Welding seam defect detection method and equipment based on multi-data fusion and medium

Abstract

The invention discloses a weld defect detection method and device based on multi-data fusion and a medium, and relates to the technical field of data detection.A detection path of a detection robot is generated based on weld data, the detection path of the detection robot is optimized by adopting a dynamic programming algorithm and curvature fitting, and the weld defect is detected based on the optimized detection path. The method comprises the following steps: acquiring target weld joint detection data, preprocessing and fusing to obtain weld joint internal structure feature and surface geometric feature data, and constructing a three-dimensional defect projection model based on the weld joint internal structure feature and surface geometric feature data to identify weld joint defects. A moving path of a detection robot is planned according to a welding seam, so that surface morphology and internal defect information of the welding seam are synchronously acquired, data fusion and defect positioning are performed based on acquired data, three-dimensional imaging is obtained, defect detection of the welding seam is performed based on the three-dimensional imaging, the defect identification precision can be improved, and the reliability and consistency of a detection result are improved. And a more visual detection result is presented.

Description

Technical Field

[0001] The present invention relates to the technical field of data detection, and in particular to a weld defect detection method, equipment and medium based on multi-data fusion. Background Art

[0002] With the rapid development of industrialization, large storage tanks, as important industrial equipment, are widely used in the fields of petroleum, chemical industry, storage and transportation. As an important connection part of the storage tank, the quality of the weld directly affects the safety and service life of the tank. The traditional method of weld defect detection is mainly based on handheld ultrasonic flaw detectors, but this method relies on the operating experience of the inspectors, which not only has low detection efficiency but also limited accuracy. In addition, although some detection methods based on automated equipment have been introduced in recent years, such as magnetic particle testing and eddy current testing, these technologies still have many problems in the detection of large storage tank welds.

[0003] Disadvantages or shortcomings of the background technology: 1. Low detection efficiency: Handheld devices have slow detection speed and are prone to missed detection or repeated detection. Especially when the length of the tank weld is long, the manual operation efficiency is significantly reduced; 2. Limited detection accuracy: Existing technologies are not effective in dealing with complex weld shapes (such as irregular welds or rough surfaces) and are not able to identify small internal defects; 3. Lack of automation and intelligence: Most devices still require manual analysis of test signals, and the accuracy and consistency of test results are easily affected by subjective factors of testers; 4. The results are not intuitive: Existing detection methods mainly present defects in numerical or simple graphical form, lacking three-dimensional visualization effects, which is not conducive to a comprehensive assessment of defect location and morphology. Summary of the invention

[0004] The technical problem to be solved by the present invention is the low efficiency, insufficient accuracy, low intelligence level and unintuitive result presentation in the existing large-scale storage tank weld defect detection. The purpose is to provide a weld defect detection method, equipment and medium based on multi-data fusion, and plan the moving path of the detection robot according to the weld. The surface morphology and internal defect information of the weld are synchronously collected based on the moving path of the robot, and data fusion and defect location are performed based on the collected data to generate a three-dimensional imaging result. The weld defect detection is performed based on the three-dimensional imaging result, which can improve the defect recognition accuracy, improve the reliability and consistency of the detection results, and present a more intuitive detection result.

[0005] The present invention is achieved through the following technical solutions: The first aspect of the present invention provides a weld defect detection method based on multi-data fusion, comprising the following specific steps: Acquire the target weld data for inspection, and generate the inspection path of the inspection robot based on the weld data; Dynamic programming algorithm and curvature fitting are used to optimize the inspection path of the inspection robot; Based on the optimized detection path, the target weld detection data is obtained; Preprocess the weld inspection data, and fuse the preprocessed data to obtain the weld internal structure characteristics and surface geometric characteristics data; Construct a three-dimensional defect projection model based on the internal structural characteristics and surface geometric characteristics of the weld; Identify weld defects based on the 3D defect projection model and generate detection results.

[0006] Furthermore, the use of dynamic programming algorithm and curvature fitting to optimize the detection path of the detection robot specifically includes: Obtain the joint angle, joint velocity, joint acceleration and joint torque of the inspection robot when it moves along the inspection path, and construct an optimization objective function; Obtain the constraints of robot motion; Based on the constraints of robot motion, the optimization objective function is solved to obtain the optimal joint angle, optimal joint velocity, optimal joint acceleration and optimal joint torque.

[0007] Furthermore, the construction of the optimization objective function specifically includes: The weights of the joint angles, joint velocities, joint accelerations, and joint torques of the detection robot are obtained when the detection robot moves along the detection path. The optimization objective function is constructed with the goal of minimizing the time the detection robot spends moving along the detection path. The calculation steps include: ; in, Q represents the optimization objective; T Indicates the total time of exercise; q represents the joint angle; represents the joint velocity; represents the joint acceleration; represents the joint jerk; τ represents the joint torque; w 1 , w 2 , w 3 , w 4 Represents the weight coefficient.

[0008] Further, the obtaining of target weld detection data specifically includes: An embedded controller is constructed to collect target weld detection data, the collected data including ultrasonic echo data, laser ranging data and weld image data, wherein the embedded controller includes FPGA and ARM processor, the FPGA is responsible for high-speed logic processing and data acquisition, and the ARM processor is responsible for advanced control and data processing.

[0009] Furthermore, the preprocessing of the weld detection data specifically includes: Preprocessing the ultrasonic echo data based on wavelet transform and time-frequency analysis includes: performing wavelet decomposition on the ultrasonic echo data to obtain wavelet transform coefficients; using modulation factors to perform time-frequency reconstruction on the wavelet transform coefficients; dynamically optimizing the reconstructed video based on adaptive time-frequency analysis to obtain optimized ultrasonic echo signals; Extracting ultrasonic signal defect features based on the optimized ultrasonic echo signal; Curvature analysis is performed on laser ranging data and image data to extract weld geometric features.

[0010] Furthermore, the calculation steps of preprocessing the ultrasonic echo data based on wavelet transform and time-frequency analysis include: ; in, represents the optimized time-frequency representation signal; Indicates j Layer wavelet transform coefficients; represents the modulation factor, which is used for time-frequency reconstruction; G ( f ) represents an adaptive filter function; J Indicates the number of wavelet decomposition layers; s Represents the signal scale parameter.

[0011] Furthermore, the fusion processing of the pre-processed data specifically includes: Dynamically adjust the weight according to the confidence of sensor data to obtain the weight under the set threshold; Based on the weight under the set threshold, the improved Kalman filter is used for state prediction and error correction, and the pre-processed data is time-synchronized to obtain time-synchronized data; Through the geometric transformation constraints between sensors, the preprocessed data are spatially aligned to obtain spatially aligned data.

[0012] Furthermore, the calculation step of dynamically adjusting the weight according to the sensor data confidence level includes: ; ; ; in, X t Represents the target state vector at time t, including the fused position information and echo features; X t-1 represents the target state vector at time t-1; F represents the state transfer matrix; G represents the control input matrix; U t Represents external control input; W t represents process noise; H represents the observation matrix; Z t represents the observation vector; V t represents the measurement noise; w i represents the Bayesian adaptive weight, which is dynamically adjusted according to the sensor signal-to-noise ratio; represents the target state vector at time t The estimated value of the state; Indicates i The target state vector at time t The estimated value of the state.

[0013] A second aspect of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a weld defect detection method based on multi-data fusion is implemented.

[0014] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a weld defect detection method based on multi-data fusion.

[0015] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1. Significantly improved inspection efficiency: The automated inspection process reduces manual intervention and can quickly complete the inspection of large-area welds; 2. High-precision detection capability: Improve the accuracy of defect recognition through multi-sensor fusion technology, especially in the detection of complex weld shapes and tiny defects; 3. Intelligent data analysis: built-in intelligent algorithm realizes automatic analysis of detection signals and improves the reliability and consistency of detection results; 4. 3D visualization: Generate 3D visualization images of defects to facilitate subsequent maintenance and quality assessment; 5. Strong applicability: The design scheme can adapt to various weld forms and working environments of storage tanks, and has broad industrial application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 This is a weld defect detection process in an embodiment of the present invention; Figure 2 Schematic diagram of three-dimensional imaging of defects in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0018] As a possible implementation, Figure 1 As shown, this embodiment provides a weld defect detection method based on multi-data fusion, including the following specific steps: acquiring detection target weld data, and generating a detection path of a detection robot based on the weld data; optimizing the detection path of the detection robot by using a dynamic programming algorithm and curvature fitting; acquiring target weld detection data based on the optimized detection path; preprocessing the weld detection data, and fusing the preprocessed data to obtain weld internal structure features and surface geometric feature data; constructing a three-dimensional defect projection model based on the weld internal structure features and surface geometric feature data; identifying weld defects based on the three-dimensional defect projection model, and generating detection results.

[0019] In some possible implementations, constructing an optimization objective function specifically includes: The weights of the joint angles, joint velocities, joint accelerations, and joint torques of the detection robot are obtained when the detection robot moves along the detection path. The optimization objective function is constructed with the goal of minimizing the time the detection robot spends moving along the detection path. The calculation steps include: ; in, Q represents the optimization objective; T Indicates the total time of exercise; q represents the joint angle; represents the joint velocity; represents the joint acceleration; represents the joint jerk; τ represents the joint torque; w1 , w 2 , w 3 , w 4 Represents the weight coefficient.

[0020] In some possible implementations, the constraints of the robot's motion include: ; in, , Respectively represent the minimum and maximum limits of joint speed; , Respectively represent the minimum and maximum limits of joint acceleration; , Represent the minimum and maximum limits of the joint torque respectively.

[0021] In some possible implementations, obtaining target weld inspection data specifically includes: An embedded controller is built to collect target weld detection data, including ultrasonic echo data, laser ranging data and weld image data. The embedded controller includes FPGA and ARM processors. FPGA is responsible for high-speed logic processing and data acquisition, and ARM processor is responsible for advanced control and data processing. That is, the embedded controller adopts an embedded controller architecture based on FPGA+ARM, which supports high-frequency acquisition and synchronous processing of multi-sensor data.

[0022] In some possible implementations, preprocessing the weld inspection data specifically includes: Preprocessing the ultrasonic echo data based on wavelet transform and time-frequency analysis includes: performing wavelet decomposition on the ultrasonic echo data to obtain wavelet transform coefficients; using modulation factors to perform time-frequency reconstruction on the wavelet transform coefficients; dynamically optimizing the reconstructed video based on adaptive time-frequency analysis to obtain optimized ultrasonic echo signals; Extracting ultrasonic signal defect features based on the optimized ultrasonic echo signal; Curvature analysis is performed on laser ranging data and image data to extract weld geometric features.

[0023] In some possible implementations, the ultrasonic signal defect characteristics include: echo delay, amplitude distribution and frequency distribution.

[0024] In some possible implementations, an optimization algorithm based on improved wavelet transform and adaptive time-frequency analysis is proposed for ultrasonic detection of weld defects to improve the accuracy and robustness of defect feature extraction. Through multi-scale decomposition and time-frequency joint analysis, high-precision processing of ultrasonic echo signals is achieved, thereby accurately extracting echo delay, amplitude and frequency characteristics. Specifically, the calculation steps for preprocessing ultrasonic echo data based on wavelet transform and time-frequency analysis include: ; in, represents the optimized time-frequency representation signal; Indicates j Layer wavelet transform coefficients; represents the modulation factor, which is used for time-frequency reconstruction; G ( f ) represents an adaptive filter function; J Indicates the number of wavelet decomposition layers; s Represents the signal scale parameter.

[0025] Through improved wavelet transform: introduce multi-scale signal enhancement method to improve the extraction capability of defect features, especially the identification of weak defects. Adaptive time-frequency analysis: use dynamic modulation strategy to optimize time-frequency distribution and improve the resolution and stability of frequency features. Multi-feature fusion: comprehensively analyze echo delay, amplitude, and frequency distribution, construct high-precision feature parameters, and achieve accurate classification of defects. Compared with the traditional time-frequency analysis method, the processing method adopted in this embodiment can improve the signal-to-noise ratio of the defect signal, enhance the resolution of tiny defects, reduce artifact interference, and provide more reliable technical support for weld defect detection.

[0026] In some possible implementations, this embodiment proposes an optimization method based on improved Kalman filtering and Bayesian adaptive weighted fusion for the multimodal fusion problem of ultrasonic echo, laser ranging data and image information. This method achieves precise alignment of data in time and space through state estimation and dynamic weight adjustment, thereby improving the accuracy and robustness of defect detection. Specifically, it includes: fusing the preprocessed data, dynamically adjusting the weights according to the confidence of the sensor data, and obtaining the weights under the set threshold; based on the weights under the set threshold, using improved Kalman filtering for state prediction and error correction, and performing time synchronization processing on the preprocessed data to obtain time-synchronized data; through geometric transformation constraints between sensors, the preprocessed data is spatially aligned to obtain spatially aligned data.

[0027] In some possible implementations, adaptive weighted fusion: combining Bayesian optimization strategy, dynamically adjusting weights according to sensor data confidence, reducing interference from low-confidence data. Time alignment optimization: using improved Kalman filter (EKF) for state prediction and error correction to achieve high-precision time synchronization. Spatial correction enhancement: based on geometric transformation constraints between sensors, the alignment accuracy of multimodal data in the spatial coordinate system can be improved. This optimization method can improve the real-time and accuracy of fused data, enhance robustness under complex working conditions, and improve the reliability of defect detection.

[0028] In some possible implementations, the calculation step of dynamically adjusting the weight according to the confidence of the sensor data includes: ; ; ; in, X t Represents the target state vector at time t, including the fused position information and echo features; X t-1 represents the target state vector at time t-1; F represents the state transfer matrix; G represents the control input matrix; U t Represents external control input; W t represents process noise; H represents the observation matrix; Z t represents the observation vector; V t represents the measurement noise; w i represents the Bayesian adaptive weight, which is dynamically adjusted according to the sensor signal-to-noise ratio; represents the target state vector at time t The estimated value of the state; Indicates i The target state vector at time t The estimated value of the state.

[0029] In some possible implementations, after constructing a 3D defect projection model based on the internal structural features and surface geometric feature data of the weld, a 3D visualization model of the weld is generated using point cloud reconstruction technology, and the defect location is displayed in color coding. In addition, an interactive viewing function of the 3D model is provided, which supports users to rotate, zoom and mark defect areas in real time.

[0030] In some possible implementations, such as Figure 2As shown in the figure, weld defects are identified based on the 3D defect projection model, including: using convolutional neural networks to classify weld defect features, supporting accurate identification of appearance defects (such as undercuts and weld bumps) and internal defects (such as slag inclusions and cracks). Integrate a Transformer-based defect diagnosis model to analyze the correlation between defect features and predict the cause and development trend of defects. Automatically generate defect diagnosis reports, including defect type, location, size and severity, and provide maintenance recommendations.

[0031] As a possible implementation, this embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, a weld defect detection method based on multi-data fusion is implemented.

[0032] As a possible implementation, this embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, a weld defect detection method based on multi-data fusion is implemented.

[0033] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A weld defect detection method based on multi-data fusion, characterized in that: The specific steps include: Acquire the target weld data for inspection, and generate the inspection path of the inspection robot based on the weld data; Dynamic programming algorithm and curvature fitting are used to optimize the inspection path of the inspection robot; Based on the optimized detection path, the target weld detection data is obtained; Preprocess the weld inspection data, and fuse the preprocessed data to obtain the weld internal structure characteristics and surface geometric characteristics data; Construct a three-dimensional defect projection model based on the internal structural characteristics and surface geometric characteristics of the weld; Identify weld defects based on the 3D defect projection model and generate detection results.

2. The weld defect detection method based on multi-data fusion according to claim 1 is characterized in that: The method of optimizing the detection path of the detection robot by using a dynamic programming algorithm and curvature fitting specifically includes: Obtain the joint angle, joint velocity, joint acceleration and joint torque of the inspection robot when it moves along the inspection path, and construct an optimization objective function; Obtain the constraints of robot motion; Based on the constraints of robot motion, the optimization objective function is solved to obtain the optimal joint angle, optimal joint velocity, optimal joint acceleration and optimal joint torque.

3. The weld defect detection method based on multi-data fusion according to claim 1 is characterized in that: The constructing of the optimization objective function specifically includes: The weights of the joint angles, joint velocities, joint accelerations, and joint torques of the detection robot are obtained when the detection robot moves along the detection path. The optimization objective function is constructed with the goal of minimizing the time the detection robot spends moving along the detection path. The calculation steps include: ； in, Q represents the optimization objective; T Indicates the total time of exercise; q represents the joint angle; represents the joint velocity; represents the joint acceleration; represents the joint jerk; τ represents the joint torque; w 1. w 2. w 3. w 4 represents the weight coefficient.

4. The weld defect detection method based on multi-data fusion according to claim 1 is characterized in that: The obtaining of target weld detection data specifically includes: An embedded controller is constructed to collect target weld detection data, the collected data including ultrasonic echo data, laser ranging data and weld image data, wherein the embedded controller includes FPGA and ARM processor, the FPGA is responsible for high-speed logic processing and data acquisition, and the ARM processor is responsible for advanced control and data processing.

5. The weld defect detection method based on multi-data fusion according to claim 1 is characterized in that: The preprocessing of the weld detection data specifically includes: Preprocessing the ultrasonic echo data based on wavelet transform and time-frequency analysis includes: performing wavelet decomposition on the ultrasonic echo data to obtain wavelet transform coefficients; using modulation factors to perform time-frequency reconstruction on the wavelet transform coefficients; dynamically optimizing the reconstructed video based on adaptive time-frequency analysis to obtain optimized ultrasonic echo signals; Extracting ultrasonic signal defect features based on the optimized ultrasonic echo signal; Curvature analysis is performed on laser ranging data and image data to extract weld geometric features.

6. The weld defect detection method based on multi-data fusion according to claim 5 is characterized in that: The calculation steps of preprocessing the ultrasonic echo data based on wavelet transform and time-frequency analysis include: ； in, represents the optimized time-frequency representation signal; Indicates j Layer wavelet transform coefficients; represents the modulation factor, which is used for time-frequency reconstruction; G ( f ) represents an adaptive filter function; J Indicates the number of wavelet decomposition layers; s Represents the signal scale parameter.

7. The weld defect detection method based on multi-data fusion according to claim 1 is characterized in that: The fusion processing of the pre-processed data specifically includes: Dynamically adjust the weight according to the confidence of sensor data to obtain the weight under the set threshold; Based on the weight under the set threshold, the improved Kalman filter is used for state prediction and error correction, and the pre-processed data is time-synchronized to obtain time-synchronized data; Through the geometric transformation constraints between sensors, the preprocessed data are spatially aligned to obtain spatially aligned data.

8. The weld defect detection method based on multi-data fusion according to claim 1 is characterized in that: The calculation step of dynamically adjusting the weight according to the sensor data confidence level includes: ； ； ； in, X t Represents the target state vector at time t, including the fused position information and echo features; X t-1 represents the target state vector at time t-1; F represents the state transfer matrix; G represents the control input matrix; U t Represents external control input; W t represents process noise; H represents the observation matrix; Z t represents the observation vector; V t represents the measurement noise; w i represents the Bayesian adaptive weight, which is dynamically adjusted according to the sensor signal-to-noise ratio; represents the target state vector at time t The estimated value of the state; Indicates i The target state vector at time t The estimated value of the state.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the weld defect detection method based on multi-data fusion as described in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the weld defect detection method based on multi-data fusion as described in any one of claims 1 to 8 is implemented.

Images