Multi-modal fusion pipeline damage positioning method

Abstract

The invention relates to a multi-modal fusion pipeline damage positioning method, which comprises the following steps of: uniformly arranging piezoelectric transducer arrays along the circumferential direction of a pipeline, generating non-frequency dispersion L (0, 2) and T (0, 1) guided waves respectively excited by 70kHz sine pulses as excitation signals, and unidirectionally spreading the excitation signals along the pipeline; the method comprises the following steps: acquiring reflection echoes at a to-be-measured section of a pipeline, and performing matched filtering, omega-k domain filtering and circumferential modal decomposition on the reflection echoes to obtain two modal components in an L (0, 2) mode and a T (0, 1) mode; performing back propagation and focusing imaging on each modal component to obtain imaging results under two modals L (0, 2) and T (0, 1); and fusing the imaging results in the two modals to obtain a multi-modal fusion pipeline damage image. Compared with the prior art, the method has the advantages that the problem that sensitivity of different defect types is difficult to consider due to single-mode optimization and the problems of signal waveform distortion and energy dispersion existing in multi-mode signals are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the fields of high-end equipment manufacturing and new energy technology, and particularly relates to a pipeline damage positioning method based on multi-modal fusion. BACKGROUND

[0002] As the core carrier for energy transportation such as oil and gas and chemical industry, the safe operation of pipeline is directly related to the national economic and ecological safety. However, during the long-term service, the pipeline is prone to defects such as cracks, holes and thinning defects due to factors such as material aging, corrosion, mechanical damage and human damage. According to statistics, more than 70% of pipeline failure accidents are caused by micro-crack propagation, and the traditional detection methods (such as body wave ultrasonic and eddy current detection) have significant shortcomings in coverage, detection efficiency and adaptability to complex working conditions. Specifically, (1) it is difficult to realize long-distance rapid screening due to the dependence on dense sensor arrangement; (2) the detection capability for hidden areas such as buried pipelines and curved pipe support sections is limited; (3) limited by linear acoustic theory, it is difficult to effectively identify early micro-cracks and other nonlinear damages.

[0003] In recent years, ultrasonic guided wave technology has become a research hotspot for pipeline nondestructive testing due to its characteristics of "low attenuation and long-distance propagation". It excites elastic waves propagating along the pipeline axis, and realizes damage positioning by using the reflection and scattering signals caused by defects. However, this technology still faces the following bottlenecks in practical application:

[0004] (1) Multi-modal and dispersion effect: there are multiple modes such as longitudinal (L), torsional (T) and flexural (F) during the propagation of guided waves in the pipeline, and the phase velocity changes with frequency (dispersion phenomenon), resulting in signal waveform distortion and energy dispersion.

[0005] (2) Limitations of single modal detection: existing researches mainly focus on the optimization of single guided wave mode (such as L(0,2) or T(0,1)), but the actual pipeline damage forms are diverse (such as circumferential cracks and axial thinning), and single mode is difficult to consider the sensitivity of different defect types. SUMMARY

[0006] The purpose of the present application is to overcome the problem of difficulty in considering the sensitivity of different defect types caused by single modal optimization and the problems of signal waveform distortion and energy dispersion existing in multi-modal signals, and to provide a pipeline damage positioning method based on multi-modal fusion.

[0007] A pipeline damage positioning method based on multi-modal fusion, the method comprising the following steps:

[0008] A piezoelectric transducer array is arranged uniformly along the circumference of the pipeline to generate 70 kHz sinusoidal pulses to excite non-dispersive L(0,2) and T(0,1) two guided wave modes as excitation signals, and the excitation signals propagate unidirectionally along the pipeline.

[0009] acquiring reflected echoes in a to-be-tested section of the pipeline, performing matched filtering, omega-k domain filtering and circumferential mode decomposition on the reflected echoes, and obtaining two mode components under L(0, 2) and T(0, 1) modes through the decomposition;

[0010] performing reverse propagation and focused imaging on each mode component to obtain imaging results under L(0, 2) and T(0, 1) modes;

[0011] fusing the imaging results under the two modes to obtain a multi-modal fusion pipeline damage image, and performing threshold segmentation on the image to obtain a pipeline defect type and coordinates of the corresponding defect.

[0012] Further, the multi-modal fusion pipeline damage image is:

[0013]

[0014] wherein, is a linear domain of the imaging result under the L(0, 2) mode, is a linear domain of the imaging result under the T(0, 1) mode, M (X, Y) is a fused imaging result, ω L and ω T are weights of the L(0, 2) mode and the T(0, 1) mode respectively.

[0015] Further, the linear domain of the imaging result under the L(0, 2) mode and the linear domain of the imaging result under the T(0, 1) mode are:

[0016]

[0017] wherein, and are the imaging results under the L(0, 2) and T(0, 1) modes respectively.

[0018] Further, specific steps of performing matched filtering, omega-k domain filtering and circumferential mode decomposition on the reflected echoes are:

[0019] performing frequency domain cross-correlation operation on the reflected echoes by using a matched filter, performing FFT on the obtained signal along a time axis and a piezoelectric transducer array space axis respectively and filtering to obtain an omega-k domain signal, decomposing the omega-k domain signal into different circumferential mode orders and setting a cutoff order to filter out high-order modes to obtain two mode components under L(0, 2) and T(0, 1) modes.

[0020] Further, the matched filter is a conjugate frequency domain response of an excitation signal.

[0021] Further, the specific steps of performing FFT along the time axis and the spatial axis of the piezoelectric transducer array respectively and filtering are as follows:

[0022] FFT is performed on the obtained signal along the time axis and the spatial axis of the piezoelectric transducer respectively, and the obtained FFT signal is reserved by a band-pass filter to obtain a signal in the ω-k domain, wherein the signal has a frequency band centered at f0 and a bandwidth of 35%, and f0 is the center frequency of the excitation signal.

[0023] Further, the specific steps of obtaining the imaging results in the L(0,2) and T(0,1) modes are as follows:

[0024] The imaging results in the L(0,2) and T(0,1) modes are obtained by reconstruction and superposition based on the modal amplitude coefficients and the phase factors of the two modal components in the L(0,2) and T(0,1) modes.

[0025] Further, the imaging results in the L(0,2) and T(0,1) modes are obtained by reconstruction and superposition based on the filtered low-order modal amplitude coefficients A(n,ω) and the phase velocity c p , the group velocity c g of the guided wave. Wherein z R is the position of the receiving sensor.

[0026] Further, the sine pulse is a signal modulated by a Hanning window.

[0027] Further, the two groups of adjacent piezoelectric transducers of the piezoelectric transducer array are separated by (n 1 / 2)λ, and the phase difference is π.

[0028] The object of the present application can be achieved by the following technical solutions:

[0029] Compared with the prior art, the present application has the following beneficial effects:

[0030] The multi-modal image fusion of the application can effectively utilize the complementary relationship between different modal guided waves, effectively avoid the problem that a single mode is difficult to detect a specific structural defect, the application selects a specific combination of L(0, 2) and T(0, 1) modes, the characteristics of the longitudinal axisymmetric mode L(0, 2) are the fastest speed, uniform axial displacement field, sensitive to defects of any depth, less radial energy leakage, suitable for long-distance flaw detection and thinning defect detection; the characteristics of the torsional axisymmetric mode T(0, 1) are significant circumferential displacement, especially sensitive to axial cracks, strong anti-interference ability, excellent non-dispersion performance, suitable for detecting small through cracks, under 70 kHz excitation, the group velocity change rates of both are less than 0.5%, avoiding signal distortion during fusion, L(0, 2) is sensitive to circumferential cracks and thinning defects, T(0, 1) is sensitive to axial cracks and small size defects, the two are orthogonal complementary in defect type sensitivity and size adaptability, after fusion, the characteristics of all types of damage such as circumferential cracks, axial cracks and thinning defects can be covered, the problem that a single mode optimization cannot simultaneously consider the sensitivity of different defect types is overcome, and the signal-to-noise ratio of the imaging result is effectively improved, and the defect positioning result is more obvious. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 It is a flowchart of the application;

[0032] Figure 2 It is a finite element simulation flowchart;

[0033] Figure 3 It is a schematic diagram of a through-wall defect finite element model;

[0034] Figure 4 It is a schematic diagram of a thinning defect finite element model;

[0035] Figure 5 It is a schematic diagram of the relationship between the pipe shape and the cutoff frequency, wherein Figure 5 a is the relationship between the pipe radius and the L(0, 2) cutoff frequency, Figure 5 b is the relationship between the pipe wall thickness and the L(0, 2) cutoff frequency, Figure 5 c is the relationship between the pipe wall thickness and the L(0, 3) cutoff frequency;

[0036] Figure 6 It is a fusion imaging algorithm flowchart;

[0037] Figure 7 It is an L(0, 2) mode imaging schematic diagram;

[0038] Figure 8 It is a T(0, 1) mode imaging schematic diagram;

[0039] Figure 9 It is a multi-modal fusion imaging schematic diagram. DETAILED DESCRIPTION

[0040] The application will be described in detail below in combination with the drawings and specific embodiments. The embodiments are implemented on the premise of the technical scheme of the application, and detailed implementation modes and specific operation processes are given, but the protection scope of the application is not limited to the following embodiments.

[0041] The application provides a pipeline damage positioning method based on multi-modal fusion, and a flow chart is shown in the figure. Figure 1 The method comprises the following steps:

[0042] A piezoelectric transducer array is uniformly arranged along the circumference of the pipeline to generate 70 kHz sinusoidal pulses to excite non-dispersive L(0,2) and T(0,1) two guided wave modes as excitation signals, and the excitation signals propagate unidirectionally along the pipeline.

[0043] Reflection echoes are acquired at the to-be-detected section of the pipeline, and the reflection echoes are subjected to matched filtering, ω-k domain filtering and circumferential mode decomposition, and two mode components under the L(0,2) and T(0,1) modes are obtained through decomposition.

[0044] Reverse propagation and focused imaging are performed on each mode component to obtain imaging results under the L(0,2) and T(0,1) modes.

[0045] The imaging results under the two modes are fused to obtain a multi-modal fusion pipeline damage image.

[0046] The application aims to solve the problems of limited coverage, insufficient resolution and serious multi-modal interference in traditional single-mode ultrasonic guided wave detection. The method first uniformly arranges a piezoelectric transducer array along the circumference of the pipeline, uses a 10-period 70 kHz sinusoidal pulse modulated by a Hanning window to excite non-dispersive L(0,2) and T(0,1) two guided wave modes, and realizes unidirectional propagation through a(n+v)λ phase difference layout between two groups of elements; then reflection echoes of N receiving nodes are collected at the to-be-detected section, and matched filtering, ω-k domain band-pass filtering and circumferential mode decomposition are sequentially performed to extract low-order dominant mode signals; then the synthetic aperture focusing technology (SAFT) is applied to the L(0,2) and T(0,1) signals to generate two high-resolution single-mode imaging images; finally, the multi-modal fusion imaging image is obtained by proportionally weighting the imaging energy of each mode in the same linear domain, and the defect focus point is extracted from the imaging image, and the normalized axial relative distance and circumferential relative angle coordinates are output. Finite element simulation and experimental verification show that the method has high sensitivity to through cracks and thinning defects, the axial positioning error is ≤3 cm, the circumferential angle error is ≤2°, the signal-to-noise ratio is significantly improved, and the artifacts are effectively suppressed. The technology takes into account the sensitivity of different defect types, and the relative coordinate label ensures the universality of the data under different pipe diameters and detection areas, and can be widely applied to online intelligent monitoring and health assessment of long-distance pipelines such as oil and natural gas.

[0047] During simulation experiment, the L(0, 2) and T(0, 1) two non-dispersive guided wave modes are excited, the guided wave is excited by a 10-period sine pulse signal modulated by a Hanning window, the center frequency is 70 kHz; the signal is transmitted by a piezoelectric transducer array arranged uniformly along the circumference of the pipeline and with a spacing less than 0.4λ (λ is the wavelength of the excited mode), so as to realize one-way propagation of a single mode in the pipeline;

[0048] Based on finite element simulation analysis of the propagation characteristics of the two guided wave modes, the pipeline model parameters (outer diameter 200 mm, wall thickness 6 mm, length 2000 mm) and sensor arrangement are optimized, and the grid size is set to be ≥λ / 8 and the time step is ≤wavelength / (phase velocity);

[0049] N receiving nodes are arranged equidistantly along the circumference of the pipeline to be measured, and the reflected echo signals of the L(0, 2) and T(0, 1) modes are collected respectively;

[0050] The received signals of each channel are respectively subjected to matched filtering and ω-k domain filtering, circumferential mode decomposition, and the L(0, 2) and T(0, 1) two modes are independently imaged by using synthetic aperture focusing technology (SAFT);

[0051] The imaging results of the L(0, 2) mode and the imaging results of the T(0, 1) mode are weighted and fused in the same linear domain to obtain a multi-modal fusion imaging graph;

[0052] The defect focusing position is extracted in the fusion graph to obtain the axial distance and the circumferential angle of the defect relative to the excitation point;

[0053] The axial distance is standardized as a relative value (defect axial distance / pipeline length), and the circumferential angle is standardized as a relative value (defect circumferential angle / 360°), and the relative coordinates of the pipeline defect are output.

[0054] The one-way propagation excitation is realized by two groups of transducers separated by (n+1 / 2)λ and with a phase difference of π, so as to enhance the forward guided wave and suppress the reverse guided wave; wherein n is an integer and λ is the wavelength of the excited mode. 1

[0055] The finite element simulation is carried out in the COMSOL environment, the simulation model is set to have an increasing damping absorption boundary to suppress the boundary echo, the total length of the absorption layer is ≥wavelength, and the Rayleigh damping coefficient is increased by m 2 .

[0056] The matched filter is the conjugate frequency domain response of the excitation signal, and the band-pass filtering bandwidth is ±35% of the center frequency, so as to extract the defect scattering component and suppress noise.

[0057] ​The circumferential modal decomposition is based on the annular array geometric parameters, and the cutoff order is set as N / 2, so as to retain low-order sensitive modes and filter out high-order modes.

[0058] The fusion weighting coefficient is automatically distributed according to the linear energy proportion of each modal imaging, so as to suppress the artifact and improve the signal-to-noise ratio in the fusion result, so that the axial positioning error is less than or equal to 3cm, and the circumferential angle error is less than or equal to 2°.

[0059] The introduction of the relative coordinate label can realize data universality under different pipe diameters and detection areas: when the relative labels are the same, the received signal characteristics of the same type of position under different models are highly similar and can be used interchangeably.

[0060] The present application also includes: threshold segmentation of the fusion image, identification and labeling of a plurality of defect positions; and output of a detection report containing defect types and relative coordinates.

[0061] The present application can also propose a pipeline damage detection system, applied to online monitoring of new energy conveying pipelines, comprising:

[0062] The excitation module generates L(0,2) and T(0,1) modal ultrasonic guided waves;

[0063] The receiving module comprises N piezoelectric sensor nodes arranged circumferentially along the pipeline, for collecting reflected echoes;

[0064] The processing module performs SAFT imaging and multi-modal fusion on the collected signals, and outputs defect relative coordinates and a detection report.

[0065] The specific steps of the present application are:

[0066] 1. Dual-mode selection and excitation

[0067] Mode selection: axial symmetric longitudinal L(0,2) mode and axial symmetric torsional T(0,1) mode are adopted, both of which exhibit excellent non-dispersive characteristics in the range of 70kHz, suitable for long-distance detection

[0068] Excitation design: a 10-period sine pulse signal modulated by a Hanning window is adopted, with a center frequency of 70kHz; in order to realize one-way propagation, the sensor array is arranged in two groups, with a group spacing of (n+ 1 / 2)λ and a phase difference of π, which can enhance forward guided waves and suppress reverse guided waves.

[0069] 2. Finite element simulation optimization

[0070] Model parameters: the outer diameter of the straight pipe section is 200mm, the wall thickness is 6mm, and the length is 2000mm; the material is cast steel (ρ=7850kg / m 3 , E=200GPa, ν=0.3)

[0071] Mesh and time step: mesh size ≤ λ / 8, time step ≤ 1 μs, to ensure high-fidelity simulation of guided waves; model boundary is set with a damping increasing absorption layer to eliminate boundary echoes

[0072] Response characteristic analysis: simulation respectively studies the change of reflection coefficient of L(0,2) and T(0,1) to through defects and thinning defects, reveals the sensitivity difference of two modes to the circumferential length, axial length and depth of defects, and verifies that T(0,1) can detect smaller defects and L(0,2) has longer propagation ability.

[0073] 3. Synthetic aperture focused imaging (SAFT)

[0074] Signal preprocessing: the received echoes are first matched filtered (based on the conjugate frequency domain response of the excitation signal), and then band-pass filtered (bandwidth ± 35%) in the ω-k domain to improve the signal-to-noise ratio.

[0075] Modal decomposition: the frequency-wavenumber domain signal is decomposed into different circumferential modes using the annular array geometry, and the low-order (≤ N / 2) sensitive mode components are retained, and the high-order noise is filtered out.

[0076] Imaging implementation: SAFT delay focusing processing is applied to L(0,2) and T(0,1) modes respectively, and two single-mode imaging images are generated.

[0077] 4. Multi-modal image fusion

[0078] The two single-mode imaging results are weighted and fused in the same linear domain, and the fusion weights of each mode are automatically allocated according to the imaging energy of each mode, taking into account the resolution and artifact suppression, to ensure that the axial error of the fused image is ≤ 3 cm and the circumferential error is ≤ 2°.

[0079] The present application has the following beneficial effects

[0080] 1) Multi-modal image fusion can effectively utilize the complementary relationship between different modal guided waves, effectively avoiding the problem that a single mode is difficult to detect a specific structural defect. The combination of L(0,2) and T(0,1) modes in the present application is irreplaceable compared to other dual-mode schemes:

[0081] 1. Frequency dispersion characteristic matching: under 70 kHz excitation, the group velocity change rate of both is less than 0.5%, avoiding signal distortion during fusion;

[0082] 2. Defect sensitivity orthogonality: L(0,2) is sensitive to circumferential cracks and thinning defects, T(0,1) is sensitive to axial cracks and small size defects, and the two are orthogonal and complementary in defect type sensitivity and size adaptability, and can cover all types of damage such as circumferential cracks, axial cracks and thinning defects after fusion.

[0083] 3. Engineering friendliness: changing the sensor direction machine delay can realize dual-mode one-way excitation without increasing hardware complexity.

[0084] 2) The multi-modal fusion imaging algorithm can identify multiple damage locations in the pipeline, and has high positioning accuracy and resolution. The axial error is not more than 3 cm, and the circumferential angle error is not more than 2°, while effectively reducing noise interference and avoiding false images.

[0085] 3) Multi-modal fusion imaging can effectively improve the signal-to-noise ratio of the imaging result, and the defect positioning result is more obvious.

[0086] 4) The fusion method has good adaptability to complex working conditions such as buried, high temperature and high pressure, and can be popularized to online intelligent monitoring of oil and gas long-distance pipelines.

[0087] 5) The present application can be integrated into an intelligent pipeline inspection robot or an online monitoring system, promoting the intelligent upgrading of the high-end equipment manufacturing field;

[0088] 6) By improving the pipeline defect detection efficiency and reducing the leakage risk of new energy transmission pipelines, it helps the sustainable development of the new energy industry.

[0089] Through multi-modal data fusion and automatic defect classification, real-time evaluation of the health status of the pipeline is realized, providing core technical support for predictive maintenance in intelligent manufacturing.

[0090] The following is a practical simulation experiment:

[0091] 1. Finite element model construction

[0092] The process of ultrasonic guided wave analysis simulation is as shown in Figure 2 .

[0093] (1) Model geometric parameters and material settings:

[0094] In the present application, considering the common physical pipeline size and the calculation time required by the simulation model, the model pipeline geometry is determined as a straight pipeline with an outer diameter of 200 mm, a wall thickness of 6 mm, and a length of 2000 mm. The pipeline material is cast steel, and its main parameters are shown in Table 1:

[0095] Table 1 Model pipeline material parameters

[0096]

[0097] (2) Grid division parameters and time parameters:

[0098] The grid size and time parameters of the model are mainly affected by the guided wave frequency, wavelength and model geometry. The following is the determination process of the main parameters.

[0099] 1) Mesh size: In order to accurately simulate the ultrasonic wave, at least 8 elements per wavelength are needed when using linear mesh

[0100] 2) Time step: In order to obtain sufficient calculation accuracy when using the finite element method to calculate the model of the defect, the distance passed by the modal guided wave in a time step should not exceed the length of a single element grid,

[0101] 3) Calculation time: The total calculation time is mainly determined by the position of the farthest defect and the position of the receiving signal sensor.

[0102] In summary, the calculation parameters of the pipe model are shown in Table 2:

[0103] Table 2 Main parameters for calculating the model pipe

[0104]

[0105] (3) Defect setting

[0106] In the present application, the defects are set to include a thinning defect with an axial length of 19 mm, a circumferential angle of 72°, a remaining wall thickness of 50%, and a distance of 2100 mm from the excitation point. A through defect with an axial length of 3 mm, a circumferential length of 40 mm, and a distance of 1400 mm from the excitation point, and its specific shape is shown in Figure 3 and Figure 4 .

[0107] (4) Boundary setting

[0108] During the simulation process, the residual back-propagating ultrasonic guided wave will undergo modal conversion when encountering the boundary of the model, and at the same time, boundary echoes will be generated, and the amplitude of the boundary echo signal is similar to that of the defect echo signal. The two signals are mixed together, increasing the difficulty of data analysis. In order to reduce the influence of the boundary echo on the simulation results and at the same time control the size of the model, an increasing damping absorption boundary is set in this study to suppress it.

[0109] In this experiment, the setting of the absorption boundary layer is as follows:

[0110] α MAX = 10f

[0111] L = 2λ

[0112] l / L = 1 / 10

[0113]

[0114] In the formula, α is the Rayleigh damping coefficient, l is the length of a single layer of absorption boundary, L is the total length of the absorption boundary, f is the excitation signal frequency, m is the layer number of the single layer of absorption boundary, and n is the power index.

[0115] 2. Mode excitation

[0116] (1) Selection of guided wave modes

[0117] In order to achieve long distance detection, the selected mode must keep the group velocity and phase velocity substantially unchanged in the working frequency band, avoiding the wave form broadening and energy attenuation caused by dispersion. Different modes have different response sensitivities to different defect types (circumferential crack, axial crack, thinning defect), and should be complementary to each other.

[0118] Longitudinal axisymmetric mode L(0, 2): fastest speed, uniform axial displacement field, sensitive to defects of any depth, less radial energy leakage, suitable for long distance detection and thinning defect detection;

[0119] Torsional axisymmetric mode T(0, 1): significant circumferential displacement, sensitive to axial cracks, strong anti-interference ability, excellent non-dispersion performance, suitable for detecting small through cracks;

[0120] (2) Excitation signal waveform and frequency optimization

[0121] Using a 10-cycle sine pulse signal modulated by a Hanning window can significantly reduce spectral leakage and suppress time-domain broadening caused by dispersion, and improve the signal-to-noise ratio of weak echoes;

[0122] As shown in Figure 5 The diameter of the pipeline (100 mm), the wall thickness (6 mm) and the mode cutoff frequency curve are selected as the working frequency of L(0, 2) and T(0, 1), which not only ensures the non-dispersion of the two, but also avoids the interference of high-order modes;

[0123] (3) Sensor arrangement

[0124] In order to only excite axisymmetric modes, the sensors are uniformly distributed along the circumference of the pipeline, and the total number must be greater than the highest mode order that can be excited, in order to filter out bending modes; The method for determining the highest mode that can be excited is as follows:

[0125] 1) Fourier transform of the excitation signal to obtain the corresponding frequency domain graph

[0126] 2) Cut the frequency at which the amplitude is -30bd on the frequency domain graph

[0127] 3) Compare the cut frequency with the dispersion curve of the pipeline to be tested to obtain the highest mode that can be excited at the cut frequency.

[0128] Experiments show that when the sensor spacing is ≥0.4λ, the imaging quality decreases significantly, and when it is ≥0.5λ, the defect cannot be identified, so the spacing should be kept <0.4λ (λ is the wavelength of the excited mode);

[0129] (4) One-way propagation implementation

[0130] The forward wave is enhanced and the backward wave is suppressed by two groups of sensors with a spacing (n 1 / 2)λ and a phase difference of π, so that the wave is transmitted in a single direction; for example, the amplitude of the reverse wave is reduced to about 1 / 5 of the forward wave by the method;

[0131] 3. Single mode imaging

[0132] In the pipe model, a distance of 2100 mm from the excitation point is set to cover the thinning defect with a circumferential range of 102°-180° and a distance of 1400 mm from the excitation point to cover the through defect at a circumferential position of 270°, the excitation mode is 70 KHz, the L(0,2) mode and the T(0,1) mode, and 16 nodes are uniformly set along the pipe circumference at a distance of 300 mm from the excitation point to receive data. The received data is subjected to synthetic aperture focusing imaging processing.

[0133] The synthetic aperture focusing algorithm flow is shown in Figure 6 , wherein:

[0134] The main function of the excitation signal generation and matched filter processing is to improve the signal-to-noise ratio and resolution of the defect echo signal. A single-frequency burst excitation signal s(t) with a Hanning window modulation is generated, with a center frequency f0, and the length of the measured signal is extended by zero padding. The received signal is subjected to frequency domain cross-correlation operation by using a matched filter (i.e. the conjugate frequency domain response of the excitation signal), to suppress noise and enhance the defect scattering components matching the excitation waveform.

[0135] The main function of the two-dimensional Fourier transform and frequency domain filtering is to extract the frequency-space features related to the defect. First, FFT is performed along the time axis and the sensor array space axis respectively to convert the signal to the frequency-wave number domain (ω-k domain). The frequency band centered at f0 with a bandwidth of 35% is retained by a band-pass filter, and high-frequency noise and low-frequency interference are removed, focusing on the effective guided wave mode energy.

[0136] The main function of the circumferential mode decomposition and filtering is to separate and select the dominant guided wave mode. Based on the geometric characteristics (outer diameter, number of array elements) of the sensor ring array, the frequency-wave number domain signal is decomposed into contributions of different circumferential mode orders n. The array sampling effect is compensated by a sinc function model, and the complex amplitude coefficient A(n,ω) of each mode is calculated. The cutoff order is set to filter out high-order modes to retain low-order mode components sensitive to defects.

[0137] The main function of the reverse propagation and focusing imaging link is to use the filtered low-order mode amplitude coefficient A(n,ω) and the phase velocity c p , group velocity c g of the guided wave to calculate the phase factor (z RFor reconstruction and stacking at the receiving sensor position, the high-resolution axial-circumferential defect focusing image is obtained by amplitude extraction and normalization, so as to accurately locate the defect position and geometric size.

[0138] The filtered low-order modal amplitude coefficient A(n, ω) and the phase velocity c p of the guided wave are used. g On the polar grid (z, θ), the phase factor z R is calculated for each frequency and each circumferential mode. For reconstruction and stacking at the receiving sensor position, the high-resolution axial-circumferential defect focusing image is obtained by amplitude extraction and normalization, so as to accurately locate the defect position and geometric size. Figure 7 Figure 8 The imaging results of L(0, 2) and T(0, 1) are calculated by the following formula:

[0139]

[0140] The meanings of the parameters are shown in the following table:

[0141]

[0142]

[0143] The imaging results are shown in Figure 7 and Figure 8 .

[0144] 4. Multi-modal fusion imaging

[0145] Using the data fusion technology to fuse the imaging results of different modes is an effective way to improve the imaging resolution and suppress the artifacts. The method for fusion imaging is:

[0146]

[0147] wherein is the linear domain of L(0, 2) mode imaging, is the linear domain of T(0, 1) mode imaging, I M (X, Y) is the fusion imaging result, is the imaging result of L(0, 2) mode, and the linear domains of the two modes are Gaussian blurred before fusion imaging to optimize the imaging effect. The fusion imaging result is shown in Figure 9 .

[0148] It can be observed from the images that the imaging results of the through defects and the thinning defects are clear and bright, which are better than the T(0, 1) and L(0, 2) single modal imaging. The fusion imaging effectively suppresses the artifacts and noise around the thinning defects in the L(0, 2) modal imaging, and more accurately indicates the position of the thinning defects. The images show that the echo signal amplitude of the through defects is higher than that of the thinning defects. Referring to the content in the fourth chapter, it can be preliminarily judged that the defect depth at 1400 mm is deeper. This effect is not possessed by the T(0, 1) and L(0, 2) single modal detection.

[0149] The white cross in the image represents the actual position of the defect, and the red circular ring represents the imaging position. The corresponding position parameters of the two are shown in Table 3:

[0150] Table 3: Pipeline damage positioning results

[0151]

[0152] This shows that the multi-modal image fusion can effectively utilize the complementary relationship between different modal guided waves, effectively avoid the problem that the single modal is difficult to detect the specific structural defects, can effectively improve the signal-to-noise ratio of the imaging result, and the defect positioning result is more obvious.

[0153] The above describes the preferred embodiments of the present application in detail. It should be understood that those skilled in the art can make many modifications and changes without creative labor according to the concept of the present application. Therefore, any technical solution obtained by logical analysis, reasoning or limited experiment on the basis of the prior art according to the concept of the present application should be within the protection scope determined by the claims.

Claims

1. A pipeline damage location method based on multi-modal fusion, characterized in that: The method comprises the following steps: A piezoelectric transducer array is evenly arranged along the circumference of the pipeline to generate two guided wave modes, non-dispersive L(0,2) and T(0,1), excited by 70kHz sinusoidal pulses. These guided waves serve as excitation signals, which propagate unidirectionally along the pipeline. Acquire the reflected echo at the test section of the pipeline, perform matched filtering, ω–k domain filtering, and circumferential modal decomposition on the reflected echo, and decompose it into two modal components under the L(0,2) and T(0,1) modes; Perform back propagation and focused imaging on each modal component to obtain imaging results in two modes: L(0,2) and T(0,1); The imaging results under the two modalities are fused to obtain a multimodal fusion pipeline damage image. The image is threshold segmented to obtain the pipeline defect type and the corresponding defect coordinates.

2. The pipeline damage location method based on multimodal fusion according to claim 1, characterized in that: The multimodal fusion pipeline damage image is: in, is the linear domain of the imaging result under the L(0,2) mode, is the linear domain of the imaging result under T(0,1) mode, I M (X, Y) is the fusion imaging result, ω L and ω T are the weights of the L(0,2) mode and the T(0,1) mode respectively.

3. The pipeline damage location method based on multimodal fusion according to claim 2, characterized in that: The linear domain of the imaging results under the L(0,2) mode and the linear domain of the imaging results under the T(0,1) mode are: in, and These are the imaging results under L(0,2) and T(0,1) modes respectively.

4. The pipeline damage location method based on multimodal fusion according to claim 1, characterized in that: The specific steps of matching filtering, ω–k domain filtering and circumferential modal decomposition of the reflected echo are as follows: The reflected echo is subjected to frequency domain cross-correlation operation using a matched filter. The obtained signal is subjected to FFT and filtering along the time axis and the spatial axis of the piezoelectric transducer array respectively to obtain the ω-k domain signal. The ω-k domain signal is decomposed into different circumferential modal orders and the cutoff order is set to filter out the high-order modes, thus obtaining two modal components under the L(0,2) and T(0,1) modes.

5. The pipeline damage location method based on multi-modal fusion according to claim 4 is characterized in that: The matched filter is the conjugate frequency domain response of the excitation signal.

6. The pipeline damage location method based on multimodal fusion according to claim 5, characterized in that: The specific steps of performing FFT and filtering along the time axis and the piezoelectric transducer array spatial axis are as follows: The obtained signal is subjected to FFT along the time axis and the spatial axis of the piezoelectric transducer respectively. The obtained FFT signal is passed through a bandpass filter to retain the frequency band centered on f0 and with a bandwidth of 35%, thereby obtaining the ω-k domain signal, where f0 is the center frequency of the excitation signal.

7. The pipeline damage location method based on multimodal fusion according to claim 4, characterized in that: The specific steps to obtain the imaging results under the two modes of L(0,2) and T(0,1) are: The modal amplitude coefficients and phase factors of the two modal components under the L(0,2) and T(0,1) modes are reconstructed and superimposed to obtain the imaging results under the L(0,2) and T(0,1) modes.

8. The pipeline damage location method based on multi-modal fusion according to claim 7, characterized in that: The imaging results of the two modes L(0,2) and T(0,1) are obtained by using the filtered low-order modal amplitude coefficient A(n,ω) and the phase velocity c of the guided wave. p , group velocity c g On the polar coordinate grid (z,θ), the phase factor is used for each frequency and each circumferential mode. where z R To receive the sensor position, reconstruction and superposition are performed, and finally it is obtained through amplitude extraction and normalization.

9. The pipeline damage location method based on multimodal fusion according to claim 1, characterized in that: The sinusoidal pulse is a signal modulated by a Hanning window.

10. The pipeline damage location method based on multi-modal fusion according to claim 1, characterized in that: Two adjacent groups of piezoelectric transducers in the piezoelectric transducer array are spaced apart by (n+1 / 2)λ, and have a phase difference of π.

Images