A fusion method for weld defect phased array detection

Abstract

The application discloses a kind of weld defect phased array detection fusion methods, belong to nondestructive testing technical field.Specifically including the following steps: S1, by 3D laser scanner to the current weld is continuously scanned and obtains about the 3D point cloud data of weld, by industrial camera to the surface of weld is scanned and obtains about the surface image data of weld, the surface image data of weld 3D point cloud data is multimodal data fusion and obtains about the first fusion data model of weld, the first fusion data model is divided into multiple regions;S2, by the first fusion data model and weld parameter and weld historical record data are matched, according to the matching result determination current weld each region exists defect possibility planning scanning strategy.The automatic detection process reduces manual participation, especially suitable for high altitude, low temperature or dangerous working condition under weld detection, reduce personnel safety risk, make up the deficiency that traditional phased array detection lacks surface geometry measurement ability.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology, and in particular to a phased array detection fusion method for weld defects. Background Technology

[0002] Modern industrial equipment relies heavily on welded components. Welds are complex, containing interfaces between various media and heat-affected zones, making them susceptible to defects during manufacturing and service life. To ensure the safety of welded structures, non-destructive testing (NDT) is essential. Traditional weld inspection methods include manual visual inspection, traditional NDT, and some automated methods based on simple single sensors. Manual inspection relies on experienced professionals visually inspecting the weld surface or using simple tools, which is time-consuming, inefficient, susceptible to human error, and difficult to perform large-area, continuous, and quantitative inspections. Furthermore, manual inspection poses significant safety risks in high-altitude, low-temperature, or hazardous environments. While traditional phased array testing can detect internal defects, it often lacks the ability to accurately measure surface geometry. Simultaneously, a single data source struggles to effectively connect surface morphology and internal structure information, failing to provide a systematic assessment of the overall weld quality. Existing inspection systems are mostly independent devices, resulting in isolated data and a lack of effective cross-sensor and cross-data source fusion methods. Summary of the Invention

[0003] Purpose of the invention: The purpose of this invention is to provide a phased array detection and fusion method for weld defects; it can solve the problem that a single data source is difficult to effectively connect surface morphology and internal structure information.

[0004] Technical Solution: To solve the above-mentioned technical problems, according to one aspect of the present invention, more specifically, a phased array detection and fusion method for weld defects, specifically including the following steps:

[0005] S1. Continuously scan the current weld seam using a 3D laser scanner to obtain 3D point cloud data about the weld seam, scan the surface of the weld seam using an industrial camera to obtain surface image data about the weld seam, perform multimodal data fusion on the 3D point cloud data and surface image data of the weld seam to obtain a first fused data model about the weld seam, and divide the first fused data model into multiple regions.

[0006] S2. Match the first fusion data model and weld parameters with the historical weld data, and determine the possibility of defects in each area of ​​the current weld based on the matching results to plan the scanning strategy.

[0007] S3. According to the planned scanning strategy, multi-angle acoustic beam scanning of the weld interior is performed using an ultrasonic phased array probe to obtain tomographic image data of the weld.

[0008] S4. Spatial registration of 3D point cloud data, surface image data and tomographic image data in a unified coordinate system, multimodal data fusion to obtain a second fused data model of the weld, and generate weld defect detection results based on the second fused data model;

[0009] S5. After binding the generated weld defect detection results with the corresponding first fusion data model, store them in the weld history record.

[0010] S6. Compare the defect detection results of each area of ​​the weld with the planned scanning strategy to obtain relevant data of the scanning strategy. Analyze and process the relevant data of the scanning strategy to obtain the scanning evaluation index of the scanning strategy. Determine whether to continue using the scanning strategy based on the evaluation index of the scanning strategy.

[0011] Furthermore, in step S1, the division conditions for dividing the first fused data model into multiple regions are actually set as needed.

[0012] Furthermore, the specific operation steps of step S2 are as follows:

[0013] S21. Match the current weld seam's 3D point cloud data and weld seam parameters with the weld seam's historical data, and select the historical weld seam data that matches the current weld seam the most.

[0014] S22. Based on the defect situation of each area of ​​the historical weld with the highest matching degree, each area of ​​the weld is determined as a high-risk area of ​​weld defects and a low-risk area of ​​weld defects.

[0015] S23. Plan denser scanning paths for areas identified as high-risk weld defects and sparser scanning paths for areas identified as low-risk weld defects.

[0016] Furthermore, in step S23, the number of scans for denser and sparser scan paths is adjusted according to actual needs during operation.

[0017] Furthermore, in step S21, the matching degree is determined based on 3D point cloud data and weld parameters, including the similarity of various parameters such as weld workpiece attributes, process parameters, and geometric features. The higher the similarity, the higher the matching degree.

[0018] Furthermore, in step S4, the weld defect detection result is a weld defect model constructed based on the parameters of the weld defect and a display of the location of the weld defect on the weld based on the second fusion data model.

[0019] Furthermore, the specific operation steps of step S6 are as follows:

[0020] S61. Compare the defect detection results of each area of ​​the weld with the planned scan to obtain the number of areas with weld defects in the planned denser scan path, the number of areas without weld defects in the planned denser scan path, the number of areas without weld defects in the planned sparser scan path, and the number of areas with weld defects in the planned sparser scan path.

[0021] S62. By analyzing the number of regions with weld defects in the planned denser scan path, the number of regions without weld defects in the planned denser scan path, the number of regions without weld defects in the planned sparser scan path, and the number of regions with weld defects in the planned sparser scan path, the scanning evaluation index of the current scanning strategy is obtained.

[0022] S63. Compare the scan evaluation index with the scan evaluation index threshold. If the scan evaluation index is greater than the scan evaluation index threshold, continue to use the scan strategy; otherwise, do not continue to use the scan strategy.

[0023] S64. After determining that the scanning strategy should not be used anymore, analyze the historical data of the scanning strategy used, adjust the scanning strategy, and replace the previous strategy with the new scanning strategy.

[0024] Furthermore, in step S62, by analyzing the number of regions with weld defects when using the current scanning strategy to plan a denser scan path each time, the number of regions without weld defects when using the current scanning strategy to plan a denser scan path each time, the number of regions without weld defects when using the current scanning strategy to plan a sparser scan path each time, and the number of regions with weld defects when using the current scanning strategy to plan a sparser scan path each time, the scanning evaluation index of the current scanning strategy is obtained as follows:

[0025]

[0026] in, This is the scanning evaluation index for the current scanning strategy. To plan the number of regions with weld defects for each time a denser scan path is used with the current scan strategy, To plan the number of areas without weld defects for each time the current scanning strategy is used, a denser scan path is planned. To plan the number of regions without weld defects for each time a sparser scan path is used with the current scan strategy. The number of regions with weld defects is calculated for each time the current scanning strategy is used to plan a sparser scan path, where n is the total number of times the current scanning strategy is used.

[0027] Furthermore, in step S64, when modifying the scanning strategy, if the proportion of times a denser scanning path in a certain area does not contain weld defects exceeds a set first upper limit, the area is adjusted to have a sparser scanning path planned; otherwise, no adjustment is made. If the proportion of times a sparser scanning path in a certain area contains weld defects exceeds a set second upper limit, the area is adjusted to have a denser scanning path planned; otherwise, no adjustment is made.

[0028] Beneficial effects:

[0029] 1. Integrating multi-source data to construct a complete detection model: 3D point cloud data of the weld is acquired through a 3D laser scanner, surface image data is acquired through an industrial camera, and internal tomographic image data is acquired through an ultrasonic phased array probe. This achieves cross-sensor fusion of surface morphology and internal structural information. Through unified coordinate system spatial registration, a second fused data model containing surface and internal defects is generated. This model can accurately construct weld defect models and locate defect positions, solving the "data silo" problem of traditional single data sources and forming a systematic judgment on weld quality. The first stage fuses 3D point cloud and surface images to generate a preliminary model for dividing regions and matching historical data. The second stage incorporates tomographic images to generate the final detection result, ensuring a deep correlation between surface and internal defect information.

[0030] 2. The intelligent scanning strategy based on historical data improves detection efficiency and targeting. By matching the current weld with historical data, the weld is divided into high-risk and low-risk areas, avoiding the blindness of the traditional "uniform scanning of the entire area". This shortens the detection time and reduces resource consumption. The scanning path density can be dynamically adjusted according to actual needs. For example, the number of scans can be increased in high-risk areas to improve the defect detection rate. By using the matching results of historical weld data, potential defect areas can be quickly located, reducing trial scans of unknown areas. This is especially suitable for detection scenarios of batch production or similar welds, improving the automation and intelligence level of the detection process.

[0031] 3. Adaptive optimization of scanning strategies enhances the system's self-learning capability. By defining a scanning evaluation index, the accuracy and efficiency of the scanning strategy are quantitatively evaluated based on data such as "the number of areas where defects are detected by dense scanning" and "the number of areas where defects are missed by sparse scanning." The formula uses logistic regression to ensure a linear correlation between the evaluation results and actual detection effects. A threshold is set to compare the evaluation index, automatically determining whether to continue using or adjust the strategy, avoiding reliance on human experience and improving the adaptability of the detection system. If the scanning strategy evaluation fails to meet the standards, the system automatically analyzes historical data and adjusts the regional risk level, forming a closed loop of "detection-evaluation-optimization" to continuously improve the accuracy of the detection strategy.

[0032] 4. Bind the defect detection results to the first fusion data model and store them in the history record to form a complete traceability chain including "surface morphology - internal defects - scanning strategy", which facilitates subsequent quality review, process improvement or responsibility tracing.

[0033] 5. Automated inspection processes reduce manual intervention, making them particularly suitable for weld inspection in high-altitude, low-temperature, or hazardous conditions. This reduces personnel safety risks and compensates for the lack of surface geometry measurement capabilities in traditional phased array inspection. Through multimodal data fusion, it enables the linkage analysis of "surface-internal" defects, avoiding the blind spots of single-technology inspection. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating the method. Detailed Implementation

[0035] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] Example 1

[0037] The first step involves continuously scanning the current weld seam using a 3D laser scanner to obtain 3D point cloud data of the weld seam, and scanning the weld seam surface using an industrial camera to obtain surface image data of the weld seam. The 3D point cloud data and surface image data of the weld seam are then fused using multimodal data to obtain a first fused data model of the weld seam. The first fused data model is then divided into multiple regions. The division conditions for the 3D point cloud data of the weld seam into multiple regions are set according to actual needs.

[0038] The second step involves matching the first fused data model and weld parameters with historical weld data. Based on the matching results, the likelihood of defects in each area of ​​the current weld is determined, and a scanning strategy is planned. The specific steps are as follows:

[0039] 1. Match the current weld seam's 3D point cloud data and weld seam parameters with historical weld seam data, including: weld seam workpiece attributes, process parameters, geometric features, etc. The higher the similarity, the higher the matching degree. Select the historical weld seam data with the highest matching degree with the current weld seam.

[0040] 2. Based on the defect situation of each area of ​​the historical weld with the highest matching degree, each area of ​​the weld is identified as a high-risk area for weld defects and a low-risk area for weld defects.

[0041] 3. For areas identified as high-risk weld defects, plan denser scanning paths; for areas identified as low-risk weld defects, plan sparser scanning paths. The number of scans for denser and sparser scanning paths can be adjusted during operation according to actual needs.

[0042] By integrating multi-source data, a complete detection model is constructed. A 3D laser scanner acquires 3D point cloud data of the weld, an industrial camera acquires surface image data, and an ultrasonic phased array probe acquires internal tomographic image data. This achieves cross-sensor fusion of surface morphology and internal structural information. Through unified coordinate system spatial registration, a second fused data model containing surface and internal defects is generated. This model can accurately construct weld defect models and locate defect positions, solving the "data silo" problem of traditional single data sources and forming a systematic judgment on weld quality. The first stage fuses 3D point cloud and surface images to generate a preliminary model for region division and historical data matching. The second stage incorporates tomographic images to generate the final detection result, ensuring a deep correlation between surface and internal defect information.

[0043] The third step involves using a phased-array ultrasonic probe to perform multi-angle acoustic beam scanning of the weld interior, based on the planned scanning strategy, to obtain tomographic image data of the weld. An intelligent scanning strategy based on historical data improves detection efficiency and targeting. By matching the current weld with historical data, the weld is divided into high-risk and low-risk areas, avoiding the blindness of traditional "uniform scanning of the entire area," shortening detection time and reducing resource consumption. The scanning path density can be dynamically adjusted according to actual needs; for example, increasing the number of scans in high-risk areas improves the defect detection rate. Using historical weld data matching results, potential defect areas can be quickly located, reducing trial scans of unknown areas. This is particularly suitable for batch production or similar weld inspection scenarios, improving the automation and intelligence level of the inspection process.

[0044] The fourth step involves spatially registering the 3D point cloud data, surface image data, and tomographic image data in a unified coordinate system, performing multimodal data fusion to obtain a second fused data model of the weld, and generating weld defect detection results based on the second fused data model. The weld defect detection results consist of a weld defect model constructed based on the parameters of the weld defect and a display of the location of the weld defect on the weld based on the second fused data model.

[0045] The fifth step is to bind the generated weld defect detection results to the corresponding first fusion data model and store them in the weld history record. This binds the defect detection results to the first fusion data model and stores them in the history record, forming a complete traceability chain that includes "surface morphology - internal defects - scanning strategy", which facilitates subsequent quality review, process improvement or responsibility tracing.

[0046] The sixth step involves comparing the defect detection results of each area of ​​the weld with the planned scanning strategy to obtain relevant data for the scanning strategy. This data is then analyzed to obtain the scanning evaluation index. Based on this index, it is determined whether to continue using the scanning strategy. The specific steps are as follows:

[0047] 1. Compare the defect detection results of each area of ​​the weld with the planned scan to obtain the number of areas with weld defects in the planned denser scan path, the number of areas without weld defects in the planned denser scan path, the number of areas without weld defects in the planned sparser scan path, and the number of areas with weld defects in the planned sparser scan path.

[0048] 2. By analyzing the number of regions with weld defects in the denser scan path, the number of regions without weld defects in the denser scan path, the number of regions without weld defects in the sparser scan path, and the number of regions with weld defects in the sparser scan path, the scanning evaluation index of the current scanning strategy is obtained.

[0049] 3. Compare the scan evaluation index with the scan evaluation index threshold. If the scan evaluation index is greater than the scan evaluation index threshold, continue to use the scan strategy; otherwise, do not continue to use the scan strategy.

[0050] 4. After determining that the scanning strategy should no longer be used, analyze the historical data of the scanning strategy used, adjust the scanning strategy, and replace the previous strategy with the new scanning strategy.

[0051] By analyzing the number of regions with weld defects when using the current scanning strategy to plan a denser scan path, the number of regions without weld defects when using the current scanning strategy to plan a denser scan path, the number of regions without weld defects when using the current scanning strategy to plan a sparser scan path, and the number of regions with weld defects when using the current scanning strategy to plan a sparser scan path, the scanning evaluation index of the current scanning strategy is obtained:

[0052]

[0053] in, This is the scanning evaluation index for the current scanning strategy. To plan the number of regions with weld defects for each time a denser scan path is used with the current scan strategy, To plan the number of areas without weld defects for each time the current scanning strategy is used, a denser scan path is planned. To plan the number of regions without weld defects for each time a sparser scan path is used with the current scan strategy. The number of regions with weld defects is calculated for each time the current scanning strategy is used to plan a sparser scan path, where n is the total number of times the current scanning strategy is used. Indicates the number of correct detections: the number of defects found in high-risk areas. Indicates correct detection: the number of low-risk areas that are free of defects and confirmed; as both increase, The value will also increase, indicating that the current scanning strategy is more suitable; This indicates a potential overscan: the number of defect-free areas in high-risk regions. This indicates a potential missed detection: the number of defects in low-risk areas that were overlooked. As both increase, A decreasing value indicates that the current scanning strategy is less applicable. The formula structure effectively handles extreme values. For cases where the result is 0 or 1, the output is normalized to reduce the impact of data fluctuations.

[0054] The scanning strategy is adaptively optimized to enhance the system's self-learning capability. By defining a scanning evaluation index, the accuracy and efficiency of the scanning strategy are quantitatively evaluated based on data such as "the number of areas where defects are detected by dense scanning" and "the number of areas where defects are missed by sparse scanning." The formula uses logistic regression to ensure a linear correlation between the evaluation results and actual detection performance. A threshold is set to compare the evaluation index, automatically determining whether to continue using or adjust the strategy, avoiding reliance on human experience and improving the adaptability of the detection system. If the scanning strategy evaluation fails to meet the standards, the system automatically analyzes historical data and adjusts the regional risk level, forming a closed loop of "detection-evaluation-optimization" to continuously improve the accuracy of the detection strategy.

[0055] When changing the scanning strategy, if the proportion of times a denser scanning path in a certain area does not contain weld defects exceeds the total number of times a denser scanning path in that area exceeds a set first upper limit, the area will be adjusted to have a sparser scanning path planned; otherwise, no adjustment will be made. If the proportion of times a sparser scanning path in a certain area contains weld defects exceeds the total number of times a sparser scanning path in that area exceeds a set second upper limit, the area will be adjusted to have a denser scanning path planned; otherwise, no adjustment will be made.

[0056] Example 2

[0057] When calculating the scan evaluation index, when =15, When =30, then we have

[0058]

[0059] At this point, 0.622 is compared with the threshold of the scan evaluation index. If the scan evaluation index is greater than the threshold, the scan strategy is continued; otherwise, the scan strategy is discontinued.

[0060] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A phased array detection fusion method for weld defects, characterized in that, Specifically, the following steps are included: S1. Continuously scan the current weld seam using a 3D laser scanner to obtain 3D point cloud data about the weld seam, scan the surface of the weld seam using an industrial camera to obtain surface image data about the weld seam, perform multimodal data fusion on the 3D point cloud data and surface image data of the weld seam to obtain a first fused data model about the weld seam, and divide the first fused data model into multiple regions. S2. Match the first fusion data model and weld parameters with the historical weld data, and determine the possibility of defects in each area of ​​the current weld based on the matching results to plan the scanning strategy. S3. According to the planned scanning strategy, multi-angle acoustic beam scanning of the weld interior is performed using an ultrasonic phased array probe to obtain tomographic image data of the weld. S4. Spatial registration of 3D point cloud data, surface image data and tomographic image data in a unified coordinate system, multimodal data fusion to obtain a second fused data model of the weld, and generate weld defect detection results based on the second fused data model; S5. After binding the generated weld defect detection results with the corresponding first fusion data model, store them in the weld history record. S6. Compare the defect detection results of each area of ​​the weld with the planned scanning strategy to obtain the relevant data of the scanning strategy. Analyze and process the relevant data of the scanning strategy to obtain the scanning evaluation index of the scanning strategy. Determine whether to continue using the scanning strategy based on the evaluation index of the scanning strategy. The specific steps of step S6 are as follows: S61. Compare the defect detection results of each area of ​​the weld with the planned scan to obtain the number of areas with weld defects in the planned denser scan path, the number of areas without weld defects in the planned denser scan path, the number of areas without weld defects in the planned sparser scan path, and the number of areas with weld defects in the planned sparser scan path. S62. By analyzing the number of regions with weld defects in the planned denser scan path, the number of regions without weld defects in the planned denser scan path, the number of regions without weld defects in the planned sparser scan path, and the number of regions with weld defects in the planned sparser scan path, the scanning evaluation index of the current scanning strategy is obtained. S63. Compare the scan evaluation index with the scan evaluation index threshold. If the scan evaluation index is greater than the scan evaluation index threshold, continue to use the scan strategy; otherwise, do not continue to use the scan strategy. S64. After determining that the scanning strategy should not be used anymore, analyze the historical data of the scanning strategy used, adjust the scanning strategy, and replace the previous strategy with the new scanning strategy. In step S62, the scanning evaluation index of the current scanning strategy is obtained by analyzing the number of regions with weld defects when planning a denser scanning path using the current scanning strategy each time, the number of regions without weld defects when planning a denser scanning path using the current scanning strategy each time, the number of regions without weld defects when planning a sparser scanning path using the current scanning strategy each time, and the number of regions with weld defects when planning a sparser scanning path using the current scanning strategy each time. ； in, This is the scanning evaluation index for the current scanning strategy. To plan the number of regions with weld defects for each time a denser scan path is used with the current scan strategy, To plan the number of areas without weld defects for each time the current scanning strategy is used, a denser scan path is planned. To plan the number of regions without weld defects for each time a sparser scan path is used with the current scan strategy. The number of regions with weld defects is calculated for each time the current scanning strategy is used to plan a sparser scan path, where n is the total number of times the current scanning strategy is used.

2. The phased array detection and fusion method for weld defects according to claim 1, characterized in that: In step S1, the division conditions for dividing the first fused data model into multiple regions are set according to actual needs.

3. The phased array detection and fusion method for weld defects according to claim 1, characterized in that: The specific steps of step S2 are as follows: S21. Match the current weld seam's 3D point cloud data and weld seam parameters with the weld seam's historical data, and select the historical weld seam data that matches the current weld seam the most. S22. Based on the defect situation of each area of ​​the historical weld with the highest matching degree, each area of ​​the weld is determined as a high-risk area of ​​weld defects and a low-risk area of ​​weld defects. S23. Plan denser scanning paths for areas identified as high-risk weld defects and sparser scanning paths for areas identified as low-risk weld defects.

4. The phased array detection and fusion method for weld defects according to claim 3, characterized in that: In step S23, the number of scans for denser and sparser scan paths is adjusted according to actual needs during operation.

5. The phased array detection fusion method for weld defects according to claim 3, characterized in that: In step S21, the matching degree is determined based on 3D point cloud data and weld parameters, including the similarity of various parameters such as weld workpiece attributes, process parameters, and geometric features. The higher the similarity, the higher the matching degree.

6. The phased array detection fusion method for weld defects according to claim 1, characterized in that: In step S4, the weld defect detection result is a weld defect model constructed based on the parameters of the weld defect and the location of the weld defect on the weld shown based on the second fusion data model.

7. The phased array detection fusion method for weld defects according to claim 1, characterized in that: In step S64, when the scanning strategy is modified, if the proportion of the number of times a denser scanning path in a certain area does not contain weld defects exceeds the total number of times a denser scanning path is planned in that area, the area is adjusted to have a sparser scanning path planned; otherwise, no adjustment is made. If the proportion of the number of times a sparser scanning path in a certain area contains weld defects exceeds the total number of times a sparser scanning path is planned in that area, the area is adjusted to have a denser scanning path planned; otherwise, no adjustment is made.

Images