Device and method for microwave nondestructive testing of electric melting joint of PE (Poly Ethylene) pipeline

Abstract

The invention provides a device and method for microwave nondestructive testing of a PE pipeline electric melting joint, the device comprises a walking module, a signal module and a control module, the walking module comprises a fixed part and a moving part, the fixed part comprises a plurality of pipeline clamps, and the pipeline clamps wrap the periphery of a PE pipeline; any two pipeline clamps are connected through a clamp connecting fitting, the moving part comprises a rotary moving mechanism, an X-axis moving mechanism and a Z-axis moving mechanism, and the signal module and the control module are connected and both installed on the walking module. According to the method, the defect detection precision is improved, the detection safety is enhanced, online real-time monitoring is realized, manual intervention is reduced, the adaptability and flexibility are improved, and comprehensive data support is provided.

Description

Technical Field

[0001] The present invention belongs to the technical field of pipeline detection, and particularly relates to a microwave non-destructive detection device and method for electrofusion joints of PE pipes. Background Art

[0002] Due to its excellent corrosion resistance, toughness, and convenient construction method, polyethylene (PE) pipes have been widely used in urban gas, water supply and drainage, and other fields. In the connection of PE pipes, electrofusion joints, as an important connection method, heat the joints and pipes through resistance wires to melt and combine them to form a firm connection. However, the quality of electrofusion joints is directly related to the safety and reliability of the entire pipeline system. If defects such as pores, poor fusion, or excessive internal stress occur during the heating or cooling process of electrofusion joints, it may lead to joint damage, thereby affecting the overall safety of the pipeline.

[0003] To ensure the connection quality of electrofusion joints, traditional methods such as manual visual inspection or destructive testing are used to evaluate the reliability of joints. However, visual inspection is difficult to detect internal defects, and destructive testing not only has high costs but also has an irreversible impact on the installed pipeline system.

[0004] In recent years, microwave non-destructive testing technology has received increasing attention in the field of pipeline detection due to its advantages such as non-contact, penetrability, and sensitivity to electromagnetic characteristics. Compared with traditional detection methods such as ultrasonic and X-ray, microwave non-destructive testing can not only penetrate non-metallic materials but also detect minute defects inside materials with high sensitivity. Through microwave detection, the fusion quality of electrofusion joints of PE pipes can be monitored in real time, and potential voids, unfused areas, or other structural defects can be identified, thereby providing reliable guarantee for the safe operation of pipelines.

[0005] Therefore, using microwave non-destructive testing technology to detect the quality of electrofusion joints of PE pipes can not only improve the efficiency and accuracy of detection but also effectively avoid the influence of human factors on detection results, reduce costs and risks, and enhance the safety and reliability of the pipeline system.

[0006] Technical Solution of the First Prior Art The existing ultrasonic detection technology is a common non-destructive testing technology that uses the propagation characteristics of ultrasonic waves in materials to detect internal defects. For the electrofusion joints of PE pipes, ultrasonic waves can penetrate the materials and reflect echoes, thereby judging the integrity of the internal structure. The good ultrasonic transmission performance of PE materials makes ultrasonic detection a feasible method for evaluating the quality of electrofusion joints.

[0007] Disadvantages of the First Prior Art 1. Insufficient sensitivity to defect shape: Ultrasonic testing has different detection sensitivities for different types of defects (such as pores, cracks, and unfused areas). For smaller pores or defects with complex geometries, it may be difficult for ultrasonic waves to accurately detect or identify them. In addition, defects parallel to the wave propagation direction (such as laminar delamination) may not be easily detected by ultrasonic waves.

[0008] 2. High requirements for surface conditions: Ultrasonic testing has high requirements for the smoothness of the pipe surface and the use of coupling agents. Since the surface of PE material is relatively smooth, a coupling agent needs to be applied during testing to ensure good propagation of ultrasonic signals. However, the on-site conditions are complex, and sometimes there may be dirt, oxide layers, or moisture on the pipe surface, which affects the propagation effect of ultrasonic waves and thus reduces the detection accuracy.

[0009] 3. High operation requirements: Ultrasonic testing requires experienced technicians to operate. Especially in the case of complex joint structures or large pipe diameters, the operators need to have a strong professional background and technical level. The placement of ultrasonic probes, angle selection, and signal processing will all affect the test results, which increases the uncertainty of the detection.

[0010] Solution of the second prior art: Existing X-ray testing is a common non-destructive testing method. Based on the penetrability of X-rays and the absorption differences of different materials for rays, it can identify internal defects of materials, such as pores, cracks, or lack of fusion. X-rays can penetrate non-metallic materials such as PE (polyethylene), so it is used to detect the quality of electrofusion joints of PE pipes.

[0011] Disadvantages of the second prior art: 1. Safety issues: X-rays have strong radioactivity and pose radiation hazards to operators and the surrounding environment. Strict safety protection measures must be taken, which increases the complexity and cost of operation.

[0012] 2. Not suitable for on-line detection: X-ray testing needs to be carried out in a special testing environment and it is difficult to achieve on-site or on-line real-time detection, which may not be very applicable when large-scale detection or rapid detection is required.

[0013] Application No. CN202322805607.4, the utility model is named a microwave system for detecting defects in hot-melt joints of polyethylene pipes, which discloses a fixture, an encoder, a single-chip microcomputer, an open waveguide, a microwave analyzer and a computer. The fixture includes clamping blocks and rollers. A notch is provided in the middle of the clamping blocks. The open waveguide is inserted into the notch. The open waveguide is connected to the microwave analyzer through a coaxial cable. The microwave analyzer is connected to the computer through a network cable. The rollers are arranged below the clamping blocks. The encoder is arranged on one side of the rollers. The encoder is connected to the single-chip microcomputer. The single-chip microcomputer is connected to the computer through a USB cable.

[0014] The defects it has are as follows: 1. Lack of automatic compensation for the change in lift-off distance: The prior art requires a constant lift-off distance between the waveguide and the pipe. Otherwise, signal changes may lead to false detection or missed detection. This design is difficult to adapt to practical applications in complex environments.

[0015] 2. Dependence on a fixed waveguide platform: The prior art usually needs to use a scanning platform to fix the waveguide, resulting in high equipment costs, complex installation and poor portability.

[0016] 3. Lack of real-time and accurate defect imaging: The existing system cannot provide real-time defect imaging, and subsequent data processing is required to judge the defect position and size. In summary, the present invention has significant advantages over the prior art in terms of portability, detection depth, error compensation and real-time imaging.

[0017] CN201910519821.9, the invention is named a method and device for detecting pipeline surface corrosion defects based on microwave non-destructive testing. The method includes collecting microwave signals reflected from the surface of the pipeline to be measured, using microwave imaging technology to perform imaging processing on the phase values of the reflection coefficients, performing edge detection processing on the imaging results to obtain defect position and width information, constructing a defect detection model, and calculating defect depth information. This invention uses the phase values of microwave reflection coefficients at different detection positions as imaging pixel points, and at the same time identifies the boundary, so as to obtain defect position and width information. At the same time, according to the relationship model between the phase difference of the reflection coefficient and the defect size, the quantitative information of the defect depth is inversely solved to realize the quantitative detection of the position, width and depth of defects on the pipeline.

[0018] The defects of a method and device for detecting pipeline surface corrosion defects based on microwave non-destructive testing are as follows: (1). Limited imaging resolution and accuracy: The prior art uses the Canny operator for edge detection. Although it can identify the defect position and width, the detection accuracy for tiny defects is still limited, and it is more sensitive to noise during the imaging process, which may lead to blurred boundaries.

[0019] (2). Limited depth detection accuracy: The existing technology estimates the defect depth through the simple linear relationship between the phase difference and the defect depth, which may lead to insufficient accuracy of the depth information, especially the accuracy decreases in the case of complex defects or shallow depths.

[0020] (3). Low environmental adaptability: The existing technology lacks an automatic compensation mechanism for the change in the lift-off distance between the probe and the pipeline, and the detection results are easily affected by the changes in the probe position and the pipeline surface condition.

[0021] (4). Poor real-time performance: The existing methods rely on a fixed scanning platform and a vector network analyzer for detection. The equipment is large in volume and poor in portability, and it is difficult to achieve real-time detection in complex pipeline structures.

[0022] In summary, the present invention has obvious advantages over the existing technology in terms of detection accuracy, environmental adaptability, real-time performance, etc. Summary of the Invention

[0023] The purpose of the present invention is to solve the defects existing in the above-mentioned existing technology, and provide a microwave non-destructive testing device for the electrofusion joint of PE pipelines, which uses the algorithm in the control module to solve problems such as low detection accuracy, low defect recognition rate, inability to monitor in real time, interference of energized joints and resistance wires, etc.

[0024] The present invention ensures accurate detection even under the change of lift-off distance through an automatic compensation mechanism. The fixture of the present invention is simple and lightweight, suitable for on-site detection applications, and improves the flexibility and practicality of the equipment. Through the real-time scanning and feedback of microwave signals, the present invention can generate real-time defect scanning images on a computer to achieve efficient and intuitive defect positioning.

[0025] The present invention adopts the following technical solutions: A microwave non-destructive testing device for the electrofusion joint of PE pipelines, comprising a walking module, a signal module, and a control module.

[0026] The walking module includes a fixed part and a moving part.

[0027] The fixed part includes a pipeline fixture. A plurality of pipeline fixtures are wrapped around the periphery of the PE pipeline, and any two pipeline fixtures are connected by a fixture connection fitting.

[0028] The moving part includes a rotary motion mechanism, an X-axis motion mechanism, and a Z-axis motion mechanism.

[0029] The rotary motion mechanism includes a driving shaft I, which is connected to the output shaft of a rotary servo motor. A rotary transmission gear is installed on the driving shaft I. The rotary transmission gear meshes with a magnetic wheel transmission gear I on the driving shaft II and a magnetic wheel transmission gear II on the driving shaft III respectively. Magnetic wheels are installed on both the driving shaft II and the driving shaft III. The magnetic wheels are adsorbed in contact with the fixture. Since the gear diameter is greater than the length of the connecting fitting gap between the fixtures.

[0030] The X-axis motion mechanism includes a support column. One end of the support column is installed on the upper part of the driving shaft I. The X-axis slide table housing is installed at the other end of the support column. The X-axis lead screw is installed inside the X-axis slide table housing. The X-axis lead screw meshes with the X-axis telescopic shaft. One end of the X-axis lead screw is installed with an X-axis servo motor. The other end of the X-axis telescopic shaft is installed with the housing of the Z-axis slide table.

[0031] The Z-axis motion mechanism includes a Z-axis slide table. There is a Z drive shaft inside the Z-axis slide table. There is a Z-axis slider inside the Z-axis slide table. The Z-axis slider is sleeved on the Z drive shaft. One side of the Z-axis slide table is open. The microwave antenna is connected to the Z-axis slider through the opening. One end of the Z drive shaft is connected to a Z-axis servo motor.

[0032] The laser locator is fixed on the microwave antenna. The control signal positioning module is installed on the laser locator. A control module and a signal module are installed inside the control signal positioning module. The laser locator is signal-connected to the control module inside the control signal positioning module. The microwave antenna is connected to the signal module. The signal module is connected to the control module.

[0033] Furthermore, it also includes a control module. The control module is installed inside the control signal positioning module, and the control module is signal-connected to the display module.

[0034] A detection method for the electrofusion joint of a PE pipeline by microwave non-destructive testing includes the following steps: Step 1. After installing and fixing this device, perform a full circumferential scan on the electrofusion joint of the PE pipeline, collect microwave signals, and perform preliminary processing on the signals to remove part of the noise and interference.

[0035] Step 2. The received microwave signals are preprocessed by the signal module. The processed microwave signal data is uploaded to the control module and transmitted to the remote monitoring center in real time through the network interface to support remote monitoring and data analysis.

[0036] Step 3. According to the microwave imaging diagram in Step 2, the control module extracts the defect characteristics of the welded joint. The control module uses machine learning algorithms to classify the characteristics and compares them with the historical defect samples in the database.

[0037] Step 4. Based on the proposed defect characteristics, the control module generates an integrity and completeness evaluation report for the electrofusion joint of the PE pipeline by itself.

[0038] Further, the installation in step 1 includes first installing the support frame of the device on the PE pipeline to be detected, making the device adapt to pipelines of different diameters through adjustable pipeline clamps, and ensuring stability. The operator uses a laser locator to align with the central area of the electrofusion joint of the PE pipeline, and then gradually adjusts the tightening degree of the device by rotating and locking the pipeline clamp, so that the device maintains a stable contact pressure with the surface of the PE pipeline.

[0039] Next, the magnetic wheel is strongly magnetically adsorbed on any pipeline clamp. The walking module is controlled by a servo motor. The rotating servo motor drives the rotating shaft Ⅰ to make the device rotate circumferentially along the PE pipeline, ensuring a 360-degree non-blind-angle scan of the electrofusion joint of the PE pipeline. The X-axis servo motor and Z-axis servo motor of the walking device are used for the operator to precisely adjust the horizontal and vertical positions of the microwave antenna, ensuring that the microwave antenna is aligned with the central area of the electrofusion joint of the PE pipeline.

[0040] After the microwave antenna is positioned in the detection area, the control module sends an instruction. After starting, the device rotates circumferentially. During the rotation process, the microwave antenna emits microwave signals at a specific frequency. The signals penetrate the PE pipeline and are reflected back to the microwave antenna.

[0041] Further, the full circumferential scan includes: first determining the initial position, recording the initial coordinates using a laser locator, and then the probe scans along the periphery of the pipeline. After each complete scan, the moving platform controls the microwave antenna to move a preset distance to the right, and then conducts the next round of full circumferential scan along the PE pipeline. This is repeated until the entire electrofusion joint area of the PE pipeline is covered.

[0042] Specifically, it includes: first aligning with the area to be detected through a laser locator, ensuring that the microwave antenna maintains a constant distance from the surface of the electrofusion joint of the PE pipeline, and at the same time ensuring that the scanning path of the microwave antenna covers the entire electrofusion joint area of the PE pipeline.

[0043] Further, it also includes an automatic obstacle avoidance function. The automatic obstacle avoidance includes: Assumption: Detect the position of the energized joint of the electrofusion joint of the PE pipeline . The current position of the microwave antenna is . The safe distance between the microwave antenna and the energized joint of the electrofusion joint of the PE pipeline . The motion speed vector of the microwave antenna is . The distance between the motion direction of the microwave antenna and the energized joint is .

[0044] Real-time calculate the Euclidean distance between the microwave antenna and the energized joint : (1) When the distance between the microwave antenna and the energized joint is less than the safe distance an obstacle avoidance is triggered: and the obstacle avoidance starts.

[0045] To avoid the energized joint of the electrofusion joint of the PE pipe, the algorithm needs to recalculate the movement direction of the probe. The movement direction of the probe is adjusted to a direction away from the obstacle using the formula:

[0046] First, calculate the unit vector of the microwave antenna and the energized joint of the electrofusion joint of the PE pipe : (2) Next, adjust the direction vector of the probe movement to: (3) where is the obstacle avoidance adjustment coefficient, which is used to control the intensity of the probe avoidance and the degree of direction correction. The adjusted microwave antenna ensures that the microwave antenna moves in a direction away from the obstacle.

[0047] After the microwave antenna avoids the obstacle, the new movement path will be achieved through the re-planned scanning route. The new probe position will be updated according to the new velocity vector and time : t (4); where is a tiny time step. This path adjustment will continue until the distance between the microwave antenna and the energized joint i.e., the probe safely avoids the obstacle.

[0048] The laser locator will continuously monitor the distance between the microwave antenna and the energized joint . If the microwave antenna approaches the energized joint and reaches the preset safe distance threshold , the obstacle avoidance operation is triggered. Once the obstacle avoidance is triggered, the movement direction of the microwave antenna will be corrected according to formula (4), that is, by adjusting the velocity vector to avoid the energized joint and re-plan the movement path. During the obstacle avoidance process, the microwave antenna will continue to scan the PE pipe along the new path.

[0049] The further step 2 includes, first, filtering and denoising the signal to eliminate the environmental electromagnetic interference during the detection process. Subsequently, the control module optimizes the signal data based on the inverse scattering algorithm. The optimized signal data generates two-dimensional or three-dimensional microwave imaging maps of the electrofusion joints of the PE pipeline through an image reconstruction algorithm to clearly display the internal structure and potential defects of the electrofusion joints of the PE pipeline.

[0050] Furthermore, the image reconstruction algorithm is the non-negative matrix factorization inverse scattering imaging algorithm, specifically including: In the scanning imaging, it is necessary to perform a step-by-step stop scan on the entire scanning plane. This process is regarded as stopping at a position to transmit and receive signals, and then moving to the next position to continue transmitting and receiving signals, so as to obtain the target echo data. For the direction of For the height direction For the distance direction. The defect target is processed by transmitting a frequency-stepped signal. The sampling microwave antenna scans sequentially along the positive x-axis direction, and the scanning space trajectory forms a plane, that is Scanning plane.

[0051] The purpose of two-dimensional imaging is to obtain the target scattering information to be obtained Along The equivalent distribution on the plane That is, the target two-dimensional image function. The electric field Is located in the x-z plane. The sampling plane during measurement coincides with the plane Plane and is parallel to the xoy coordinate plane. The electric field quantity of the target scattering field at each sampling point is expressed as , and the ultimate goal of imaging is to reconstruct the Scattering field strength distribution corresponding to the target image plane. The known quantity of the scattering field strength Is the one corresponding to z = 0 Propagating along the positive z direction Distance generated.

[0052] Given a non-negative matrix (where all elements ), the goal of NMF is to find two non-negative matrices: ( Is the basis matrix), ( Is the coefficient matrix) Such that: WH (5); That is, through the matrix And the matrix The product of to approximate the original matrix , where Is the dimension of the low-dimensional space, satisfying min Let \(m\) denote the number of rows of matrix \(V\), and \(n\) denote the number of columns of matrix \(V\).

[0053] The goal of NMF is to minimize the difference between the original matrix and the decomposed matrix . Commonly used objective functions are based on squared error or KL divergence.

[0054] (6) where denotes the Frobenius norm.

[0055] Objective function based on KL divergence: ;

[0056] The main constraint condition in NMF is that all matrix elements must be non - negative, i.e.: (8); Update rule for matrix \(W\): (9); Update rule for matrix \(H\): (10); where denotes element - wise multiplication, and the division operation is also element - wise.

[0057] NMF effectively achieves dimensionality reduction by decomposing the high - dimensional matrix into low - dimensional matrices and . Matrix represents a set of basis vectors or patterns, and matrix represents the representation of data on these basis vectors.

[0058] For the spatial - domain spectral plane electric - field distribution d d (11); where and represent the spatial wave - number components corresponding to the spatial coordinates \(x\) and \(y\) respectively.

[0059] Studying the spectral theory in near - field azimuth imaging, from the spectral theory at the plane \(z = 0\): (12); where \(Z_0\) is the lift - off distance, is the wave - number vector, is the wave frequency. is the circular - domain function, expressed as: (13); Therefore, the transfer function is: (14); The circular domain function circ restricts the integration region within the circular region. It can be seen that not all components of the plane wave spectrum contribute to the superposition of the total field. Only the part of the plane spectrum whose propagation direction satisfies the circular domain function circ contributes. From Equation (1-4), it is known that during the target imaging process from the scanning plane to z = 0, the spatial wave spectrum is like a two-dimensional band-pass filter. Its amplitude-frequency responses within, on, and outside the circular domain with a radius of the wave number vector are 1, 1 / 2, and 0, respectively. Within the passband, given that the echo data on the sampling plane z = z0 is , according to the wave spectrum expansion theory: d d (15); Among them,

[0060] is the spatial domain spectrum, which is obtained by performing a two-dimensional Fourier transform FT2D on the echo signal on the plane, that is: d (16); Among them, and are the corresponding ranges of the scanning plane.

[0061] So far, through the two-dimensional inverse Fourier transform of the two-dimensional spatial spectrum, the algorithm derivation is completed, and the scattering distribution function of the target can be obtained. That is, the target imaging expression is: (17).

[0062] Further, the machine learning algorithm in Step 3 is the electrofusion joint defect detection algorithm for PE pipes, including: the signal module performs preprocessing of normalizing the acquired microwave reflection signal.

[0063] The preprocessed signal is sliced according to the preset slice width and step size to generate multiple signal segments of the same length, ensuring that each generated slice signal has a fixed length.

[0064] It is automatically labeled according to whether it contains defect parts (such as holes, slag inclusions, or lack of fusion) in the electrofusion joint of the polyethylene pipe.

[0065] A method for extracting multi-dimensional features in the time domain and frequency domain for each sliced signal and the complete signal with tags; these features include but are not limited to the amplitude, phase, spectral features, power spectral density (PSD), etc. of the signal.

[0066] Screen the prepared feature dataset, and then use the Random Forest Classifier to train the screened feature dataset to construct a defect detection classification model.

[0067] Furthermore, in the feature dataset screening stage, first screen out the features that have a significant impact on defect classification through the Mann-Whitney U test, and then optimize the feature set in combination with recursive feature elimination (RFE), and finally retain the most important ten features.

[0068] Among them, the Mann-Whitney U test can effectively compare the differences between defects and normal signals in a certain feature, and screen out the features with significant differences between the two types of samples. The recursive feature elimination algorithm gradually improves the model performance by removing features with smaller weights, and finally retains the features crucial for defect recognition. This method is a non-parametric statistical method used to compare the differences between defect signals and normal signals in a certain feature; for a given feature, it is divided into two groups: the defect sample group and the normal sample group, and the U value is calculated: (19); Among them, is the number of defect samples, is the number of normal samples, is the sum of the ranks of the defect sample group. If the distribution of the feature in the two groups is significantly different, it means that the feature has a greater discrimination ability for the classification task, otherwise it is weaker. By performing the Mann-Whitney U test on each feature, select the features with a p-value less than the threshold of 0.01 as important candidate features and retain them for the next recursive feature elimination.

[0069] The recursive feature elimination (RFE) algorithm recursively trains the model and removes the least important features. Based on the support vector classifier training, calculate the weight of each feature, and measure the importance of the feature according to the absolute value of the weight. The features with smaller absolute values of the weight are considered to contribute less to the classification result. The feature importance function is defined as: (20); Among them, X i is the feature vector representing the i-th signal segment, is the weight of the classifier decision. Remove the features with smaller weights and repeat the training process until ten of the most important features are retained; and recursively perform feature selection through the following optimization process: (21); Where C is the regularization parameter that controls the influence of the weight size in the feature selection process, b is the bias term, the offset of the decision boundary of the classifier, which is used to adjust the relative position of the hyperplane, Xi is the feature, is the label of the i-th sample, and is the weight of the classifier's decision.

[0070] Furthermore, in the random forest classifier, the extracted feature dataset is defined by the feature vector as: (18); Where, represents the feature vector of the i-th signal segment, and the feature dimension includes time-domain features and frequency-domain features. Each feature vector corresponds to a label , where, represents the normal area corresponding to the signal segment, indicates that the signal segment contains a defect area.

[0071] During the training process, the random forest classifier builds multiple decision trees and uses the majority voting method to determine the final classification result. Each tree is trained independently, and the nodes of the tree are constructed by randomly selecting features. Through this ensemble method, the random forest can effectively avoid overfitting and enhance the ability to identify different types of defects. The performance of the defect detection classification model is evaluated by metrics such as cross-validation, accuracy, precision, recall, and F1-score, and its generalization ability is verified on the test set. Finally, the trained random forest classifier can automatically detect new weld signals and determine whether they contain defects, thus realizing the automation and intelligence of defect detection.

[0072] Advantages of the present invention: Through normalized signal processing and slice analysis, the present invention can improve the sensitivity to micro-defects (such as holes, slag inclusions, lack of fusion), and eliminate the interference of resistance wires, thereby effectively identifying and locating potential defects in electrofusion joints and ensuring pipeline quality. The automation feature of the present invention reduces the complexity of manual operations, simplifies the operation process, enabling detection to be carried out smoothly even at a low skill level, and enhancing the usability. Through the non-contact microwave detection method, the safety risk of operators during the detection process is reduced, and potential hazards caused by direct contact are avoided. An automatic obstacle avoidance algorithm is integrated, which can effectively handle the energized joints in electrofusion joints, avoid the interference of energized joints, and ensure the smooth progress of the detection process. The present invention supports online real-time detection, can continuously monitor during the pipeline operation process, timely discover and handle defects, and improves the pipeline maintenance efficiency and safety. Description of the Drawings

[0073] Figure 1It is a module diagram of a device for non-destructive microwave detection of electrofusion joints of PE pipes according to the present invention.

[0074] Figure 2 It is a device for non-destructive microwave detection of electrofusion joints of PE pipes according to the present invention.

[0075] Figure 3 It is a schematic diagram of a fixture of a device for non-destructive microwave detection of electrofusion joints of PE pipes provided by the present invention.

[0076] Figure 4 It is a schematic diagram of the structure of the rotary motion mechanism according to the present invention.

[0077] Figure 5 It is a schematic diagram of the structure of the Z-axis motion mechanism according to the present invention.

[0078] Figure 6 It is a schematic diagram of the structure of the X-axis motion mechanism according to the present invention.

[0079] In the figure: 1 - PE pipe, 2 - electrofusion joint of PE pipe, 3 - internal resistance wire, 4 - power-on joint, 5 - outer shell of electrofusion joint of PE pipe, 6 - laser locator, 7 - microwave antenna, 8 - Z-axis servo motor, 9 - Z-axis slide, 10 - X-axis lead screw, 11 - outer shell of X-axis slide, 12 - X-axis servo motor, 13 - control signal positioning module, 14 - drive shaft Ⅰ, 15 - support pillar, 16 - rotary servo motor, 17 - gear, 18 - pipe fixture Ⅰ, 19 - fixture connection fitting Ⅰ, 20 - pipe fixture Ⅱ, 21 - fixture connection fitting Ⅱ, 22 - pipe fixture Ⅲ, 23 - fixture connection fitting Ⅲ, 24 - magnetic wheel drive gear Ⅰ, 25 - magnetic wheel drive gear Ⅱ, 26 - rotary drive gear, 27 - drive shaft Ⅰ, 28 - drive shaft Ⅱ, 29 - Z drive shaft, 30 - Z-axis slider, 31 - X-axis telescopic shaft; 110 - signal module, 120 - control module, 130 - walking module, 140 - display module. Specific embodiments

[0080] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without making creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention.

[0081] In view of the defects existing in the existing detection technology for electrofusion joints of PE pipes, especially problems such as insufficient detection accuracy, low defect recognition rate, inability to monitor in real time, interference from energized joints and resistance wires, etc., the present invention proposes a microwave non-destructive detection device and method for electrofusion joints of PE pipes. By emitting microwave signals of a specific frequency, this device can penetrate the pipe wall and detect the defects of the electrofusion joint. At the same time, advanced signal processing algorithms are adopted to eliminate noise interference in a complex environment, ensuring the stability and detection accuracy of the signal. This method can monitor the physical state of the electrofusion joint in real time, accurately identify the location and nature of welding defects, effectively avoid false alarms and missed detections in traditional detection methods, greatly improve the efficiency and reliability of pipeline detection, and achieve intelligent and highly accurate detection of electrofusion joints of pipes.

[0082] As Figures 1-6 shown, a microwave non-destructive detection device for electrofusion joints of PE pipes according to the present invention includes a traveling module 130, a signal module 110, a control module 120, and a display module 140.

[0083] The traveling module 130 includes a Z-axis servo motor 8, a Z-axis slide 9, a Z drive shaft 29, a Z-axis slider 30, an X-axis lead screw 10, an X-axis slide housing 11, an X-axis telescopic shaft 31, an X-axis servo motor 12, a drive shaft I 14, a rotary drive gear 26, a magnetic wheel drive gear I 24, a magnetic wheel drive gear II 25, a drive shaft II 27, a drive shaft III 28, a support 15, a rotary servo motor 16, a magnetic wheel 17, a pipe positioning fixture I 18, a fixture connection fitting I 19, a pipe fixture II 20, a fixture connection fitting II 21, a pipe fixture III 22, and a fixture connection fitting III 23.

[0084] In the present invention, the connection between two PE pipes 1 is an electrofusion joint 2 of the PE pipe. Inside the electrofusion joint 2 of the PE pipe, there is an internal resistance wire 3 of the electrofusion joint of the PE pipe. Outside the electrofusion joint 2 of the PE pipe is the outer shell 5 of the electrofusion joint of the PE pipe. On the outer shell 5 of the electrofusion joint of the PE pipe, there is an energized joint 4 of the electrofusion joint of the PE pipe.

[0085] The pipe clamp Ⅰ 18, pipe clamp Ⅱ 20, and pipe clamp Ⅲ 22 are wrapped around the periphery of the PE pipe 1. The pipe clamp Ⅰ 18 and the pipe clamp Ⅱ 20 are connected by the clamp connection fitting Ⅰ 19. The pipe clamp Ⅱ 20 and the pipe clamp Ⅲ 22 are connected by the clamp connection fitting Ⅱ 21. The pipe clamp Ⅲ 22 and the pipe clamp Ⅰ 18 are connected by the clamp connection fitting Ⅲ 23. Of course, according to the outer diameter of the PE pipe 1, other numbers of pipe clamps and clamp connection fittings can also be set, and there is no limit on the quantity here. The pipe clamp is used to fix and support the PE pipe 1 to ensure its stability during the test. The clamp can prevent the PE pipe 1 from moving or deforming due to external forces or environmental changes, and at the same time, it can also provide the necessary support and convenience during the detection. The clamp connection fitting is used to connect the support structures. These fittings can ensure the firm connection between the clamp and the pipe, enhance the overall structural strength and stability of the system, and ensure that the PE pipe 1 will not become loose or fall off during the detection process.

[0086] As Figure 4 shown, the rotary motion mechanism: The magnetic wheel 17 is adsorbed on the clamp 18 (the material is iron, steel, etc.). Since the diameter of the gear 17 is larger than the length of the gap between the clamp connection fittings, it can stably pass through the gap between the clamp connection fittings, so it can achieve 360° rotation. The drive shaft Ⅰ 14 is connected to the output shaft of the rotary servo motor 16. A rotary transmission gear 26 is installed on the drive shaft Ⅰ 14. The rotary transmission gear 26 meshes with the magnetic wheel transmission gear Ⅰ 24 on the drive shaft Ⅱ 27 and the magnetic wheel transmission gear Ⅱ 25 on the drive shaft Ⅲ 28 respectively. Magnetic wheels 17 are installed on both the drive shaft Ⅱ 27 and the drive shaft Ⅲ 28. The magnetic wheels 17 help to reduce mechanical wear and maintenance costs. The rotary servo motor 16 is used to drive the entire detection device to achieve multi-angle rotation during the detection process, so as to cover all directions of the electrofusion joint 2 of the PE pipe.

[0087] The X-axis motion mechanism: One end of the support column 15 is installed on the upper part of the drive shaft Ⅰ 14. The X-axis slide table housing 11 is installed on the other end of the support column 15. The X-axis lead screw 10 is installed inside the X-axis slide table 11 housing. The X-axis lead screw 10 meshes with the X-axis telescopic shaft 31. One end of the X-axis lead screw 10 is installed with the X-axis servo motor 12. The other end of the X-axis telescopic shaft 31 is installed with the housing of the Z-axis slide table 9. When the X-axis servo motor 12 receives the instruction from the control module 120, it starts to generate a rotary motion and transmits this motion to the X-axis lead screw 10. The rotation of the X-axis lead screw 10 drives the X-axis telescopic shaft 31 to achieve precise horizontal movement along the preset guide rail. The horizontal movement of the X-axis telescopic shaft 31 further drives the overall relevant detection components to achieve synchronous left and right movement, ensuring that the microwave antenna 7 always remains in the best detection position.

[0088] In addition, the left - right movement is monitored in real - time and fine - tuned through the feedback signal of the control module 120, ensuring the smoothness of the telescopic shaft movement and the accuracy of detection and positioning. In this way, the microwave antenna 7 can cover a wider detection area in the X - axis direction, conduct a comprehensive defect detection on the pipeline, improve the detection sensitivity and reliability of the system. This design can adapt to different detection requirements and meet the precise detection of the electro - fusion joints of the pipeline at different horizontal positions and angles. Through the coordinated movement of the X - axis and the Z - axis, the entire detection system achieves a full coverage in the three - dimensional space, further improving the detection efficiency and accuracy, and ensuring the safe operation of the pipeline.

[0089] Z - axis movement mechanism: Inside the Z - axis slide 9, there is a Z - drive shaft 29. There is a Z - axis slider 30 inside the Z - axis slide 9, and the Z - axis slider 30 is sleeved on the Z - drive shaft 29. One side of the Z - axis slide 9 is open, and the microwave antenna 7 is connected to the Z - axis slider 30 through the opening. One end of the Z - drive shaft 29 is connected to the Z - axis servo motor 8. When the Z - axis servo motor 8 receives the command information from the control module 120, it starts to generate a rotational movement and transmits this movement to the Z - drive shaft 29. The rotation of the Z - drive shaft 29 drives the Z - axis slider 30 to move precisely up and down along the preset guide rail. The vertical movement of the Z - axis slider 30 further drives the microwave antenna 7 and its related detection components to achieve synchronous up - and - down movement as a whole, ensuring that the microwave antenna always remains in the optimal detection position.

[0090] The laser locator 6 is fixed on the microwave antenna 7, the control signal positioning module 13 is installed on the laser locator 6, the control module 120 is installed inside the control signal positioning module 13, the laser locator 6 is signal - connected to the control module 120 inside the control signal positioning module 13, and the microwave antenna 7 is connected to the signal module 110. The signal module 110 is connected to the control module 120.

[0091] The laser locator 6 is used to precisely locate the specific position of this device relative to the pipeline. It emits a laser beam to determine the distance and angle between this device and the electro - fusion joint 2 of the PE pipeline, so as to ensure that the microwave antenna 7 can accurately align with the detection area of the electro - fusion joint 2 of the PE pipeline, avoid position deviation. The laser locator 6 improves the positioning accuracy and ensures the accuracy of the detection data.

[0092] The microwave antenna 7 is the core component responsible for transmitting and receiving microwave signals in this device. It is used to transmit microwaves of a specific frequency, penetrate the wall of the PE pipeline, and reflect back to the internal structure of the electro - fusion joint 2 of the PE pipeline. The received signal is used to analyze the state of the electro - fusion joint 2 of the PE pipeline and identify internal welding defects such as cracks or bubbles. The performance of the microwave antenna 7 directly affects the detection sensitivity and accuracy.

[0093] The Z-axis servo motor 8 is responsible for controlling the movement in the Z-axis direction (vertical direction). This precise vertical movement ensures that the microwave antenna 7 can adjust its height according to the detection needs, maintaining an appropriate distance from pipes of different sizes to optimize the signal transmission and reception effects. The high-precision control of the servo motor improves the flexibility and adaptability of the detection device.

[0094] The Z-axis slide 9 provides a guiding structure for vertical movement in this detection device. Cooperating with the Z-axis servo motor 8, it enables the stable movement of the device in the vertical direction. The Z-axis slide 9 ensures the stable and precise up-and-down movement of the device, helping the antenna to align with the area to be detected and adapt to pipe structures of different heights.

[0095] The X-axis lead screw 10 is used to convert the rotational motion of the X-axis servo motor 12 into linear motion, enabling the device to move along the X-axis (horizontal direction). The threaded structure of the X-axis lead screw 10 ensures the accuracy and stability during the movement, helping the device to perform fine positioning in the horizontal direction and ensuring a comprehensive scan of different parts of the electrofusion joint 2 of the PE pipe.

[0096] The X-axis slide housing 11 protects the X-axis lead screw 10 and the internal moving parts, ensuring that they are protected from the external environment (such as dust, moisture, or mechanical damage), thus maintaining the smooth operation of the X-axis slide. The X-axis slide housing 11 also provides structural support for the entire X-axis movement system, extending the service life of the equipment and ensuring that the device can still operate normally under harsh conditions.

[0097] The X-axis servo motor 12 is used to drive the horizontal movement of the X-axis lead screw 10. By precisely controlling the rotation of the motor, the device can move smoothly along the X-axis, ensuring that the device can cover every detection area of the electrofusion joint 2 of the PE pipe. It cooperates with the X-axis lead screw 10 to complete precise positioning, improving the detection efficiency and accuracy.

[0098] The signal module 110 is used to receive and process the signals collected through microwave non-destructive testing, and analyze and judge the detection results of the electrofusion joint 2 of the PE pipe.

[0099] The walking module 130 is used to ensure that the device can move smoothly and precisely along the position of the electrofusion joint 2 of the PE pipe to achieve a full-range and real-time detection of the electrofusion joint 2 of the PE pipe.

[0100] The control module 120 is installed within the control signal positioning module 13, and the control modules 120 are all signal-connected to the display module 140. The control module 120 ensures the synchronization of signal acquisition and device movement of this apparatus by controlling the walking module, processes the detection data in real time, avoids any time delay, and thus provides real-time detection results. This not only greatly improves the detection signal but also enables stable operation in complex environments, especially suitable for long-distance or unattended detection tasks. The control modules 120 cooperate closely to ensure that each signal acquisition and processing is precisely controlled, reducing errors.

[0101] When a defect is detected in the electrofusion joint 2 of the PE pipeline, the laser locator 6 accurately marks the defect position, the signal module 110 provides detailed signal analysis, and the control module 120 is responsible for recording and summarizing this information to form a systematic defect report for subsequent maintenance and management. Through the flexible control of the control module 120, the system can adapt to different types and specifications of pipelines and adjust the detection parameters according to environmental changes, ensuring the wide applicability and flexible operation of the detection system.

[0102] The signal module 110 is used to receive and process the signals collected through the microwave antenna 7 and analyze and judge the detection results of the electrofusion joint 2 of the PE pipeline. The control module 120 plays a role in core management and coordination to ensure the efficient and stable operation of the entire apparatus.

[0103] The display module 140 is used to provide visual information on the detection process and results for the operator to ensure that the detection data can be presented intuitively.

[0104] A detection method for non-destructively detecting the electrofusion joint of a PE pipeline by microwave includes the following steps: Step 1. After installing and fixing this apparatus, perform a full circumferential scan on the electrofusion joint 2 of the PE pipeline, collect microwave signals, and perform preliminary processing on the signals to remove some noise and interference.

[0105] Specifically, it includes: First, install the support frame of this apparatus on the PE pipeline to be detected, and make this apparatus adapt to pipelines of different diameters through the adjustable pipe clamps I 18, pipe clamps II 20, and pipe clamps III 22, and ensure stability. The operator uses the laser locator 6 to align with the central area of the electrofusion joint 2 of the PE pipeline, and then gradually adjusts the fastening degree of the equipment by rotating and locking the pipe clamps 18, pipe clamps II 20, and pipe clamps III 22 to keep a stable contact pressure between this apparatus and the surface of the PE pipeline 1, preventing this apparatus from sliding or misaligning during the detection process.

[0106] Next, the magnetic adsorption rotary motion mechanism is strongly magnetically adsorbed on any pipe clamp. The walking module is controlled by a servo motor. The rotary servo motor 16 drives the rotary shaft 14 to make the device rotate circumferentially along the PE pipe, ensuring a 360-degree non-blind-angle scan of the electrofusion joint of the PE pipe. The X-axis servo motor 12 and Z-axis servo motor 8 of the walking device are used for the operator to precisely adjust the horizontal and vertical positions of the microwave antenna 7, ensuring that the microwave antenna 7 is aligned with the central area of the electrofusion joint 2 of the PE pipe.

[0107] After the microwave antenna 7 is positioned in the detection area, the control module 120 sends an instruction to start the device to rotate circumferentially. During the rotation process, the microwave antenna 7 emits microwave signals at a specific frequency, and the signals penetrate the PE pipe 1 and are reflected back to the microwave antenna 7.

[0108] Among them, the full circumferential scan includes: first, determine the initial position, record the initial coordinates using the laser locator, and then the probe scans along the circumference of the pipe. After each full circumferential scan is completed, the moving platform controls the microwave antenna 7 to move a preset distance to the right and then conducts the next full circumferential scan along the PE pipe 1. This is repeated until the entire area of the electrofusion joint 2 of the PE pipe is covered.

[0109] Furthermore, to ensure stability, first align the laser locator 6 with the area to be detected, ensure that the microwave antenna 7 maintains a constant distance from the surface of the electrofusion joint 2 of the PE pipe, and at the same time ensure that the scanning path of the microwave antenna 7 covers the entire area of the electrofusion joint 2 of the PE pipe. To deal with the energized joint 4 of the electrofusion joint of the PE pipe that may be encountered during the detection process, the system is equipped with an automatic obstacle avoidance function. The laser locator 6 detects the energized position of the electrofusion structure. When an energized joint is detected, the system will automatically adjust the scanning path to prevent the microwave antenna 7 from contacting the energized area.

[0110] The automatic obstacle avoidance is a real-time obstacle avoidance algorithm based on the data of the laser locator 6, mainly by detecting information such as the distance and relative position of the obstacle (energized joint) to dynamically adjust the running path of the device.

[0111] Assume: the position of the energized joint is detected . The current position of the microwave antenna 7 is . The safe distance between the microwave antenna 7 and the energized joint 4 of the electrofusion joint 2 of the PE pipe (this distance is a set threshold to prevent the microwave antenna 7 from colliding with the energized joint). The motion velocity vector of the microwave antenna 7 is . The distance between the motion direction of the microwave antenna 7 and the energized joint is .

[0112] Calculate the Euclidean distance between the microwave antenna and the energized joint in real time : (1); When the distance between the microwave antenna and the energized joint is less than the safety distance , obstacle avoidance is triggered: , start obstacle avoidance.

[0113] To avoid the energized joint, the algorithm needs to recalculate the movement direction of the probe. Obstacle avoidance can be achieved by changing the direction of the velocity vector of the probe. The following formula can be used to adjust the movement direction of the probe to a direction away from the obstacle:

[0114] First, calculate the unit vector of the microwave antenna 7 and the energized joint 4 of the electrofusion joint of the PE pipe : (2); Next, adjust the direction vector of the probe movement to: (3); Among them, is the obstacle avoidance adjustment coefficient, which is used to control the intensity of the probe avoidance and the degree of direction correction. The adjusted microwave antenna 7 ensures that the microwave antenna 7 moves in a direction away from the obstacle.

[0115] After the microwave antenna 7 avoids the obstacle, the new movement path will be achieved through a re-planned scanning route. The new probe position will be updated according to the new velocity vector and time : t (4); Among them is a small time step. This path adjustment will continue until the distance between the microwave antenna 7 and the energized joint , that is, the probe safely avoids the obstacle.

[0116] The laser locator 6 will continuously monitor the distance between the microwave antenna 7 and the energized joint . If the microwave antenna approaches the energized joint and reaches the preset safety distance threshold , the obstacle avoidance operation is triggered. Once the obstacle avoidance is triggered, the movement direction of the microwave antenna 7 will be corrected according to the above formula, that is, by adjusting the velocity vector , it avoids the energized connector and re-plans the movement path. During the obstacle avoidance process, the microwave antenna will continue to scan the PE pipe 1 along the new path. The obstacle avoidance algorithm ensures that the microwave antenna will not collide with the energized connector, and the scanning task proceeds unhindered. The core advantage of this algorithm is that it can monitor the energized connector in real time and dynamically adjust the path, making the detection process safer and more reliable. In addition, since the adjustment speed vector and path planning are calculated based on the relative position and real-time distance of the microwave antenna, the obstacle avoidance process is fast and flexible, capable of adapting to complex detection environments.

[0117] By introducing an automatic obstacle avoidance algorithm based on the above formula, the control module 120 can intelligently avoid the energized connector in the electrofusion joint 2 of the PE pipe, ensuring the safe operation of the device. Through real-time distance monitoring, speed vector adjustment, and dynamic path planning, the detection process does not affect the scanning and data collection of the electrofusion joint 2 of the PE pipe by microwave non-destructive testing while ensuring safety.

[0118] Step 2. After the control module 120 receives the microwave signal data transmitted back by the signal module 110, it performs multi-level processing. It also includes that the received microwave signal is preprocessed by the signal module 110, including corresponding processing such as data filtering, amplification, and denoising. The processed microwave signal data is uploaded to the control module 120 and transmitted to the remote monitoring center in real time through the network interface to support remote monitoring and data analysis.

[0119] Specifically, first, the signal is filtered and denoised to eliminate possible environmental electromagnetic interference during the detection process. Subsequently, the control module 120 optimizes the accuracy of the signal data based on the inverse scattering algorithm. The optimized signal data generates a two-dimensional or three-dimensional microwave imaging map of the electrofusion joint 2 of the PE pipe through the image reconstruction algorithm to clearly display the internal structure and potential defects of the electrofusion joint 2 of the PE pipe.

[0120] Furthermore, the image reconstruction algorithm is the non-negative matrix factorization inverse scattering imaging algorithm, specifically including: In the scanning imaging, it is necessary to perform a step-by-step stop scanning on the entire scanning plane. This process can be regarded as stopping at a position to transmit and receive signals, and then moving to the next position to continue transmitting and receiving signals, so as to obtain the target echo data. For the direction. For the height direction. For the range direction. The defect target is processed by transmitting a frequency-stepped signal. The sampling microwave antenna 7 scans sequentially along the positive x-axis direction, and the scanning space trajectory forms a plane, that is Scanning plane.

[0121] The purpose of two-dimensional imaging is to obtain the target scattering information to be obtained Along Equivalent distribution on the plane That is, the target two-dimensional image function. Electric field Is located in the x-z plane. The sampling plane during measurement coincides with the plane Surface, parallel to the xoy coordinate plane, and the electric field quantity of the target scattering field at each sampling point can be expressed as , and the ultimate goal of imaging is to reconstruct the corresponding Scattering field strength distribution on the target image plane. Known quantity of scattering field strength Is the one corresponding to z = 0 Propagating along the +Z direction Distance generation.

[0122] Non-negative matrix factorization (NMF, Non-negative Matrix Factorization) is a method for decomposing a non-negative matrix into the product of two non-negative matrices.

[0123] Given a non-negative matrix (where all elements ), the goal of NMF is to find two non-negative matrices: (basis matrix), (coefficient matrix) Such that: WH (5); That is, approximate the original matrix through the product of matrix and matrix , where is the dimension of the low-dimensional space, satisfying min .

[0124] The goal of NMF is to minimize the difference between the original matrix and the decomposed matrix . Commonly used objective functions are based on squared error or KL divergence.

[0125] (6); Where represents the Frobenius norm.

[0126] Objective function based on KL divergence: (7); The main constraint condition in NMF is that all matrix elements must be non-negative, that is: (8); Update rule for matrix W: (9); Update rule for matrix H: (10); Among them, represents the element-wise product (Hadamard product), and the division operation is also element-wise.

[0127] NMF effectively realizes dimensionality reduction by decomposing the high-dimensional matrix into the low-dimensional matrices and . The matrix represents a set of basis vectors or patterns, and the matrix represents the representation of the data on these basis vectors.

[0128] For the spatial-domain spectral plane electric field distribution

[0129] d d (11); Among them, and respectively represent the spatial wave number components corresponding to the spatial coordinates x and y.

[0130] Studying the wave spectrum theory in near-field azimuth imaging, from the wave spectrum theory at the plane z = 0, we know that: (12); Among them, Zo: lift-off distance, is the wave number vector, is the wave frequency. is the circular domain function, expressed as: (13); So the transfer function is: (14); The circular domain function circ restricts the integration region within the circular region. It can be seen that not all components of the plane wave spectrum contribute to the superposition of the total field. Only the part of the plane spectrum whose propagation direction satisfies the circular domain function circ contributes. From Equation (1-4), it can be seen that during the target imaging process from the scanning plane to z = 0, the spatial wave spectrum is like a two-dimensional band-pass filter. Its amplitude-frequency responses within, on, and outside the circular domain with a radius of the wave number vector are 1, 1 / 2, and 0 respectively. In the passband, given that the echo data on the sampling plane z = z0 is , according to the wave spectrum expansion theory, we can obtain: d d (15); Among them,

[0131] For the spatial domain spectrum, it can be obtained from the echo signal on the plane through two-dimensional Fourier transform FT2D, that is: After two-dimensional Fourier transform FT2D, we have: d (16); where and are the corresponding ranges of the scanning plane.

[0132] So far, through the two-dimensional inverse Fourier transform of the two-dimensional spatial spectrum, the algorithm derivation is completed, and the scattering distribution function of the target can be obtained. That is, the target imaging expression is: (17);

[0133] Step 3. According to the microwave imaging diagram in Step 2, the control module extracts the defect features of the welded joint, including common defects such as incomplete welding, bubbles, cracks or lack of fusion. The control module uses machine learning algorithms to classify the features and compares them with the historical defect samples in the database to further improve the accuracy and reliability of recognition.

[0134] The specific machine learning algorithm is the defect detection algorithm for the electrofusion joint of PE pipes: The signal module performs normalization preprocessing on the acquired microwave reflection signal. The purpose of normalization is to standardize the microwave reflection signals obtained under different detection conditions, eliminate non-target fluctuations caused by external factors such as the sensitivity of the signal acquisition device, the detection environment, or the change in the position of the microwave waveguide probe, and ensure the consistency and comparability of subsequent analysis and processing. This normalization step plays an important role in signal processing, especially when analyzing and processing common defects such as holes, slag inclusions, and lack of fusion in the hot melt joints of polyethylene pipes. The standardized signal can make subsequent processing steps such as feature extraction and defect recognition more stable and accurate; in addition, the normalization process can also improve the comparability of microwave reflection signals between different detection batches, enabling the detection algorithm to maintain robustness under various environmental conditions, thereby improving the performance and reliability of the overall detection system.

[0135] The preprocessed signal is sliced according to the preset slice width and step size to generate multiple signal segments of the same length, ensuring that each generated slice signal has a fixed length. Through the slicing process, the original signal can be refined into multiple local signal segments, facilitating subsequent independent feature analysis and defect recognition for each segment. For each slice signal, further determine whether it contains defect parts (such as holes, slag inclusions or lack of fusion) in the hot melt joint of the polyethylene pipe

[0136] Automatic annotation is carried out, and this automatic annotation process lays the foundation for subsequent supervised learning tasks, enabling the system to learn the model through the annotated training data, thereby realizing automatic defect identification and classification. By using the annotated slice data, supervised learning can effectively improve the detection accuracy of the microwave detection algorithm for three common defects, ensuring that the model has high robustness and accuracy under different working conditions.

[0137] A multi-dimensional feature extraction method in the time domain and frequency domain is applied to each labeled slice signal and the complete signal to fully capture the potential defect information in the signal; these features include but are not limited to the amplitude, phase, spectral characteristics, power spectral density (PSD), etc. of the signal. The purpose of feature extraction is to convert the one-dimensional microwave reflection signal into a high-dimensional feature vector through mathematical calculations, thereby establishing a feature vector data set for weld defect identification and classification, and further improving the learning and prediction accuracy of the model.

[0138] The made feature data set is screened, and then the Random Forest Classifier is used to train the screened feature vectors to construct a defect detection and classification model.

[0139] In the feature screening stage, first, the features that have a significant impact on defect classification are screened out through the Mann-Whitney U test, and then the feature set is optimized by combining recursive feature elimination (RFE). Finally, the most important ten features are retained. The Mann-Whitney U test can effectively compare the differences in a certain feature between defects and normal signals, and screen out the features with significant differences between the two types of samples. The recursive feature elimination algorithm gradually improves the model performance by removing the features with smaller weights, and finally retains the features crucial for defect identification. This method is a non-parametric statistical method used to compare the differences in a certain feature between defect signals and normal signals; for a given feature, it is divided into two groups: the defect sample group and the normal sample group, and the U value is calculated: (19); Among them, is the number of defect samples, is the number of normal samples, is the sum of the ranks of the defect sample group. If the distribution of a feature in the two groups is significantly different, it means that this feature has a greater discrimination ability for the classification task, and vice versa. By performing the Mann-Whitney U test on each feature, the features with a p-value less than the threshold of 0.01 are selected as important candidate features and retained for the next recursive feature elimination.

[0140] The Recursive Feature Elimination (RFE) algorithm trains a model recursively and removes the least important features. Based on the support vector classifier training, the weights of each feature are calculated, and the importance of the features is measured according to the absolute value of the weights. The features with smaller absolute values of the weights are considered to contribute less to the classification result. The feature importance function is defined as: (20); where X i is the feature vector representing the i-th signal segment, is the weight of the classifier decision.

[0141] Remove the features with smaller weights and repeat the training process until ten most important features are retained; and recursively perform feature selection through the following optimization process: (21); where C is the regularization parameter, controlling the influence of the weight size in the feature selection process, b is the bias term, the offset of the classifier's decision boundary, used to adjust the relative position of the hyperplane, Xi is the feature, is the label of the i-th sample, is the weight of the classifier decision.

[0142] The PE pipeline capacitance defect detection algorithm based on microwave signals can significantly improve the detection accuracy and efficiency of electrofusion joint defects. By normalizing the collected microwave signals, the influence of external factors is effectively eliminated, making the subsequent analysis more consistent and comparable. The slicing process divides the signal into multiple segments of fixed length, facilitating independent analysis and feature extraction. At the same time, using the Random Forest Classifier for feature classification can efficiently distinguish different types of defects. In addition, the Mann-Whitney U test evaluates the importance of features to ensure that the selected features contribute significantly to defect recognition. Recursive Feature Elimination (RFE) further optimizes the feature set to improve the model performance by gradually eliminating redundant features. Finally, the trained model performs defect recognition on each signal slice and generates a detailed detection report, providing key information such as defect type and location. This series of processes ensures the safe operation of the pipeline, provides reliable data support for pipeline maintenance, and meets the requirements of modern pipeline monitoring.

[0143] Use the Random Forest Classifier to train the extracted feature dataset to construct a defect detection classification model. Random Forest is an ensemble learning method that can significantly improve the classification accuracy and robustness by constructing multiple decision trees and combining their output results. During the training process, the classifier adjusts the model parameters through the ensemble learning of multiple decision trees so that it can predict whether the signal segment contains defects according to the input features.

[0144] The feature vector in the random forest classifier is defined as: (18); where represents the feature vector of the i-th signal segment, and the feature dimensions include time-domain features and frequency-domain features. Each feature vector corresponds to a label , where represents the normal region corresponding to the signal segment, indicates that the signal segment contains a defective region.

[0145] During the training process, the random forest classifier constructs multiple decision trees and uses the majority voting method to determine the final classification result. Each tree is trained independently, and the nodes of the tree are constructed by randomly selecting features. Through this ensemble method, the random forest can effectively avoid overfitting and enhance the ability to identify different types of defects. The performance of the model is evaluated by metrics such as cross-validation, accuracy, precision, recall, and F1-score, and its generalization ability is verified on the test set. Finally, the trained random forest classifier can automatically detect new weld signals and determine whether they contain defects, thus realizing the automation and intelligence of defect detection.

[0146] Through high-precision microwave signal acquisition, intelligent data processing, and comprehensive defect detection, combined with a fully automatic scanning and control module, the present invention not only improves the detection efficiency but also greatly reduces the error of manual operation, ensuring the safety and reliability of the detection of the electrofusion joint 2 of the PE pipe.

[0147] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A microwave non-destructive testing device for electrofusion joints of PE pipes, characterized in that , including a walking module, a signal module, and a control module. The walking module includes a fixed part and a moving part. The fixed part includes pipe clamps, and multiple pipe clamps surround the PE pipe. Any two pipe clamps are connected by a clamp connection fitting. The moving part includes a rotary motion mechanism, an X-axis motion mechanism, and a Z-axis motion mechanism. The signal module is connected to the control module and both are installed on the walking module.

2. The microwave non-destructive testing device for electrofusion joints of PE pipes according to claim 1, characterized in that, The rotary motion mechanism includes a drive shaft Ⅰ, which is connected to the output shaft of a rotary servo motor. A rotary transmission gear is installed on the drive shaft Ⅰ. The rotary transmission gear meshes with a magnetic wheel transmission gear Ⅰ on the drive shaft Ⅱ and a magnetic wheel transmission gear Ⅱ on the drive shaft Ⅲ respectively. Magnetic wheels are installed on both the drive shaft Ⅱ and the drive shaft Ⅲ, and the magnetic wheels are adsorbed in contact with the clamp. Since the gear diameter is greater than the length of the clearance between the clamp connection fittings.

3. The microwave non-destructive testing device for the electrofusion joint of PE pipes according to claim 1, wherein, The X-axis motion mechanism includes a support column. One end of the support column is installed on the upper part of the drive shaft Ⅰ. The X-axis slide table housing is installed on the other end of the support column. The X-axis lead screw is installed inside the X-axis slide table housing. The X-axis lead screw meshes with the X-axis telescopic shaft. One end of the X-axis lead screw is installed with an X-axis servo motor, and the other end of the X-axis telescopic shaft is installed with the housing of the Z-axis slide table.

4. The microwave non-destructive testing device for the electrofusion joint of PE pipes according to claim 1, characterized in that, The Z-axis motion mechanism includes a Z-axis slide table. The Z-axis slide table has a Z drive shaft inside. There is a Z-axis slider inside the Z-axis slide table. The Z-axis slider is sleeved on the Z drive shaft. One side of the Z-axis slide table is open, and the microwave antenna is connected to the Z-axis slider through this opening. One end of the Z drive shaft is connected to a Z-axis servo motor.

5. The microwave non-destructive testing device for electrofusion joints of PE pipes according to claim 1, wherein The laser locator is fixed on the microwave antenna. The control signal positioning module is installed on the laser locator. The control module and the signal module are installed inside the control signal positioning module. The laser locator is signal-connected to the control module inside the control signal positioning module. The microwave antenna is connected to the signal module; It further includes a control module. The control module is installed inside the control signal positioning module, and the control module is signal-connected to the display module.

6. A detection method for the electrofusion joint of a PE pipeline by microwave non-destructive testing, characterized in that, It includes the following steps: Step 1. After installing and fixing the device as described in Claim 1, perform a full circumferential scan on the electrofusion joint of the PE pipe, collect microwave signals, and perform preliminary processing on the signals to remove some noise and interference; Step 2. The received microwave signals are preprocessed by the signal module. The processed microwave signal data is uploaded to the control module and transmitted to the remote monitoring center in real time through the network interface. The remote monitoring center supports remote monitoring and data analysis; Step 3. According to the microwave imaging diagram in Step 2, the control module extracts the defect characteristics of the welded joint. The control module uses machine learning algorithms to classify the characteristics and compares them with the historical defect samples in the database; Step 4. Based on the proposed defect characteristics, the control module generates an integrity and completeness evaluation report for the electrofusion joint of the PE pipe by itself.

7. The method according to claim 6, wherein The full circumferential scan in Step 1 further includes an automatic obstacle avoidance function. The automatic obstacle avoidance includes: Hypothesis: Detect the position of the energized joint of the electrofusion joint of the PE pipeline , the current position of the microwave antenna is , the safety distance between the microwave antenna and the energized joint of the electrofusion joint of the PE pipeline , the motion speed vector of the microwave antenna is , the distance between the motion direction of the microwave antenna and the energized joint is , Calculate the Euclidean distance between the real-time computing microwave antenna and the energized connector : （1） When the distance between the microwave antenna and the energized connector is less than the safe distance obstacle avoidance is triggered: , start obstacle avoidance: To avoid the energized joint of the electrofusion joint of the PE pipe, the algorithm needs to recalculate the movement direction of the probe and use the formula to adjust the movement direction of the probe to a direction away from the obstacle: First, calculate the unit vector of the energized joint between the microwave antenna and the electrofusion joint of the PE pipe : （2） Next, adjust the direction vector of the probe movement to: （3） Among them, is the obstacle avoidance adjustment coefficient, which is used to control the intensity of the probe's avoidance of obstacles and the degree of direction correction; the adjusted microwave antenna ensures that the microwave antenna moves away from the obstacle; After the microwave antenna avoids an obstacle, the new movement path will be achieved through a re-planned scanning route, and the new probe position will be updated according to the new velocity vector and time Update: t （4） wherein is a tiny time step, and this path adjustment will continue until the distance between the microwave antenna and the energized connector , that is, the probe safely avoids obstacles; The laser locator continuously monitors the distance between the microwave antenna and the energized connector. If the microwave antenna approaches the energized connector and reaches the preset safety distance threshold, an obstacle avoidance operation is triggered. Once the obstacle avoidance is triggered, the movement direction of the microwave antenna will be corrected according to formula (4), that is, by adjusting the velocity vector to avoid the energized connector and re-plan the movement path. During the obstacle avoidance process, the microwave antenna will continue to scan the PE pipeline along the new path.

8. The method according to claim 6, wherein Step 2 further includes image reconstruction. The image reconstruction algorithm is a non-negative matrix factorization inverse scattering imaging algorithm, specifically including: In scanning imaging, it is necessary to perform a step-by-stop scan over the entire scanning plane. This process is regarded as stopping at a position to transmit and receive signals, and then moving to the next position to continue transmitting and receiving signals, so as to obtain target echo data. is the direction of is the height direction, is the range direction. The defect target is processed by transmitting a frequency-stepped signal, and the sampling microwave antenna scans sequentially along the positive x-axis direction. The scanning spatial trajectory forms a plane, that is scanning plane; The purpose of two-dimensional imaging is to obtain the target scattering information to be acquired along equivalent distribution on the plane That is, the target two-dimensional image function, the electric field is located in the x-z plane. The sampling plane during measurement coincides with the plane plane and is parallel to the xoy coordinate plane. The electric field quantity of the target scattering field at each sampling point is expressed as , and the ultimate goal of imaging is to reconstruct the scattering field strength distribution corresponding to the target image plane. The known quantity of the scattering field strength is the one corresponding to z = 0 propagating along the positive z direction distance generation; Given a non - negative matrix , where all elements , the goal of NMF is to find two non - negative matrices: , as the basis matrix, as the coefficient matrix; such that: WH(5) That is, approximate the original matrix by the product of matrix and matrix , where is the dimension of the low-dimensional space, satisfying min ; The goal of NMF is to minimize the difference between the original matrix and the decomposed matrix The commonly used objective function is based on squared error or KL divergence; (6) where denotes the Frobenius norm; The objective function based on KL divergence: （7） The main constraint in NMF is that all matrix elements must be non - negative, i.e.: （8） The update rule for matrix W: （9） The update rule for matrix H: （10） wherein, represents an element-wise product, and the division operation is also element-wise; NMF effectively realizes dimensionality reduction by decomposing a high-dimensional matrix into low-dimensional matrices and ; matrix represents a set of basis vectors or patterns, and matrix represents the representation of data on these basis vectors; The spatial domain spectrum Plane electric field distribution d d (11) Among them, and respectively represent the spatial wave number components corresponding to the spatial coordinates x and y; Research on the spectral theory in near - field azimuth imaging. According to the spectral theory at the plane z = 0: （12） where Z0 is the lift-off distance, is the wave number vector, is the wave frequency, is the circular domain function, expressed as: （13） So the transfer function is: （14） The circular domain function circ restricts the integration region within the circular region of. It can be seen that not all components of the plane wave spectrum contribute to the superposition of the total field. Only the part of the plane spectrum whose propagation direction satisfies the circular domain function circ contributes. From Equation (1-4), it is known that during the process of imaging the target from the scanning plane to z = 0, the spatial wave spectrum is like a two-dimensional band-pass filter. Its amplitude-frequency responses within, on, and outside the circular domain with a radius of the wave number vector are 1, 1 / 2, and 0 respectively. Within the passband, the echo data on the sampling plane z = z0 is known to be , and according to the wave spectrum expansion theory: d d (15) Among them, is the spatial domain spectrum, which is obtained by performing a two-dimensional Fourier transform FT2D on the echo signal on the plane, that is: d （16） Among them, and are the corresponding ranges of the scanning plane; So far, through the two - dimensional spatial spectrum to perform two - dimensional inverse Fourier transform, the algorithm derivation is completed, and the scattering distribution function of the target is obtained. That is, the target imaging expression is: 。 9. The method according to claim 6, wherein The machine learning algorithm in step 3 is the defect detection algorithm for the electro - fusion joint of PE pipes, including: The signal module pre - processes the acquired microwave reflection signal by normalization, including the following steps: Step 1. Slice the pre - processed signal according to a preset slice width and step size to generate multiple signal segments of the same length, ensuring that each generated slice signal has a fixed length; Step 2. Automatically label according to whether it contains the defective part in the hot - melt joint of the polyethylene pipe; Step 3. Extract multi - dimensional features in the time domain and frequency domain for each labeled slice signal and the complete signal, and use a random forest classifier to train the made feature dataset to construct a classification model for defect detection; Among them, in the random forest classifier, the extracted feature dataset is defined according to the feature vector as: （18） Among them, represents the feature vector of the i-th signal segment, and the feature dimensions include time-domain features and frequency-domain features. Each feature vector corresponds to a label , among which, represents the normal region corresponding to the signal segment, indicates that the signal segment contains a defective region.

10. The method according to claim 9, wherein Step 3 also includes: Screening the made feature dataset before the random forest classifier is trained: Step 1. In the feature dataset screening stage, first, screen out the features that have a significant impact on defect classification through the Mann - Whitney U test, and then optimize the feature set in combination with recursive feature elimination, and finally retain the most important ten features; Among them, the Mann - Whitney U test includes: Given a feature, divide it into two groups: the defective sample group and the normal sample group, and calculate the U value: （19） Among them, is the number of defective samples, is the number of normal samples, is the sum of the ranks of the defective sample group. If the distribution of a feature in the two groups is significantly different, it means that the feature has a large discriminatory ability for the classification task, and vice versa. By performing the Mann-Whitney U test on each feature, features with a p-value less than the threshold of 0.01 are selected as important candidate features; Recursive feature elimination algorithm: The feature importance function is defined as: （20） Among them, X i is the feature vector representing the i-th signal segment, is the weight of the classifier decision; Remove the features with smaller weights and repeat the training process until ten of the most important features are retained; and perform feature selection recursively through the following optimization process: （21） where C is the regularization parameter that controls the influence of the weight magnitude during the feature selection process, b is the bias term, the offset of the decision boundary of the classifier, which is used to adjust the relative position of the hyperplane. is the feature. is the label of the i-th sample, is the weight for the classifier's decision.

Images