and the concentration of abnormal nodes to establish a segment risk mapping and obtain a defect risk distribution map of the weld segment.

[0038] Specifically, S4 includes: S401: Obtain weld segment division parameters, call the distribution map of key nodes for abnormal welding current, cluster information of abnormal arc length nodes, and welding arc amplitude energy status nodes, map amplitude nodes, energy nodes, length nodes, and current nodes to the corresponding weld segments according to the spatial location of the weld, count the number of nodes and node spacing in each weld segment, and establish a correlation table for abnormal nodes in weld segments; details are as follows: The weld segmentation parameters are obtained, including the segment length division benchmark and the node spacing threshold. The segment length division benchmark is determined according to the curtain wall welding process requirements and is set to 0.5m, meaning that every 0.5m weld is divided into one weld segment. The node spacing threshold is 0.02m and is used to determine the correlation between nodes.

[0039] The system retrieves the distribution map of critical nodes for welding current anomalies, the cluster information of abnormal arc length nodes, and the welding arc amplitude and energy status nodes. Based on the spatial location of the weld, it maps amplitude nodes, energy nodes, length nodes, and current nodes to the corresponding weld segments. It then counts the number of amplitude nodes, energy nodes, length nodes, and current nodes within each weld segment, as well as the node spacing, and establishes a correlation table for abnormal nodes in each weld segment.

[0040] For example, a weld section (from station 0.5m to 1.0m) contains 3 amplitude nodes, 2 energy nodes, 4 length nodes, and 5 current nodes. The minimum spacing between each type of node is 0.015m, and the maximum spacing is 0.185m.

[0041] S402: Based on the abnormal node association table of weld segments, calculate the ratio of amplitude nodes, energy nodes, length nodes, current nodes to segment length for each weld segment, compare the ratio with the preset node density limit, compare the node spacing between adjacent weld segments with the preset adjacent spacing limit, determine the abnormal node density area and the association status of adjacent nodes, and generate a potential defect range table for the weld segment; details are as follows: The defect range is determined based on the abnormal node association table of weld segments. The node density is obtained by calculating the ratio of the number of amplitude nodes, energy nodes, length nodes, and current nodes to the segment length (0.5m) for each weld segment. The node density is compared with a preset node density limit, which is determined based on the node density distribution under normal welding conditions and is set to 8 nodes / m. The node spacing between adjacent weld segments is compared with a preset adjacent spacing limit, which is set to 0.03m. When the node density is greater than 8 nodes / m and the node spacing between adjacent weld segments is less than 0.03m, it is identified as an abnormally dense node area and an adjacent node association state, generating a potential defect range table for the weld segment.

[0042] The innovation of this step lies in its ability to identify potential defect areas within a weld segment through a dual assessment of node density and adjacent spacing. Node density reflects the concentration of anomalies, while adjacent spacing reflects the spatial continuity of anomalies; combining the two can effectively locate weld segments requiring focused attention.

[0043] S403: Based on the potential defect range table of weld segments, associate the spatial location of the weld, the concentration level of abnormal nodes, and the association status of adjacent nodes. Match the risk level of each weld segment according to the preset risk classification rules, establish a key-value mapping between the weld segment number and the risk level, arrange the key-value mapping according to the spatial location of the weld, and obtain the defect risk distribution map of the weld segment; as detailed below: Risk classification is performed based on the potential defect range table of weld segments. The risk level of each weld segment is matched according to the spatial location of the weld, the concentration level of abnormal nodes, and the association status of adjacent nodes, based on preset risk classification rules. The risk classification rules are based on specific values ​​of node density and adjacent spacing: a node density greater than 20 / m and an adjacent spacing less than 0.015m is considered high risk; a node density of 12 to 20 / m and an adjacent spacing of 0.015 to 0.025m is considered medium risk; and a node density of 8 to 12 / m and an adjacent spacing of 0.025 to 0.03m is considered low risk. A key-value mapping is established between the weld segment number and the risk level, and the weld segment defect risk distribution map is obtained after arranging them according to the spatial location of the weld.

[0044] For example, if a weld segment has a node density of 22 nodes / m and an adjacent spacing of 0.012m, it matches the high-risk level rule and is determined to be a high-risk weld segment.

[0045] S5: Call the defect risk distribution map of the weld segment, sort and filter the abnormal concentration areas of the weld trajectory, mark the location and risk level of key abnormal nodes of the weld segment, analyze the risk concentration pattern of each segment and establish early warning signs, and obtain the curtain wall welding defect early warning sign table.

[0046] Specifically, S5 includes: S501: Retrieve the defect risk distribution map of the weld segment, extract the weld segment number, spatial start point, spatial end point, risk level, and number of abnormal nodes according to the weld trajectory sequence, calculate the difference in risk level and the difference in the number of abnormal nodes between adjacent weld segments, arrange the weld segments in descending order of risk level, use preset risk level limits and preset node number limits to filter areas with concentrated anomalies, and generate a weld trajectory anomaly ranking list; details are as follows: The defect risk distribution map of the weld segment is retrieved, and the weld segment number, spatial start point, spatial end point, risk level, and number of abnormal nodes are extracted according to the weld trajectory sequence. The differences in risk level and the number of abnormal nodes between adjacent weld segments are calculated, and the weld segments are arranged in descending order of risk level. Preset risk level limits and preset node number limits are used to filter areas with concentrated anomalies. The preset risk level limit is medium risk and above (i.e., high risk and medium risk), and the preset node number limit is 15. A ranked list of abnormal weld trajectories is generated.

[0047] For example, a weld trajectory contains 20 weld segments, of which 5 segments are high-risk, 8 segments are medium-risk, and 7 segments are low-risk. After screening, the 13 high-risk and medium-risk segments are retained, and an anomaly sorting list is generated by arranging them in descending order of risk level.

[0048] S502: Based on the weld trajectory anomaly sorting list, extract the spatial location and risk level of key anomaly nodes within the anomaly concentration area, compare the spatial interval between nodes with the preset node proximity limit, associate the node number, node location, and risk level within the same weld segment, establish a node risk mapping according to the weld segment number, and generate a key node risk labeling table; details are as follows: Key nodes are labeled based on the weld trajectory anomaly sorting list. The spatial location and risk level of key anomaly nodes within the anomaly concentration area are extracted, and the spatial interval between nodes is compared with a preset node proximity limit of 0.02m. Node numbers, node locations, and risk levels within the same weld segment are associated, and a node risk mapping is established according to the weld segment number, generating a key node risk labeling table.

[0049] This step, through node-level risk labeling, can accurately locate the position of each abnormal node that needs attention, providing specific location information for subsequent early warning indicators.

[0050] S503: Based on the critical node risk labeling table, count the consecutive frequency of risk levels and the number of clusters of critical abnormal nodes for each weld segment. Compare the consecutive frequency with the preset continuous risk limit, and compare the cluster number with the preset cluster number limit to determine the risk concentration pattern for each segment. Associate the weld segment number, critical abnormal node location, risk level, and risk concentration pattern to establish early warning indicators and obtain the curtain wall welding defect early warning indicator table; details are as follows: Risk concentration patterns are determined based on the critical node risk labeling table. The number of consecutive risk levels and the number of clusters of critical abnormal nodes for each weld segment are counted. The number of consecutive risk levels is compared with a preset continuous risk limit (3 times). The number of clusters is compared with a preset cluster limit (8 nodes). The risk concentration patterns for each segment are determined as follows: a high concentration pattern is defined as 3 or more consecutive risk levels and 8 or more clusters; a medium concentration pattern is defined as 2 consecutive risk levels and 5 to 8 clusters; and a low concentration pattern is defined as 1 consecutive risk level and less than 5 clusters.

[0051] By associating weld segment numbers, key abnormal node locations, risk levels, and risk concentration patterns, early warning indicators are established, and a curtain wall welding defect early warning indicator table is obtained. The benefit of this step is that by determining the risk concentration pattern, the overall risk trend of the weld segment can be identified, providing the construction party with clear early warning levels and location information, facilitating targeted quality control measures. Table 2 lists examples of early warning indicators for some weld segments.

[0052] Table 2: Examples of Early Warning Signs for Welding Defects in Curtain Walls As shown in Table 2, weld segments 3 and 7 are classified as red alerts, indicating that these two segments have high-risk welding defects and require immediate quality inspection and rectification. Weld segment 11 is classified as an orange alert, weld segment 15 as a yellow alert, and weld segment 19 as a blue alert, corresponding to different risk levels and handling priorities, respectively.

[0053] like Figure 2 As shown, this embodiment also provides a curtain wall welding defect early warning system based on big data analysis, which implements a curtain wall welding defect early warning method based on big data analysis. The system includes: The welding amplitude acquisition module collects the welding arc amplitude signal and instantaneous energy signal of the welding power source of the welding robot arm at the construction site for early warning of welding defects in curtain walls. It organizes the amplitude and energy data in chronological order, analyzes the amplitude changes and energy deviations of adjacent nodes, marks the nodes with abrupt amplitude changes and records the energy state, establishes the correspondence between amplitude and energy nodes, and obtains the welding arc amplitude and energy state nodes. The welding length monitoring module monitors the changes in welding arc length during the welding process, identifies continuous length abnormal nodes and arranges their spatial positions, counts the duration and fluctuation range of abnormalities, analyzes the length relationship of nodes within the cluster, calls spatial distribution information to divide abnormal node clusters, and obtains information on arc length abnormal node clusters. The welding current analysis module collects welding current signals based on the welding arc amplitude energy state nodes and arc length abnormal node cluster information, calculates the instantaneous deviation of each node and compares the difference between the previous and subsequent samples, counts the number of consecutive abnormal fluctuations, analyzes the abnormal relationship between key nodes and adjacent nodes, filters the abnormal concentrated area of ​​weld trajectory and marks the location of key nodes, calls the amplitude energy and length state for abnormal comparison, and obtains the distribution map of key nodes with abnormal welding current. The weld risk mapping module combines the distribution map of key nodes with abnormal welding current, the cluster information of nodes with abnormal arc length, and the energy status nodes of welding arc amplitude to perform joint anomaly analysis on the weld segment. It compares the relationship between the node anomaly concentration area and the neighboring nodes, calls the concentrated state of abnormal nodes to establish the defect risk mapping of the weld segment, and obtains the defect risk distribution map of the weld segment. The weld seam early warning and identification module calls up the weld seam segment defect risk distribution map, sorts and filters abnormal areas of weld seam trajectory, marks high-risk weld seam segments and key abnormal node locations, establishes early warning identification, and obtains the curtain wall welding defect early warning identification table.

[0054] It is understandable that the above system and method have the same execution process and the same effect, so they will not be described again here.

[0055] This embodiment provides a method and system for early warning of welding defects in curtain walls based on big data analysis. It solves the problem of relying solely on single parameter statistics and threshold judgment in the prior art. It realizes the joint analysis of multi-dimensional data such as amplitude, energy, length, and current during the welding process, quantifies node anomalies, cluster characteristics and weld segment risks, establishes a precise early warning mechanism, and improves the accuracy of construction quality control.

[0056] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent.

Claims

1. A method for early warning of welding defects in curtain walls based on big data analysis, characterized in that, include: S1: Collect the welding arc amplitude signal and instantaneous energy signal of the welding power source of the welding robot arm at the construction site for early warning of welding defects in the curtain wall. Organize the amplitude and energy data in chronological order, compare the amplitude changes and energy deviations of adjacent nodes, mark the nodes with abrupt amplitude changes and record the energy status, call the node status to establish the amplitude-energy correspondence, and obtain the welding arc amplitude-energy status nodes. S2: Monitor the changes in welding arc length during the welding process, identify continuous abnormal nodes and arrange their spatial positions, count the duration and fluctuation range of abnormalities, analyze the relationships and abnormal states of nodes within the cluster, call spatial distribution information to form abnormal cluster labels, and obtain information on abnormal arc length node clusters. S3: Based on the welding arc amplitude energy state node and arc length abnormal node cluster information, the welding current signal is collected, the instantaneous deviation of each node is calculated and the difference before and after is compared, the number of abnormal fluctuations is counted, the relationship between key nodes and neighboring nodes is analyzed, the amplitude energy and length state are called to mark the location of abnormal nodes, and the distribution map of abnormal welding current key nodes is obtained. S4: Combining the distribution map of key nodes with abnormal welding current, cluster information of nodes with abnormal arc length, and nodes with energy status of welding arc amplitude, perform joint anomaly analysis on the weld segment, compare the node anomaly concentration area with the status of adjacent nodes, call the spatial distribution and anomaly concentration situation to form a segment risk mapping, and obtain the defect risk distribution map of the weld segment.

2. The method for early warning of welding defects in curtain walls based on big data analysis according to claim 1, characterized in that, The welding arc amplitude energy state nodes include amplitude anomaly points, energy anomaly points, and node state mappings. The arc length anomaly node cluster information includes anomaly intervals, anomaly amplitudes, and cluster spatial relationships. The welding current anomaly key node distribution map includes current deviation points, continuous anomaly points, and anomaly concentration areas. The weld segment defect risk distribution map includes anomaly dense areas, adjacent related areas, and potential defect segments.

3. The method for early warning of welding defects in curtain walls based on big data analysis according to claim 1, characterized in that, The steps for obtaining S1 include: S101: Collect the arc amplitude signal and instantaneous energy signal of the welding robot arm at the construction site for early warning of welding defects in curtain walls. Align the sampling time according to the time sequence of the weld trajectory, extract the amplitude value, energy value, trajectory position and sampling number, verify the continuous status of the sampling number, remove the sampling points corresponding to the missing sampling time, and generate the weld trajectory amplitude energy sequence. S102: Based on the weld trajectory amplitude energy sequence, calculate the amplitude difference between adjacent sampling points, compare the absolute value of the amplitude difference with the preset amplitude change limit, determine the interval by comparing the energy value with the preset energy reference interval, mark the sampling points that exceed the preset amplitude change limit as amplitude change nodes, associate the amplitude change nodes, trajectory position, sampling time, energy up state, energy stable state, and energy down state, and establish a node amplitude energy association table; S103: Call the node amplitude energy association table, read the amplitude direction, energy state, and trajectory position interval between adjacent amplitude change nodes segment by segment, compare the consistency between the amplitude direction change state and the energy deviation direction, determine the matching relationship between the number of energy state continuation points and the trajectory position interval, mark the node segment with the amplitude direction reversed and the energy deviation direction continued in the same direction as the node abnormality, associate the abnormality mark with the node number, and obtain the welding arc amplitude energy state node.

4. The method for early warning of welding defects in curtain walls based on big data analysis according to claim 1, characterized in that, The steps for obtaining S2 include: S201: Monitor the change in arc length during the welding process, extract the sampling time, weld space position, and arc length value according to continuous sampling points, calculate the difference between the arc length value and the preset arc length benchmark value, compare the absolute value of the difference with the preset length deviation limit, mark the sampling points that exceed the preset length deviation limit as length deviation nodes, and obtain the length deviation node sequence. S202: Based on the length deviation node sequence, the difference in sampling time between adjacent length deviation nodes and the spatial position interval of the weld are statistically analyzed. Nodes with continuous sampling time and adjacent spatial positions are merged. The duration and length fluctuation amplitude of each segment are calculated. The duration is compared with the preset duration limit, and the length fluctuation amplitude is compared with the preset fluctuation amplitude limit. The fluctuation interval is marked, and a table of abnormal length fluctuation intervals is generated. S203: Call the length abnormal fluctuation interval table, arrange abnormal nodes according to the spatial location of the weld, compare the length change direction and length difference of adjacent nodes within the same interval, associate adjacent fluctuation intervals according to spatial location intervals, compare the duration, number of nodes, and length fluctuation amplitude of each interval, filter the area where the duration exceeds the preset duration limit and the number of nodes reaches the preset number of nodes, establish cluster association, and obtain arc length abnormal node cluster information.

5. The method for early warning of welding defects in curtain walls based on big data analysis according to claim 1, characterized in that, The steps for obtaining S3 include: S301: Collect welding process current signal, match weld space position according to sampling time, extract current value, sampling sequence number, adjacent previous node, adjacent next node, calculate the difference between current value and preset current reference value, compare the absolute value of difference with preset current deviation limit, compare the change direction of adjacent previous node difference with adjacent next node difference, and obtain welding current deviation node sequence. S302: Based on the welding current deviation node sequence, count the number of sampling points that continuously exceed the preset current deviation limit, compare the number of sampling points with the preset fluctuation limit, extract the spatial position interval of adjacent abnormal nodes, determine the state that the deviation direction of key nodes is consistent with that of adjacent nodes, mark the position of abnormal nodes, and generate a table of current abnormality concentrated areas. S303: Call the current anomaly concentration area table, associate the welding arc amplitude energy state node with the arc length anomaly node cluster information, map the current anomaly concentration area, amplitude energy state node, and length anomaly node cluster according to the weld spatial location, compare the spatial overlap relationship and anomaly persistence state of the three types of nodes, mark the distribution location of key nodes, and obtain the welding current anomaly key node distribution map.

6. The method for early warning of welding defects in curtain walls based on big data analysis according to claim 1, characterized in that, The steps for obtaining S4 include: S401: Obtain weld segment division parameters, call the distribution map of key nodes for abnormal welding current, cluster information of abnormal arc length nodes and welding arc amplitude energy status nodes, map amplitude nodes, energy nodes, length nodes and current nodes to the corresponding weld segments according to the spatial location of the weld, count the number of nodes and node spacing of each weld segment, and establish a weld segment abnormal node association table. S402: Based on the abnormal node association table of the weld segment, calculate the ratio of the number of amplitude nodes, energy nodes, length nodes, and current nodes to the segment length of each weld segment, compare the ratio with the preset node density limit, compare the node spacing of adjacent weld segments with the preset adjacent spacing limit, determine the abnormal node density area and the association status of adjacent nodes, and generate a potential defect range table for the weld segment. S403: Based on the potential defect range table of the weld segment, associate the spatial location of the weld, the concentration level of abnormal nodes, and the association status of adjacent nodes, match the risk level of each weld segment according to the preset risk classification rules, establish a key-value mapping between the weld segment number and the risk level, arrange the key-value mapping according to the spatial location of the weld, and obtain the defect risk distribution map of the weld segment.

7. The method for early warning of welding defects in curtain walls based on big data analysis according to claim 1, characterized in that, It also includes S5: calling the defect risk distribution map of the weld segment, sorting and filtering the abnormal concentration areas of the weld trajectory, marking the location and risk level of key abnormal nodes of the weld segment, analyzing the risk concentration pattern of each segment and establishing early warning signs, and obtaining the curtain wall welding defect early warning sign table.

8. The method for early warning of welding defects in curtain walls based on big data analysis according to claim 7, characterized in that, The steps for obtaining S5 include: S501: Call the defect risk distribution map of the weld segment, extract the weld segment number, spatial start point, spatial end point, risk level, and number of abnormal nodes in the order of weld trajectory, calculate the difference in risk level and the difference in the number of abnormal nodes between adjacent weld segments, arrange the weld segments in descending order of risk level, use the preset risk level limit and preset node number limit to screen the abnormal concentrated area, and generate a weld trajectory abnormal sorting list. S502: Based on the weld trajectory anomaly sorting list, extract the spatial location and risk level of key anomaly nodes in the anomaly concentration area, compare the spatial interval between nodes with the preset node proximity limit, associate the node number, node location, and risk level within the same weld segment, establish a node risk mapping according to the weld segment number, and generate a key node risk labeling table. S503: Based on the aforementioned key node risk labeling table, count the number of consecutive risk levels and the number of clusters of key abnormal nodes for each weld segment, compare the number of consecutive risk levels with the preset continuous risk limit, compare the number of clusters with the preset cluster number limit, determine the risk concentration pattern for each segment, associate the weld segment number, key abnormal node location, risk level, and risk concentration pattern, establish early warning indicators, and obtain the curtain wall welding defect early warning indicator table.

9. A method for early warning of welding defects in curtain walls based on big data analysis according to claim 7, characterized in that, In S5, the curtain wall welding defect early warning identification table includes key abnormal nodes, risk levels, and early warning identification.

10. A curtain wall welding defect early warning system based on big data analysis, characterized in that, The system is used in the curtain wall welding defect early warning method based on big data analysis as described in any one of claims 1-9, and the system comprises: The welding amplitude acquisition module collects the welding arc amplitude signal and instantaneous energy signal of the welding power source of the welding robot arm at the construction site for early warning of welding defects in curtain walls. It organizes the amplitude and energy data in chronological order, analyzes the amplitude changes and energy deviations of adjacent nodes, marks the nodes with abrupt amplitude changes and records the energy state, establishes the correspondence between amplitude and energy nodes, and obtains the welding arc amplitude and energy state nodes. The welding length monitoring module monitors the changes in welding arc length during the welding process, identifies continuous length abnormal nodes and arranges their spatial positions, counts the duration and fluctuation range of abnormalities, analyzes the length relationship of nodes within the cluster, calls spatial distribution information to divide abnormal node clusters, and obtains information on arc length abnormal node clusters. The welding current analysis module, based on the welding arc amplitude energy state node and arc length abnormal node cluster information, collects welding current signals, calculates the instantaneous deviation of each node and compares the sampling difference before and after, counts the number of consecutive abnormal fluctuations, analyzes the abnormal relationship between key nodes and adjacent nodes, filters the abnormal concentrated area of ​​weld trajectory and marks the location of key nodes, calls amplitude energy and length state for abnormal comparison, and obtains the distribution map of abnormal welding current key nodes. The weld risk mapping module combines the distribution map of key nodes with abnormal welding current, cluster information of nodes with abnormal arc length, and nodes with energy status of welding arc amplitude to perform joint anomaly analysis on the weld segment. It compares the relationship between the node anomaly concentration area and adjacent nodes, calls the state of the abnormal node concentration to establish the defect risk mapping of the weld segment, and obtains the defect risk distribution map of the weld segment. The weld seam early warning and identification module calls the defect risk distribution map of the weld seam segment, sorts and filters abnormal areas of the weld seam trajectory, marks the high-risk weld seam segments and key abnormal node locations, establishes early warning identification, and obtains the curtain wall welding defect early warning identification table.

Images