In-situ eddy-magnetic memory composite inspection method and system for service pipe

Abstract

This invention discloses an in-situ eddy current-magnetic memory composite flaw detection method and system for in-service pipe materials. The method includes attaching a composite probe in-situ to the surface of the in-service pipe material. The composite probe includes at least an eddy current detection unit, a magnetic memory detection unit, and a temperature detection unit. Eddy current signals, magnetic memory signals, and the temperature of the in-service pipe material at the location of the composite probe are simultaneously acquired from the surface of the in-service pipe material. Drift compensation is performed on the eddy current signals based on the real-time detected temperature of the in-service pipe material to obtain compensated eddy current signals. Noise reduction processing is performed on the magnetic memory signals and the compensated eddy current signals. The denoised dual-modal signals are input into a neural network classifier, which outputs defect identification results. The defect identification results include at least the determination of surface cracks and / or stress concentration areas. This invention enables in-situ, high-precision detection in extreme high-temperature and high-pressure environments without disassembling the pipe material, significantly shortening the detection cycle and reducing downtime losses.

Description

Technical Field

[0001] This invention relates to the field of pipe testing technology, specifically to an in-situ eddy current-magnetic memory composite flaw detection method and system for in-service pipes. Background Technology

[0002] High-temperature and high-pressure (working temperature 300~650℃, working pressure 10~30MPa) pipes are core components in aerospace, nuclear power and energy fields. Long-term operation in extreme environments can easily lead to defects such as surface cracks, stress concentration and intergranular corrosion, which directly affect equipment safety and lifespan.

[0003] Currently, the inspection of pipes used in high-temperature and high-pressure applications mainly relies on conventional eddy current testing and magnetic memory testing. Traditional eddy current testing, due to insufficient probe temperature resistance and significant signal drift susceptibility, typically requires removing the pipe from the equipment and conducting testing at room or medium temperature. It identifies surface and near-surface defects through electromagnetic induction, but removing the pipe at high temperatures can easily cause deformation and incurs substantial downtime costs. Magnetic memory testing, on the other hand, is based on the orientation changes of magnetic domains caused by material stress concentration. It infers stress distribution and potential defect areas by measuring the leakage magnetic field. While this method can be performed without removing the pipe, it suffers from low signal-to-noise ratios in environments with strong electromagnetic interference, easily leading to missed or false defects and poor accuracy. Therefore, existing technologies struggle to achieve accurate and reliable detection of surface cracks and stress concentration areas simultaneously under high-temperature and high-pressure conditions without removing the pipe. Summary of the Invention

[0004] In view of the shortcomings of the prior art, the technical problem to be solved by the present invention is: how to provide an eddy current-magnetic memory composite flaw detection method that can perform in-situ inspection without disassembling the pipe and has accurate and reliable defect detection results.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0006] An in-situ eddy current-magnetic memory composite flaw detection method for in-service pipe materials includes the following steps:

[0007] (1) The composite probe is attached to the surface of the pipe in service in situ. The composite probe includes at least an eddy current detection unit, a magnetic memory detection unit and a temperature detection unit.

[0008] (2) Simultaneously collect eddy current signals, magnetic memory signals and service pipe temperature at the location of the composite probe on the surface of the service pipe, and perform drift compensation on the eddy current signals based on the real-time detected service pipe temperature to obtain the compensated eddy current signals.

[0009] (3) The magnetic memory signal and the compensated eddy current signal are denoised, and the denoised dual-mode signal is input into the neural network classifier to output the defect identification result. The defect identification result includes at least the determination of surface cracks and / or stress concentration areas.

[0010] In this invention, a composite probe is attached in situ to the surface of the in-service pipe, enabling online detection without disassembling the pipe. This avoids the significant economic losses and pipe deformation risks associated with downtime disassembly. Simultaneous acquisition of eddy current signals, magnetic memory signals, and pipe temperature combines the dual advantages of eddy current sensitivity to surface cracks and magnetic memory sensitivity to stress concentration, overcoming the limitations of single detection methods. Drift compensation of the eddy current signal based on real-time temperature effectively solves the problem of eddy current signal distortion under high-temperature conditions. After noise reduction, the signal is input into a neural network classifier to output defect identification results, achieving automatic and accurate defect classification. Ultimately, this invention enables in-situ detection of the pipe without disassembly under high temperature and high pressure conditions, simultaneously identifying surface cracks and stress concentration areas, with accurate and reliable detection results.

[0011] As an optimization, in step (2), the eddy current signal drift compensation method includes: according to the material type of the service pipe, calling the pre-stored temperature-impedance compensation correspondence corresponding to the material, and determining the compensation coefficient corresponding to the real-time service pipe temperature based on the correspondence, and using the compensation coefficient to correct the amplitude of the eddy current signal. The conductivity and permeability of different materials change differently with temperature, and their eddy current signal temperature drift characteristics also differ significantly. By pre-establishing a dedicated temperature-impedance compensation correspondence for different materials and calling it according to the material type during detection, accurate compensation for material compatibility is achieved, avoiding the compensation error caused by using a uniform compensation coefficient for different materials.

[0012] As an optimization, the pre-stored temperature-impedance compensation relationship is obtained through pre-calibration. This includes measuring the eddy current signal impedance deviation of the in-service pipe at different temperatures using samples of the same material as the in-service pipe, calculating the compensation coefficient at each temperature point, and establishing a fitting curve or corresponding table of temperature and impedance compensation coefficients. By measuring the eddy current signal impedance deviation of samples of the same material as the in-service pipe at different temperatures, calculating the compensation coefficient at each temperature point, and establishing a fitting curve or corresponding table of temperature and compensation coefficients, the entire operating temperature range is covered. For any temperature not at the calibrated temperature point, accurate compensation coefficients can be obtained through linear interpolation or curve fitting.

[0013] As an optimization, in step (3), the denoising process is performed using a 5-level decomposition of the db8 wavelet. The db8 wavelet has excellent orthogonality and compact support characteristics. The 5-level decomposition can effectively separate the signal and noise at different scales. In addition, the threshold can be dynamically adjusted in combination with a high-temperature noise model, which can filter out various types of noise to the maximum extent while retaining the defect feature signal.

[0014] As an optimization, in step (3), an eddy current amplitude-phase map is constructed based on the compensated eddy current signal, and a leakage magnetic field gradient is calculated based on the magnetic memory signal to construct a magnetic memory leakage magnetic field gradient map. The neural network classifier is a dual-channel convolutional neural network, and the inputs of the dual-channel convolutional neural network are the eddy current amplitude-phase map and the magnetic memory leakage magnetic field gradient map. The eddy current amplitude-phase map can intuitively reflect the impedance change characteristics caused by defects such as surface cracks, while the magnetic memory leakage magnetic field gradient map highlights the magnetic field change rate characteristics in the stress concentration area. The dual-channel convolutional neural network extracts and fuses features from the two maps respectively, making full use of the complementary information of the two modal signals.

[0015] As an optimization, the output of the dual-channel convolutional neural network includes four categories: defect-free, surface crack, stress concentration zone, and intergranular corrosion. Surface cracks are detected primarily by eddy currents, stress concentration zones are detected primarily by magnetic memory, intergranular corrosion requires dual-modal joint identification, and defect-free represents the normal state. The four-category results are directly output through deep learning, eliminating the need for manual interpretation.

[0016] As an optimization, based on the leakage magnetic field gradient, the gradient extremum method is used. Calculate the stress value in the stress concentration zone, where k is the stress-magnetic field coefficient related to the material of the pipe in service. The residual life of the pipe is calculated using a fatigue life model based on the stress values ​​in the stress concentration zone, given the extreme values ​​of the leakage magnetic field gradient. Furthermore, the stress values ​​in the stress concentration zone can be calculated and used as input to the fatigue life model. Combined with parameters such as crack depth, the residual safe life of the pipe can be predicted.

[0017] This invention also discloses an in-situ eddy current-magnetic memory composite flaw detection system for in-service pipe materials to implement the above-described method, comprising a detection unit and a signal processing unit. The detection unit includes an eddy current detection module, a magnetic memory detection module, and a temperature detection module. The signal processing unit includes a temperature compensation module, a denoising module, and a neural network classification module. The output terminals of the eddy current detection module and the temperature detection module are electrically connected to the input terminal of the temperature compensation module, respectively. The temperature compensation module is used to correct the eddy current signal output by the eddy current detection module based on the in-service pipe material temperature value output by the temperature detection module. The output terminals of the temperature compensation module and the magnetic memory detection module are electrically connected to the input terminal of the denoising module, respectively. The denoising module is used to denoise the corrected eddy current signal and the magnetic memory signal. The output terminal of the denoising module is electrically connected to the input terminal of the neural network classification module, which is used to classify the denoised dual-modal signal and output the defect identification result. The modules work collaboratively to achieve fully automated processing from signal acquisition to defect identification.

[0018] As an optimization, the temperature compensation module is further configured to: based on the material type of the pipe in service, call the pre-stored temperature-impedance compensation correspondence corresponding to that material, determine the compensation coefficient corresponding to the real-time temperature of the pipe in service based on the temperature-impedance compensation correspondence, and use this compensation coefficient to correct the amplitude of the eddy current signal. This achieves material-adaptive temperature compensation at the system level, automatically matching compensation parameters for different pipe materials without manual intervention.

[0019] As an optimization, the system further includes a calculation and evaluation unit. The input of the calculation and evaluation unit is electrically connected to the output of the signal processing unit, and is used to receive the defect identification result output by the neural network classification module. When the defect identification result is a stress concentration area, the calculation and evaluation unit calculates the stress value of the stress concentration area based on the magnetic memory signal, and calculates the remaining life based on the stress value using the Miner fatigue life model. This realizes a complete intelligent evaluation chain from detecting the stress concentration area to quantifying the stress magnitude and predicting the remaining life, further improving the accuracy and efficiency of detection.

[0020] Compared with existing technologies, this invention can achieve in-situ detection in extreme high-temperature and high-pressure environments without disassembling the pipe, which greatly shortens the detection cycle and significantly reduces downtime losses. Through material-adaptive temperature compensation and noise reduction processing, combined with dual-channel neural network classification, it achieves high-precision detection of surface defects and stress concentration areas of pipes of various materials. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention;

[0022] Figure 2 This is a cross-sectional view of the composite probe in this invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0024] like Figure 1As shown, the in-situ eddy current-magnetic memory composite flaw detection method for in-service pipes in this specific embodiment includes the following steps:

[0025] (1) The composite probe is attached to the surface of the pipe in service in situ. The composite probe includes at least an eddy current detection unit, a magnetic memory detection unit and a temperature detection unit.

[0026] (2) Simultaneously collect eddy current signals, magnetic memory signals and service pipe temperature at the location of the composite probe on the surface of the service pipe, and perform drift compensation on the eddy current signals based on the real-time detected service pipe temperature to obtain the compensated eddy current signals.

[0027] (3) The magnetic memory signal and the compensated eddy current signal are denoised, and the denoised dual-mode signal is input into the neural network classifier to output the defect identification result. The defect identification result includes at least the determination of surface cracks and / or stress concentration areas.

[0028] In step (2), the drift compensation method for the eddy current signal includes: according to the material type of the service pipe, calling the pre-stored temperature-impedance compensation correspondence corresponding to the material, and determining the compensation coefficient corresponding to the real-time service pipe temperature based on the correspondence, and using the compensation coefficient to correct the amplitude of the eddy current signal.

[0029] The pre-stored temperature-impedance compensation relationship is obtained through pre-calibration, including measuring the eddy current signal impedance deviation of the service pipe at different temperatures using samples of the same material as the service pipe, calculating the compensation coefficient at each temperature point, and establishing a fitting curve or corresponding table of temperature and impedance compensation coefficient.

[0030] In step (3), the denoising process is performed by using db8 wavelet 5-layer decomposition.

[0031] In step (3), an eddy current amplitude-phase map is constructed based on the compensated eddy current signal, and a leakage magnetic field gradient is calculated based on the magnetic memory signal and a magnetic memory leakage magnetic field gradient map is constructed. The neural network classifier is a dual-channel convolutional neural network, and the input of the dual-channel convolutional neural network is the eddy current amplitude-phase map and the magnetic memory leakage magnetic field gradient map.

[0032] The output of the dual-channel convolutional neural network includes four types of results: defect-free, surface crack, stress concentration zone, and intergranular corrosion.

[0033] Based on the leakage magnetic field gradient, the gradient extremum method is used. Calculate the stress value in the stress concentration zone, where k is the stress-magnetic field coefficient related to the material of the pipe in service. The residual life of the pipe is calculated using a fatigue life model based on the stress value in the stress concentration zone, where the leakage magnetic field gradient is the extreme value.

[0034] An in-situ eddy current-magnetic memory composite flaw detection system for in-service pipe materials to implement the above-described method includes a detection unit and a signal processing unit. The detection unit includes an eddy current detection module, a magnetic memory detection module, and a temperature detection module. The signal processing unit includes a temperature compensation module, a denoising module, and a neural network classification module. The output terminals of the eddy current detection module and the temperature detection module are electrically connected to the input terminal of the temperature compensation module, respectively. The temperature compensation module is used to correct the eddy current signal output by the eddy current detection module based on the in-service pipe material temperature value output by the temperature detection module. The output terminals of the temperature compensation module and the magnetic memory detection module are electrically connected to the input terminal of the denoising module, respectively. The denoising module is used to denoise the corrected eddy current signal and the magnetic memory signal. The output terminal of the denoising module is electrically connected to the input terminal of the neural network classification module, which is used to classify the denoised dual-modal signal and output the defect identification result.

[0035] The temperature compensation module is further configured to: call the pre-stored temperature-impedance compensation correspondence corresponding to the material type of the pipe in service, determine the compensation coefficient corresponding to the real-time temperature of the pipe in service based on the temperature-impedance compensation correspondence, and use the compensation coefficient to correct the amplitude of the eddy current signal.

[0036] The system further includes a calculation and evaluation unit. The input of the calculation and evaluation unit is electrically connected to the output of the signal processing unit. It is used to receive the defect identification result output by the neural network classification module. When the defect identification result is a stress concentration area, the calculation and evaluation unit calculates the stress value of the stress concentration area according to the magnetic memory signal, and calculates the remaining life based on the stress value using the Miner fatigue life model.

[0037] In its specific implementation, this invention provides a magnetic-hydraulic composite clamp for pipes made of magnetic materials, comprising a permanent magnet adsorption unit, a hydraulic clamping unit, and a curvature adaptive base. The permanent magnet adsorption unit uses two N45 neodymium iron boron permanent magnets symmetrically arranged on both sides of the clamp, with a single magnet having an adsorption force ≥100N, used for initial fixation of the clamp to the pipe surface. The hydraulic clamping unit includes a miniature hydraulic cylinder (stroke ±5mm, thrust 500N), a pressure sensor, and a manual pressure regulating valve. The front end of the hydraulic cylinder piston rod is connected to a composite probe mounting base, allowing precise control of the contact pressure between the probe and the pipe (adjustable from 0.3-0.5MPa). The curvature adaptive base adopts an arc-shaped structure with a built-in elastic rubber pad (5mm thick), adapting to the pipe curvature (R≥50mm) to ensure a gap between the probe and the pipe ≤0.5mm.

[0038] like Figure 2 As shown, 1- 1. Ceramic encapsulation housing; 2. Eddy current coil; 3. Magnetic memory sensor; 4. Fiber Bragg grating; 5. Water-cooling channel; 6. Sapphire window. The composite probe uses... The probe features a ceramic housing (99.5% purity) with an internal sapphire window (≥90% transmittance) and a miniature water-cooling channel. This miniature water-cooling channel spirals around the probe's interior (between the eddy current coil and the ceramic housing), with a rectangular cross-section (1mm wide × 0.8mm high), a 2mm spacing between the spiral coils, and a total length of approximately 150mm. The inlet and outlet are located on the same side of the probe's tail. It connects to an external water-cooling system via a φ2mm ceramic conduit, with a cooling water flow rate of 0.5L / min, maintaining the probe's internal temperature below 80℃ (at an external ambient temperature of 650℃). The coil uses a polyimide-glass fiber composite insulation layer (400℃ heat resistance), and signal transmission utilizes an optical fiber-twisted pair composite cable (common-mode rejection ratio ≥80dB).

[0039] For each pipe material (GH4169, 316H, TC4), temperature-impedance compensation calibration was performed in a laboratory environment beforehand. The impedance deviation of the eddy current signal was measured at temperatures of 100℃, 200℃, 300℃, 400℃, 500℃, 600℃, and 650℃ (adjusted according to the upper limit of the material's temperature resistance), and the compensation coefficient was calculated for each temperature point. A fitting curve between temperature and the compensation coefficient was established using polynomial fitting or piecewise linear interpolation and stored in the signal processing unit. During testing, the corresponding compensation coefficient was retrieved based on the real-time temperature.

[0040] Pipe material to be inspected: GH4169 nickel-based high-temperature alloy pipe (Φ50×5mm, 320℃ / 15MPa), with a fatigue crack of 0.21mm depth on the surface and a stress concentration zone of 590MPa inside.

[0041] GH4169 is a ferromagnetic material. The clamp is initially attached to the pipe surface via the permanent magnet adsorption unit of the magnetic-hydraulic composite clamp. The hydraulic clamping unit fine-tunes the pressure to 0.4 MPa, and the pressure sensor indicates a contact gap of 0.3 mm. The water cooling system is then activated (flow rate 0.5 L / min). The fiber optic grating temperature sensor detects a probe contact area temperature of 320℃. The temperature-impedance compensation relationship for GH4169 material is applied (pre-calibrated, compensation coefficient K=1.09 at 320℃). The original eddy current signal amplitude is 120 mV, which is compensated to 130.8 mV. The eddy current frequency... With a sampling rate of 150kHz and a magnetic memory sampling rate of 1kHz, a 10-second signal was acquired synchronously. The FPGA parallel acquisition module achieved a time synchronization accuracy of ≤1μs. A 5-layer decomposition of the db8 wavelet was used for denoising, improving the signal-to-noise ratio to 22dB. An eddy current amplitude-phase spectrum was constructed based on the compensated eddy current signal. The leakage magnetic field gradient was calculated based on the magnetic memory signal, and a gradient spectrum was constructed and input into a dual-channel convolutional neural network. Due to the presence of significant peaks (gradient ≥5μT / mm) in the magnetic memory gradient spectrum, the dual-channel convolutional neural network outputs surface cracks and stress concentration areas. The extreme values ​​of the leakage magnetic field gradient were extracted. The k of GH4169 is 89.4 MPa / (μT / mm) (pre-calibrated experimentally). Substituting this into the formula... The calculated stress value σ = 590 MPa was obtained. Referring to the GH4169 SN curve (stress-life curve), the fatigue life corresponding to 590 MPa was found. In the next cycle, the actual operating load frequency is 0.1Hz (one cycle every 10 seconds). Substituting this into the Miner fatigue life model, the remaining life is calculated. Hour.

[0042] Pipe material to be inspected 2: 316H austenitic stainless steel pipe (Φ159×10mm, 350℃ / 12MPa), containing 0.15mm precision intergranular corrosion, stress concentration zone 550MPa.

[0043] 316H is a non-ferromagnetic material, and permanent magnets cannot attract it; therefore, other methods are used for auxiliary fixation. The temperature of the fiber optic grating temperature sensor probe contact area is 350℃. The temperature-impedance compensation relationship for 316H material is applied (pre-calibrated, compensation coefficient K=1.07 at 350℃). The original eddy current signal amplitude is 150mV, which is compensated to 160.5mV. The eddy current frequency is 150kHz, the magnetic memory sampling rate is 1kHz, and 10 seconds of signal are acquired synchronously. The FPGA parallel acquisition module achieves time synchronization accuracy ≤1μs. After denoising using 5-layer decomposition with db8 wavelets, eddy current amplitude-phase maps and magnetic memory leakage field gradient maps are constructed and input into a dual-channel convolutional neural network. The magnetic memory gradient map shows stress concentration characteristics (gradient ≥5μT / mm), and the dual-channel convolutional neural network outputs intergranular corrosion and stress concentration areas. The extreme values ​​of the leakage field gradient are extracted. The k of 316H is 76.4 MPa / (μT / mm) (pre-calibrated experimentally). Substituting this into the formula... The calculated stress value σ = 550 MPa was obtained. Referring to the 316H SN curve (stress-life curve), the fatigue life corresponding to 550 MPa was found. In the next cycle, the actual operating load frequency is 0.08Hz. Substituting this into the Miner fatigue life model, the remaining life is calculated. Hour.

[0044] Pipe material to be inspected 3: TC4 titanium alloy pipe (Φ38×4mm, 300℃ / 10MPa), containing a 0.1mm deep fatigue crack, with a stress concentration zone of 500MPa.

[0045] TC4 is a non-ferromagnetic material, and permanent magnets cannot attract it; therefore, other methods are used for auxiliary fixation. The fiber optic grating temperature sensor detects a contact area temperature of 300℃. The temperature-impedance compensation relationship for TC4 material is applied (pre-calibrated, compensation coefficient K=1.04 at 300℃). The original eddy current signal amplitude is 180mV, which is compensated to 187.2mV. The eddy current frequency is increased to 200kHz (to improve the signal-to-noise ratio of the titanium alloy detection), the magnetic memory sampling rate is 1kHz, and 10-second signals are acquired synchronously. The FPGA parallel acquisition module achieves time synchronization accuracy ≤1μs. After denoising using 5-layer decomposition with db8 wavelets, eddy current amplitude-phase maps and magnetic memory leakage field gradient maps are constructed and input into a dual-channel convolutional neural network. The magnetic memory gradient map shows stress concentration characteristics (gradient ≥5μT / mm), and the dual-channel convolutional neural network outputs surface cracks and stress concentration areas. The extreme values ​​of the leakage field gradient are extracted. TC4's k = 54.9 MPa / (μT / mm) (pre-calibrated experimentally), substituting into the formula... The calculated stress value σ = 500 MPa was obtained. Referring to the TC4 SN curve (stress-life curve), the fatigue life corresponding to 500 MPa was determined. In the next cycle, the actual operating load frequency is 0.12Hz. Substituting this into the Miner fatigue life model, the remaining life is calculated. Hour.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit the technical solutions. Those skilled in the art should understand that any modifications or equivalent substitutions to the technical solutions of the present invention without departing from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A method of in-situ eddy current-magnetic memory composite inspection of a service pipe material, characterized in that: Includes the following steps: (1) The composite probe is attached to the surface of the pipe in service in situ. The composite probe includes at least an eddy current detection unit, a magnetic memory detection unit and a temperature detection unit. (2) Simultaneously collect eddy current signals, magnetic memory signals and service pipe temperature at the location of the composite probe on the surface of the service pipe, and perform drift compensation on the eddy current signals based on the real-time detected service pipe temperature to obtain the compensated eddy current signals. (3) The magnetic memory signal and the compensated eddy current signal are denoised, and the denoised dual-mode signal is input into the neural network classifier to output the defect identification result. The defect identification result includes at least the determination of surface cracks and / or stress concentration areas.

2. The method of in-situ eddy current-magnetic memory composite inspection of a service pipe as claimed in claim 1, wherein: In step (2), the drift compensation method for the eddy current signal includes: according to the material type of the service pipe, calling the pre-stored temperature-impedance compensation correspondence corresponding to the material, and determining the compensation coefficient corresponding to the real-time service pipe temperature based on the correspondence, and using the compensation coefficient to correct the amplitude of the eddy current signal.

3. The in-situ eddy current-magnetic memory composite flaw detection method for in-service pipes according to claim 2, characterized in that: The pre-stored temperature-impedance compensation relationship is obtained through pre-calibration, including measuring the eddy current signal impedance deviation of the service pipe at different temperatures using samples of the same material as the service pipe, calculating the compensation coefficient at each temperature point, and establishing a fitting curve or corresponding table of temperature and impedance compensation coefficient.

4. The in-situ eddy current-magnetic memory composite flaw detection method for in-service pipes according to claim 1, characterized in that: In step (3), the denoising process is performed by using db8 wavelet 5-layer decomposition.

5. The in-situ eddy current-magnetic memory composite flaw detection method for in-service pipes according to claim 1, characterized in that: In step (3), an eddy current amplitude-phase map is constructed based on the compensated eddy current signal, and a leakage magnetic field gradient is calculated based on the magnetic memory signal and a magnetic memory leakage magnetic field gradient map is constructed. The neural network classifier is a dual-channel convolutional neural network, and the input of the dual-channel convolutional neural network is the eddy current amplitude-phase map and the magnetic memory leakage magnetic field gradient map.

6. The in-situ eddy current-magnetic memory composite flaw detection method for in-service pipes according to claim 5, characterized in that: The output of the dual-channel convolutional neural network includes four types of results: defect-free, surface crack, stress concentration zone, and intergranular corrosion.

7. The in-situ eddy current-magnetic memory composite flaw detection method for in-service pipes according to claim 5, characterized in that: Based on the leakage magnetic field gradient, the gradient extremum method is used. Calculate the stress value in the stress concentration zone, where k is the stress-magnetic field coefficient related to the material of the pipe in service. The residual life of the pipe is calculated using a fatigue life model based on the stress value in the stress concentration zone, where the leakage magnetic field gradient is the extreme value.

8. An in-situ eddy current-magnetic memory composite flaw detection system for service pipes used to implement the method of claim 1, characterized in that: It includes a detection unit and a signal processing unit. The detection unit includes an eddy current detection module, a magnetic memory detection module, and a temperature detection module. The signal processing unit includes a temperature compensation module, a denoising module, and a neural network classification module. The outputs of the eddy current detection module and the temperature detection module are electrically connected to the input of the temperature compensation module. The temperature compensation module is used to correct the eddy current signal output by the eddy current detection module based on the service pipe temperature value output by the temperature detection module. The outputs of the temperature compensation module and the magnetic memory detection module are electrically connected to the input of the denoising module. The denoising module is used to denoise the corrected eddy current signal and the magnetic memory signal. The output of the denoising module is electrically connected to the input of the neural network classification module. The neural network classification module is used to classify the denoised dual-modal signal and output the defect identification result.

9. The in-situ eddy current-magnetic memory composite flaw detection system for in-service pipes according to claim 8, characterized in that: The temperature compensation module is further configured to: call the pre-stored temperature-impedance compensation correspondence corresponding to the material type of the pipe in service, determine the compensation coefficient corresponding to the real-time temperature of the pipe in service based on the temperature-impedance compensation correspondence, and use the compensation coefficient to correct the amplitude of the eddy current signal.

10. The in-situ eddy current-magnetic memory composite flaw detection system for in-service pipes according to claim 8, characterized in that: The system further includes a calculation and evaluation unit. The input of the calculation and evaluation unit is electrically connected to the output of the signal processing unit. It is used to receive the defect identification result output by the neural network classification module. When the defect identification result is a stress concentration area, the calculation and evaluation unit calculates the stress value of the stress concentration area according to the magnetic memory signal, and calculates the remaining life based on the stress value using the Miner fatigue life model.

Images