Method for detecting defects in a bend of a pressure pipeline

Abstract

The application discloses a kind of pressure pipeline elbow defect detection methods, it is related to pipeline inspection detection technical field, it is based on annular transducer array waveguide excitation and reception and fuses deep learning technology, can adapt to the fine, efficient and intelligent development of complex structure area corrosion damage detection.The pressure pipeline elbow defect detection method comprises the following steps: step S1, measures pipeline curvature radius, outer diameter and wall thickness;Step S2, arrange sensor array and collect pipeline defect data;Step S3, calculate the dispersion data of pipeline waveguide;Step S4, iteration inversion pipeline real defect size.

Description

Technical Field

[0001] This invention relates to the field of pipeline inspection and testing technology, and in particular to a method for detecting defects in elbows of pressure pipelines. Background Technology

[0002] Currently, pressure pipelines, as crucial fluid transport devices in industries such as energy and chemicals, play a vital role in the stability of the entire system through their safe operation. Elbows, as key connecting components in pressure pipelines, not only withstand pressure surges from the internal media but also frequently operate in areas of abrupt flow velocity changes and intense turbulence, making them highly susceptible to failure modes such as corrosion, erosion, and stress concentration. Once corrosion and perforation occur at an elbow, it can lead to serious accidents such as media leakage and explosions, resulting in significant economic losses and safety hazards.

[0003] In existing technologies, commonly used methods for detecting corrosion in elbows mainly include ultrasonic thickness measurement, radiographic testing, and magnetic particle testing. Among these, ultrasonic thickness measurement is sensitive to localized corrosion, but it suffers from problems such as coupling difficulties and severe echo signal attenuation when detecting elbows with complex geometries. Radiographic testing, while capable of acquiring two-dimensional projection images, has limited ability to identify minute defects and is also complex to operate and poses radiation safety risks. Magnetic particle and eddy current testing technologies are mostly limited to the identification of surface or near-surface defects, making it difficult to effectively assess buried defects or deep corrosion.

[0004] The inventors of this application have discovered that guided wave testing technology, due to its long propagation distance, wide coverage, and strong adaptability to structurally complex areas, has gradually become an important means of monitoring the integrity of pressure pipelines. Especially in structurally irregular or geometrically complex areas, such as elbows, tees, and weld areas, guided wave testing can effectively supplement the shortcomings of traditional non-destructive testing methods. However, the extraction of defect information in traditional guided wave testing mainly relies on manual experience or simple signal processing methods, making it difficult to accurately identify the complex wave field response and defect characteristics in elbow areas.

[0005] With the development of artificial intelligence, especially deep learning technology, new solutions have been provided for pattern recognition and feature inversion of complex nondestructive testing signals. The inventors of this application have discovered that combining deep learning with guided wave detection can achieve high-precision identification and inversion of defect types, shapes, and corrosion levels by fully mining signal features, thereby significantly improving the automation and intelligence level of corrosion detection. Summary of the Invention

[0006] The purpose of this invention is to provide a method for detecting defects in pressure pipeline elbows, which is based on guided wave excitation and reception of a ring transducer array and integrates deep learning technology, and can adapt to the refined, efficient and intelligent development of corrosion damage detection in complex structural areas.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A method for detecting defects in pressure pipeline elbows includes the following steps: Step S1: Measure the pipe's radius of curvature, outer diameter, and wall thickness; Step S2: Deploy the sensor array and collect pipeline defect data; Step S3: Calculate the guided wave dispersion data of the pipeline; Step S4: Iteratively invert the true defect size of the pipeline; Step S4 specifically includes: Step S41, Two-dimensional unfolding: Unfold along the outer contour line of the pipe bend to obtain a length of Width The matrix is ​​rectangular, and the two ring transducer arrays are positioned on the two longer sides of the matrix; where, The bending angle of the pipe is calculated based on the lengths of the three sides of the triangle measured in step S1; r is the mean diameter of the pipe. Given the outer diameter r1 and wall thickness d, the mean diameter r is obtained as follows: ; Step S42, Mesh division: Divide the rectangular area into a mesh according to the required measurement accuracy. The rectangle obtained in step S1 can be divided into square meshes with a length and width of 2mm. Step S43, Parameter Settings: Elbow detection is achieved by imbuing the material with anisotropic characteristics to realize two-dimensional modeling of the pipeline. The anisotropy of the pipeline elbow is reflected in the circumferential direction of the pipeline; the parameters to be set are: Among them, density Based on the original density of the pipe; speed of sound Based on the calculated phase velocity dispersion data, set the anisotropy parameters. The quality factor is calculated according to formulas (6)-(9). Set to infinity without considering sound wave attenuation; (6) (7) (8) (9) in, This represents the circumferential angle value of the pipe, which is the change in length / circumferential direction of the pipe after it is unfolded in two dimensions; Step S44, Forward Wave Field Simulation: The forward simulation uses the finite difference method to solve the pseudo-acoustic anisotropy equation; based on the actual position of the transducer array installed on the pipe bend, ultrasonic signals are emitted and received at the corresponding positions in the two-dimensional model; Step S45, Deep Learning Network Iterative Defect Inversion: Step S451, Data Acquisition and Preparation: Generate pipe models with different defect sizes using a random script, and generate forward wave fields according to the pressure pipe elbow modeling scheme in Steps S41-S44; each set of array data corresponds to a two-dimensional model diagram with corrosion defects; Step S452, Model Training: Train the prepared dataset using a 1D-CNN model; Step S453, Defect Inversion: Input the actual collected data into the trained model to finally obtain a model diagram with defects.

[0008] In practical applications, step S1 specifically includes: Step S11: Select three points at equal intervals on the curved part of the pipe. Let the three measurement points be A, B, and C. Step S12: Measure the lengths of segments AB, BC, and AC with a measuring tape, then wrap the measuring tape around the pipe to obtain the pipe circumference L. The radius of curvature R and the outer diameter r1 of the pipe can be calculated according to formulas (1)-(4). Step S13: Use a thickness gauge to measure the pipe wall thickness to obtain the pipe wall thickness d; set up Pipe circumference ; (1) (2) (3) (4).

[0009] Further, step S2 specifically includes: Step S21: Arrange annular transducer arrays on both sides of the pipe bend, with each annular transducer array arranged at equal intervals along the circumference of the pipe. Step S22, the data acquisition process is as follows: Assume two ring arrays, namely array A and array B. The signal generation module generates a signal, which is amplified by the power amplification module and then excites one transducer of array A to emit an ultrasonic signal. All transducers of array B receive this ultrasonic signal. The data is collected by the data acquisition card and stored in the PC. Then, the next transducer of array A emits an ultrasonic signal, and all transducers of array B receive this signal, and so on, until all transducers of array A emit an ultrasonic signal once.

[0010] Furthermore, step S3 specifically includes: Step S31: Based on the pipe thickness, material and pipe diameter, calculate the pipe dispersion data using formula (5); (5) in, Let r1 be the density, r1 be the outer diameter of the pipe, and d be the pipe wall thickness. These refer to the materials of the pipes.

[0011] Compared with existing technologies, the pressure pipeline elbow defect detection method of the present invention has the following advantages: The pressure pipeline elbow defect detection method provided by this invention uses a pseudo-acoustic anisotropy equation to approximate the propagation process of guided waves in the elbow, which greatly reduces the complexity of three-dimensional elastic wave forward modeling and thus effectively improves forward modeling efficiency. At the same time, by combining a deep learning network model to reconstruct the defect shape and training the network with massive amounts of data, high-precision detection of pipeline defects can be effectively achieved. Attached Figure Description

[0012] Figure 1 A schematic flowchart of the pressure pipeline elbow defect detection method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the transducer array data acquisition device in the pressure pipeline elbow defect detection method provided in this embodiment of the invention; Figure 3 This is a schematic diagram of the phase velocity dispersion curve of the L(0,1) mode guided wave in the pressure pipeline in the pressure pipeline elbow defect detection method provided in the embodiment of the present invention. Figures 4(a)-4(e) are schematic diagrams of parameter modeling in the forward simulation process of the pressure pipeline elbow defect detection method provided in the embodiments of the present invention; Figure 5 This is a schematic diagram of forward wave field simulation in the pressure pipeline elbow defect detection method provided in this embodiment of the invention; Figure 6 This is a schematic diagram of the deep learning network model in the pressure pipeline elbow defect detection method provided in this embodiment of the invention; Figure 7 This is a schematic diagram of the reconstruction results in the pressure pipeline elbow defect detection method provided in the embodiment of the present invention. Detailed Implementation

[0013] For ease of understanding, the method for detecting defects in pressure pipeline elbows provided by the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0014] This invention provides a method for detecting defects in pressure pipeline elbows, such as... Figures 1-7 As shown, it includes the following steps: Step S1: Measure the pipe's radius of curvature, outer diameter, and wall thickness; Step S2: Deploy the sensor array and collect pipeline defect data; Step S3: Calculate the guided wave dispersion data of the pipeline; Step S4: Iteratively invert the true defect size of the pipeline; The above step S4 specifically includes: Step S41, Two-dimensional unfolding: Unfold along the outer contour line of the pipe bend to obtain a length of Width The matrix is ​​rectangular, and the two ring transducer arrays are positioned on the two longer sides of the matrix; where, The bending angle of the pipe is calculated based on the lengths of the three sides of the triangle measured in step S1; r is the mean diameter of the pipe. Given the outer diameter r1 and wall thickness d, the mean diameter r is obtained as follows: ; Step S42, Mesh division: Divide the rectangular area into a mesh according to the required measurement accuracy. The rectangle obtained in step S1 can be divided into square meshes with a length and width of 2mm. Step S43, Parameter Settings: Elbow detection is achieved by imbuing the material with anisotropic characteristics to realize two-dimensional modeling of the pipeline. The anisotropy of the pipeline elbow is reflected in the circumferential direction of the pipeline; the parameters to be set are: Among them, density Based on the original density of the pipe; speed of sound Based on the calculated phase velocity dispersion data, set the anisotropy parameters. The quality factor is calculated according to formulas (6)-(9). Set to infinity without considering sound wave attenuation; (6) (7) (8) (9) in, This represents the circumferential angle value of the pipe, which is the change in length / circumferential direction of the pipe after it is unfolded in two dimensions; Step S44, Forward Wave Field Simulation: The forward simulation uses the finite difference method to solve the pseudo-acoustic anisotropy equation; based on the actual position of the transducer array installed on the pipe bend, ultrasonic signals are emitted and received at the corresponding positions in the two-dimensional model; Step S45, Deep Learning Network Iterative Defect Inversion: Step S451, Data Acquisition and Preparation: Use a random script to generate pipe models with different defect sizes, and generate forward wave fields according to the pressure pipe elbow modeling scheme in Steps S41-S44 above; each set of array data corresponds to a two-dimensional model diagram with corrosion defects; Step S452, Model Training: Use a 1D-CNN model to train the prepared dataset; Step S453, Defect Inversion: Input the actual collected data into the trained model to finally obtain a model diagram with defects.

[0015] Compared with existing technologies, the pressure pipeline elbow defect detection method described in this embodiment of the invention has the following advantages: In the pressure pipeline elbow defect detection method provided by the present invention, the propagation process of guided waves in the elbow is approximated by the pseudo-acoustic anisotropy equation, which greatly reduces the complexity of three-dimensional elastic wave forward modeling and thus effectively improves the forward modeling efficiency. At the same time, the defect shape is reconstructed by combining a deep learning network model and the network is trained with massive amounts of data, which can effectively achieve high-precision detection of pipeline defects.

[0016] In practical applications, step S1 above may specifically include: Step S11: Select three points at equal intervals on the curved part of the pipe. Let the three measurement points be A, B, and C. Step S12: Measure the lengths of segments AB, BC, and AC with a measuring tape, then wrap the measuring tape around the pipe to obtain the pipe circumference L. The radius of curvature R and the outer diameter r1 of the pipe can be calculated according to formulas (1)-(4). Step S13: Use a thickness gauge to measure the pipe wall thickness to obtain the pipe wall thickness d; set up Pipe circumference ; (1) (2) (3) (4).

[0017] Furthermore, step S2 above may specifically include: Step S21: Arrange annular transducer arrays on both sides of the pipe bend, with each annular transducer array arranged at equal intervals along the circumference of the pipe. Step S22, the data acquisition process is as follows: Assume two ring arrays, namely array A and array B. The signal generation module generates a signal, which is amplified by the power amplification module and then excites one transducer of array A to emit an ultrasonic signal. All transducers of array B receive this ultrasonic signal. The data is collected by the data acquisition card and stored in the PC. Then, the next transducer of array A emits an ultrasonic signal, and all transducers of array B receive this signal, and so on, until all transducers of array A emit an ultrasonic signal once.

[0018] Furthermore, step S3 above may specifically include: Step S31: Based on the pipe thickness, material and pipe diameter, calculate the pipe dispersion data using formula (5); (5) in, Let r1 be the density, r1 be the outer diameter of the pipe, and d be the pipe wall thickness. These refer to the materials of the pipes. Specific Implementation Example 1 With a 90° bend ( Steel pipe (density ρ: 7932 kg / m³) 3 Taking the following as an example: Young's modulus: 216.9 GPa, Poisson's ratio: 0.2865, pipe inner diameter: r1=0.1015 m, pipe outer diameter: r2=0.1095 m, pipe middle diameter: 0.1055 m, radius of curvature: R=0.333 m.

[0020] Figure 2 This is a schematic diagram of a data acquisition device for a pressure pipeline bend. Annular transducer arrays (with a center frequency of 40kHz) are arranged on both sides of the pipeline bend. Each annular array has 30 transducers, and they are arranged at equal intervals along the circumference of the pipeline. The data collection process is as follows: Two ring arrays are designated as array A (transmitter array) and array B (receiver array). A signal is generated by a signal generation module, amplified by a power amplifier module, and then used to excite one transducer in array A to emit an ultrasonic signal. All transducers in array B receive this ultrasonic signal. The data is collected by a data acquisition card and stored in a PC. Subsequently, the next transducer in array A emits an ultrasonic signal, and all transducers in array B receive this signal, and so on, until all transducers in array A have emitted an ultrasonic signal once, resulting in a total of 30×30 waveform data.

[0021] After data acquisition, it is necessary to calculate the guided wave dispersion data of the current pipeline. Given the inner and outer diameters of the pipeline and its material parameters, the dispersion curve of the guided wave in the pipeline can be calculated, such as... Figure 3As shown; since lower-order modes are easier to excite in the pipeline, only the L(0,1) mode in the pipeline is calculated here.

[0022] After obtaining the dispersion data of the pipeline, it is necessary to perform forward modeling simulation on the propagation of guided waves in the pipeline bend. The efficiency of forward modeling simulation directly determines the speed of defect reconstruction. Here, the three-dimensional pipeline structure is simplified to two-dimensional structure modeling to improve the efficiency of forward modeling simulation. That is, the subsequent step is to iteratively invert the actual defect size of the pipeline, and the specific steps are as follows: (1) Two-dimensional unfolding: unfolding along the outer contour line of the pipe bend, a length of [length missing] can be obtained. Width The rectangle has two ring transducer arrays positioned on the top and bottom sides of the matrix. Indicates the bending angle of the pipe (here it is) ); (2) Mesh division: Divide the mesh according to the required measurement accuracy. Generally, the rectangle obtained in step (1) above can be divided into square meshes with a length and width of 2mm. (3) Parameter settings: Unlike straight pipes, the detection of elbows is mainly achieved by giving the material anisotropic characteristics to realize the two-dimensional model of the pipe. The anisotropy of the pipe elbow is mainly reflected in the circumferential direction of the pipe. The main parameters to be set are: ρ, ν, ε, δ, ϱ. Among them, the density ρ can be set according to the original density of the pipe. The initial model of the iterative inversion is a uniform defect-free model. In addition, the L(0,1) mode wave with a frequency of 40kHz is selected as the detection wave. The anisotropic parameters ε and δ are calculated according to formulas (6)-(9). The quality factor ϱ can be set to 1000 without considering the attenuation of the sound wave. The modeling results of the five parameters are shown in Figures 4(a)-4(e). (4) Forward wave field simulation: Forward simulation uses finite difference to solve the anisotropic equation of pseudo-acoustic waves; based on the actual position of the transducer array installed on the pipe bend, ultrasonic signals are emitted and received at the corresponding positions of the two-dimensional model; the initial model is a uniform and defect-free model, and the wave field simulation results are as follows. Figure 5 As shown, the lower circle represents the transmitting transducer, and the upper triangle represents the receiving transducer. This wave field is the 40kHz single-frequency wave field obtained when the first transmitting transducer in the lower left corner transmits a signal. (5) Defects of iterative inversion in deep learning networks: A schematic diagram of the network model is shown below. Figure 6 As shown, the actual acquired wave field is subjected to a Fourier transform (fft) to obtain the 40kHz signals at all receiving transducers, which are then input to... Figure 6 In the trained network shown, the final result is... Figure 7 The reconstruction result shown indicates that the central black area represents the region where defects exist, corresponding to... Figure 2 The area where the defect is located.

[0023] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for detecting defects in pressure pipeline elbows, characterized in that, Includes the following steps: Step S1: Measure the pipe's radius of curvature, outer diameter, and wall thickness; Step S2: Deploy the sensor array and collect pipeline defect data; Step S3: Calculate the guided wave dispersion data of the pipeline; Step S4: Iteratively invert the true defect size of the pipeline; Step S4 specifically includes: Step S41, Two-dimensional unfolding: Unfold along the outer contour line of the pipe bend to obtain a length of Width The matrix is ​​rectangular, and the two ring transducer arrays are positioned on the two longer sides of the matrix; where, The bending angle of the pipe is calculated based on the lengths of the three sides of the triangle measured in step S1; r is the mean diameter of the pipe. Given the outer diameter r1 and wall thickness d, the mean diameter r is obtained as follows: ; Step S42, Mesh division: Divide the rectangular area into a mesh according to the required measurement accuracy. The rectangle obtained in step S1 can be divided into square meshes with a length and width of 2mm. Step S43, Parameter Settings: Elbow detection is achieved by imbuing the material with anisotropic characteristics to realize two-dimensional modeling of the pipeline. The anisotropy of the pipeline elbow is reflected in the circumferential direction of the pipeline; the parameters to be set are: Among them, density Based on the original density of the pipe; speed of sound Based on the calculated phase velocity dispersion data, set the anisotropy parameters. The quality factor is calculated according to formulas (6)-(9). Set to infinity without considering sound wave attenuation; (6) (7) (8) (9) in, This represents the circumferential angle value of the pipe, which is the change in length / circumferential direction of the pipe after it is unfolded in two dimensions; Step S44, Forward Wave Field Simulation: The forward simulation uses the finite difference method to solve the pseudo-acoustic anisotropy equation; based on the actual position of the transducer array installed on the pipe bend, ultrasonic signals are emitted and received at the corresponding positions in the two-dimensional model; Step S45, Deep Learning Network Iterative Defect Inversion: Step S451, Data Acquisition and Preparation: Generate pipe models with different defect sizes using a random script, and generate forward wave fields according to the pressure pipe elbow modeling scheme in Steps S41-S44; each set of array data corresponds to a two-dimensional model diagram with corrosion defects; Step S452, Model Training: Train the prepared dataset using a 1D-CNN model; Step S453, Defect Inversion: Input the actual collected data into the trained model to finally obtain a model diagram with defects.

2. The method for detecting defects in pressure pipeline elbows according to claim 1, characterized in that, Step S1 specifically includes: Step S11: Select three points at equal intervals on the curved part of the pipe. Let the three measurement points be A, B, and C. Step S12: Measure the lengths of segments AB, BC, and AC with a measuring tape, then wrap the measuring tape around the pipe to obtain the pipe circumference L. The radius of curvature R and the outer diameter r1 of the pipe can be calculated according to formulas (1)-(4). Step S13: Use a thickness gauge to measure the pipe wall thickness to obtain the pipe wall thickness d; set up Pipe circumference ; (1) (2) (3) (4)。 3. The method for detecting defects in pressure pipeline elbows according to claim 2, characterized in that, Step S2 specifically includes: Step S21: Arrange annular transducer arrays on both sides of the pipe bend, with each annular transducer array arranged at equal intervals along the circumference of the pipe. Step S22, the data acquisition process is as follows: Assume two ring arrays, namely array A and array B. The signal generation module generates a signal, which is amplified by the power amplification module and then excites one transducer of array A to emit an ultrasonic signal. All transducers of array B receive this ultrasonic signal. The data is collected by the data acquisition card and stored in the PC. Then, the next transducer of array A emits an ultrasonic signal, and all transducers of array B receive this signal, and so on, until all transducers of array A emit an ultrasonic signal once.

4. The method for detecting defects in pressure pipeline elbows according to claim 3, characterized in that, Step S3 specifically includes: Step S31: Based on the pipe thickness, material and pipe diameter, calculate the pipe dispersion data using formula (5); (5) in, Let r1 be the density, r1 be the outer diameter of the pipe, and d be the pipe wall thickness. These refer to the materials of the pipes.

Images