阅读说明：本技术 基于锁相放大的多频旋转磁场的金属缺陷检测装置及方法 (Metal defect detection device and method based on multi-frequency rotating magnetic field of phase-locked amplification ) 是由 刘金海 李松凯 汪刚 马大中 冯健 卢森骧 于 2019-09-26 设计创作，主要内容包括：本发明提供一种基于锁相放大的多频旋转磁场的金属缺陷检测装置及方法,涉及电磁无损检测技术领域。该方法首先通过正弦激励信号发生模块产生正弦电流激励信号为检测探头提供激励,检测探头检测到电压信号并通过信号处理模块对信号数据处理后输入到缺陷数据识别模块,缺陷数据识别模块基于可扩展处理平台ZYNQ采用归一化动态阈值法对管道金属数据进行归一化,识别出管道缺陷特征,同时对异常数据进行滤波；将识别到的缺陷特征数据输入到基于随机森林的缺陷角度识别模块对缺陷的角度进行识别,得到最终的缺陷分类结果。该装置及方法能够实现非接触耦合检测及金属近表面的任意方向缺陷检测,同时可以实现缺陷任意角度的识别。(The invention provides a device and a method for detecting metal defects of a multi-frequency rotating magnetic field based on phase-locked amplification, and relates to the technical field of electromagnetic nondestructive detection. The method comprises the steps that firstly, a sinusoidal current excitation signal is generated by a sinusoidal excitation signal generation module to provide excitation for a detection probe, the detection probe detects a voltage signal, the signal data are processed by a signal processing module and then input to a defect data identification module, the defect data identification module normalizes pipeline metal data by adopting a normalized dynamic threshold method based on an extensible processing platform ZYNQ, pipeline defect characteristics are identified, and meanwhile abnormal data are filtered; and inputting the identified defect characteristic data into a defect angle identification module based on a random forest to identify the angle of the defect, so as to obtain a final defect classification result. The device and the method can realize non-contact coupling detection and any direction defect detection of the near surface of the metal, and can realize the identification of any angle of the defect.)

1. The utility model provides a metal defect detection device of multifrequency rotating magnetic field based on lock is enlarged which characterized in that: the system comprises a sinusoidal excitation signal generation module, a detection probe, a signal processing module, a defect data identification module and a defect angle identification module based on random forests;

the sinusoidal excitation signal generation module is used for generating a sinusoidal current excitation signal and providing excitation for a detection probe for detecting metal defects; the detection probe comprises a rotating magnetic field detection coil and a TMR giant magneto-resistance sensor; the rotating magnetic field detection coil comprises three rectangular coils which form an angle of 120 degrees with each other, and the three rectangular coils are connected through a triangular magnetic yoke which forms an angle of 120 degrees with each other and are used for generating a three-phase rotating magnetic field so as to induce a three-phase rotating eddy current on the surface of a test piece to be detected; the TMR giant magnetoresistance sensor is placed in the center of the rotating magnetic field detection coil and used for detecting the magnetic field change caused by the rotating magnetic field detection coil and a test piece to be detected and transmitting a detected voltage signal to the signal processing module; the signal processing module carries out frequency-selecting filtering and analog-to-digital conversion on the detected voltage signal and then inputs the voltage signal to the defect data identification module; the defect data identification module receives the data transmitted by the signal processing module, and performs normalization processing and defect identification processing on the data to obtain defect data and sends the defect data to the defect angle identification module; and the defect angle identification module receives the defect data transmitted by the defect data identification module and identifies the angle of the defect through an angle identification algorithm.

2. The apparatus for detecting metal defects based on a phase-locked amplified multi-frequency rotating magnetic field according to claim 1, wherein: the sinusoidal excitation signal generation module comprises a DDS chip and a voltage-current conversion module; the DDS chip generates a voltage signal, which is then converted into a current signal by a voltage-current conversion module.

3. The apparatus for detecting metal defects based on a phase-locked amplified multi-frequency rotating magnetic field according to claim 1, wherein: the signal processing module comprises a lock-in amplifier module and an AD conversion module; the locking amplifier module is used for carrying out frequency-selective filtering and amplifying on the voltage signal detected by the TMR giant magneto-resistance sensor and then inputting the voltage signal into the AD conversion module; the AD conversion module converts the analog signal into a digital signal and inputs the converted digital signal into the defect data identification module.

4. The apparatus for detecting metal defects based on a phase-locked amplified multi-frequency rotating magnetic field according to claim 1, wherein: the defect data identification module processes data output by the signal processing module by adopting a normalized dynamic threshold value method based on the extensible processing platform ZYNQ, normalizes the pipeline data, calculates a dynamic threshold value for identifying the metal defect characteristics of the pipeline, filters abnormal data while obtaining the defect data characteristics, and inputs the identified defect characteristic data to the defect angle identification module based on the random forest.

5. The apparatus for detecting metal defects based on a phase-locked amplified multi-frequency rotating magnetic field according to any one of claims 1 to 4, wherein: the random forest based angle defect recognition module is used for collecting D through training samples on an upper computer_aConstructing a defect angle identification model, and testing the sample set D_bVerifying the accuracy of the identification model, and performing iterative feedback on the identification model to obtain an optimal angle identification model; through the established optimal defect angle identification model, the characteristic vector X of the detection signal is equal to { B ] in the defect angle analysis process_angle，B_heighObtaining a corresponding label vector Y ═ angle, h }, wherein angle is the angle of the defect, h is the depth of the defect, and B is the depth of the defect_angleAs a difference in the magnetic field of the X-axis component of the defect, B_heighIs the defect Z-axis component magnetic field difference.

6. A metal defect detection method based on a multi-frequency rotating magnetic field amplified by a lock phase, which adopts the detection device of claim 5 to detect metal defects, and is characterized in that: the method comprises the following steps:

step 1, generating a sinusoidal current excitation signal through a sinusoidal excitation signal generation module, and generating a three-phase rotating magnetic field through a rotating magnetic field detection coil in a detection probe so as to induce a three-phase rotating eddy current on the surface of a test piece to be detected; the TMR giant magnetoresistance sensor in the detection probe detects the magnetic field change caused by the rotating magnetic field detection coil and the test piece to be detected, and transmits the detected voltage signal to the signal processing module;

step 2, the signal processing module carries out frequency-selective filtering and analog-to-digital conversion on the detected voltage signal and inputs the voltage signal to the defect data identification module;

step 3, the defect data identification module processes the data output by the signal processing module by adopting a normalized dynamic threshold method based on the extensible processing platform ZYNQ, normalizes the metal data of the pipeline, calculates a dynamic threshold for identifying the defect characteristics of the pipeline, filters abnormal data while obtaining the defect data characteristics, and inputs the identified defect characteristic data into a defect angle identification module based on a random forest;

and 4, identifying the angle of the defect by the defect angle identification module through an angle identification algorithm to obtain a final defect classification result.

7. The method of claim 6, wherein the method comprises: the specific method of the step 3 comprises the following steps:

step 3.1: establishing initial threshold lambda of pipe metal defect angle and defect height_angle，λ_heigh；

Step 3.2: the acquired X-axis magnetic field data and Z-axis magnetic field data are normalized by depending on the background magnetic field, and the following formula is shown:

wherein, B_{x_back}For detecting the X-axis component of the magnetic field when it is defect-free, B_xFor detecting the X-axis component of the magnetic field when there is a defect, B_{z_back}For detecting the Z-component of the magnetic field in the absence of defects, B_zDetecting a Z-axis component of the magnetic field when the defect is detected;

step 3.3: b after normalization_angle，B_heighAnd a threshold lambda_angle，λ_heighBy comparison, when B_angle≥λ_angleWhen it is, B_angleMarking as a significandAccording to the method, the angle of the metal defect of the pipeline is expressed; all the same thing as B_heigh≥λ_heighWhen it is, B_heighAnd marking the data as valid data for expressing the depth of the metal defect of the pipeline.

8. The method of claim 7, wherein the method comprises: the specific method of the step 4 comprises the following steps:

the specific method comprises the following steps:

step 4.1: operating a metal defect detection device in a pipeline with known position and size, and recording a sample data set X of experimental data₁And sample tag set Y₁；

The sample data set comprises data of pipe metal defect angles and defect heights; the sample label set comprises sample labels of the corresponding relation between the actual defect angle and the defect height of the sample data set pipeline metal defect angle data and the defect height data respectively;

step 4.2: establishing a pipeline defect detection simulation model by using finite element simulation software, and recording a sample data set X of simulation data₂And sample tag set Y₂；

Step 4.3: randomly selecting a sample; by applying to a sample data set X₁And X₂In the method, random sampling with replacement is carried out to construct a certain number of new test sample sets, wherein random sampling with replacement is carried out for M times, and N data are sampled each time, thereby constructing M test sample sets X_m；

Step 4.4: randomly selecting a sample label; from the sample tag set Y₁And Y₂Random sampling without putting back is carried out, and a certain number of new test sample label sets are constructed; randomly sampling k sample labels without putting back, calculating the information gain of the samples, and then selecting an optimal label;

step 4.5: constructing a decision tree; determining node selection of the decision tree by using an information entropy method, namely selecting the node sequence of the decision tree by calculating the information entropy of each node so as to construct the decision tree;

step 4.6: cutting a decision tree; cutting the decision tree established in the step 4.5 by using a cost complexity pruning method;

step 4.7: random forest voting classification is carried out; repeating the steps 4.5-4.6 to construct L decision trees, training the L decision trees aiming at the test samples respectively to obtain L results, and forming a voting result set T by the results_lAnd calculating a voting result T_iAnd T_jEuclidean distance d (T) therebetween_i，T_j) The following formula shows:

wherein, T_iFor the ith voting result, T_jFor the jth voting result, i ≠ 1, …, L, j ≠ 1, …, L, and i ≠ j; deeply cutting the decision tree corresponding to the classification result with the maximum distance, and then classifying again until the distance of the voting result is less than a threshold value d_stopAnd stopping cutting to obtain a final defect classification result.

Technical Field

The invention relates to the technical field of electromagnetic nondestructive testing, in particular to a metal defect detection device and method of a multi-frequency rotating magnetic field based on phase-locked amplification.

Background

Due to the transportation mode of pipeline transportation, the transportation capacity is large, and the efficiency is high; the transportation cost is low; generally, no pollution is generated; the transportation vehicle is usually buried underground, is safe and reliable, is not easily limited by external conditions, and the like, and becomes a fifth great transportation vehicle after transportation by roads, railways, water ways, aviation and the like. Since the pipeline is buried deeply in the ground or under the sea for a long time, the pipe wall of the pipeline is damaged by natural corrosion, and the like, and the problems possibly cause safety accidents such as pipeline leakage in the future. Therefore, a system capable of accurately detecting the pipeline defects in time and positioning the defects is urgently needed, and the safe operation of the pipeline system is protected.

The common nondestructive testing method comprises the following steps: eddy Current Test (ECT), Radiographic Test (RT), Ultrasonic Test (UT), magnetic particle test (MT), and liquid Penetration Test (PT). Other non-destructive testing methods: acoustic emission inspection (AE), thermographic/infrared (TIR), Leak Test (LT), ac field measurement technique (ACFMT), magnetic flux leakage test (MFL), far field test detection method (RFT), ultrasonic diffraction time difference method (TOFD), and the like. However, these methods have their own application range and corresponding limitations: some require reagent coupling; some tested agents show higher cleanliness; some defects have low recognition accuracy and cannot recognize the angles of the defects.

Disclosure of Invention

The invention aims to solve the technical problem of the prior art, and provides a metal defect detection device and method based on a multi-frequency rotating magnetic field amplified by a lock phase, so as to realize the identification of defect angles in a pipeline.

In order to solve the technical problems, the technical scheme adopted by the invention is as follows: on one hand, the invention provides a metal defect detection device based on a multi-frequency rotating magnetic field amplified by phase locking, which comprises a sinusoidal excitation signal generation module, a detection probe, a signal processing module, a defect data identification module and a defect angle identification module based on random forests;

the sinusoidal excitation signal generation module is used for generating a sinusoidal current excitation signal and providing excitation for a detection probe for detecting metal defects; the detection probe comprises a rotating magnetic field detection coil and a TMR giant magneto-resistance sensor; the rotating magnetic field detection coil comprises three rectangular coils which form an angle of 120 degrees with each other, and the three rectangular coils are connected through a triangular magnetic yoke which forms an angle of 120 degrees with each other and are used for generating a three-phase rotating magnetic field so as to induce a three-phase rotating eddy current on the surface of a test piece to be detected; the TMR giant magnetoresistance sensor is placed in the center of the rotating magnetic field detection coil and used for detecting the magnetic field change caused by the rotating magnetic field detection coil and a test piece to be detected and transmitting a detected voltage signal to the signal processing module; the signal processing module carries out frequency-selecting filtering and analog-to-digital conversion on the detected voltage signal and then inputs the voltage signal to the defect data identification module; the defect data identification module receives the data transmitted by the signal processing module, and performs normalization processing and defect identification processing on the data to obtain defect data and sends the defect data to the defect angle identification module; and the defect angle identification module receives the defect data transmitted by the defect data identification module and identifies the angle of the defect through an angle identification algorithm.

Preferably, the sinusoidal excitation signal generation module comprises a DDS chip and a voltage-current conversion module; the DDS chip generates a voltage signal, and then the voltage signal is converted into a current signal through a voltage-current conversion module;

preferably, the signal processing module comprises a lock-in amplifier module and an AD conversion module; the locking amplifier module is used for carrying out frequency-selective filtering and amplifying on the voltage signal detected by the TMR giant magneto-resistance sensor and then inputting the voltage signal into the AD conversion module; the AD conversion module converts the analog signal into a digital signal and inputs the converted digital signal into the defect data identification module.

Preferably, the defect data identification module processes the data output by the signal processing module by using a normalized dynamic threshold method based on the scalable processing platform ZYNQ, normalizes the pipeline data, calculates a dynamic threshold for identifying the metal defect characteristics of the pipeline, filters abnormal data while obtaining the defect data characteristics, and inputs the identified defect characteristic data to the defect angle identification module based on the random forest.

Preferably, the random forest based angle defect identification module is used for training a sample set D on an upper computer_aConstructing a defect angle identification model, and testing the sample set D_bVerifying the accuracy of the identification model, and performing iterative feedback on the identification model to obtain an optimal angle identification model; through the established optimal defect angle identification model, the characteristic vector X of the detection signal is equal to { B ] in the defect angle analysis process_angle，B_heighObtaining a corresponding label vector Y ═ angle, h }, wherein angle is the angle of the defect, h is the depth of the defect, and B is the depth of the defect_angleAs a difference in the magnetic field of the X-axis component of the defect, B_heighIs the defect Z-axis component magnetic field difference.

On the other hand, the invention also provides a metal defect detection method based on the multi-frequency rotating magnetic field amplified by the phase lock, which comprises the following steps:

step 1, generating a sinusoidal current excitation signal through a sinusoidal excitation signal generation module, and generating a three-phase rotating magnetic field through a rotating magnetic field detection coil in a detection probe so as to induce a three-phase rotating eddy current on the surface of a test piece to be detected; the TMR giant magnetoresistance sensor in the detection probe detects the magnetic field change caused by the rotating magnetic field detection coil and the test piece to be detected, and transmits the detected voltage signal to the signal processing module;

step 2, the signal processing module carries out frequency-selective filtering and analog-to-digital conversion on the detected voltage signal and inputs the voltage signal to the defect data identification module;

step 3, the defect data identification module processes the data output by the signal processing module by adopting a normalized dynamic threshold method based on the extensible processing platform ZYNQ, normalizes the metal data of the pipeline, calculates a dynamic threshold value used for identifying the defect characteristics of the pipeline, filters abnormal data while obtaining the defect data characteristics, and inputs the identified defect characteristic data into the defect angle identification module based on the random forest, and the specific method comprises the following steps:

step 3.1: establishing initial threshold lambda of pipe metal defect angle and defect height_angle，λ_heigh；

Step 3.2: the acquired X-axis magnetic field data and Z-axis magnetic field data are normalized by depending on the background magnetic field, and the following formula is shown:

wherein, B_{x_back}For detecting the X-axis component of the magnetic field when it is defect-free, B_xFor detecting the X-axis component of the magnetic field when there is a defect, B_{z_back}For detecting the Z-component of the magnetic field in the absence of defects, B_zDetecting a Z-axis component of the magnetic field when the defect is detected;

step 3.3: b after normalization_angle，B_heighAnd a threshold lambda_angle，λ_heighBy comparison, when B_angle≥λ_angleWhen it is, B_angleMarking the data as valid data for expressing the angle of the metal defect of the pipeline; all the same thing as B_heigh≥λ_heighWhen it is, B_heighMarking the data as valid data for expressing the depth of the metal defect of the pipeline;

step 4, the defect angle identification module identifies the angle of the defect through an angle identification algorithm to obtain a final defect classification result, and the specific method comprises the following steps:

step 4.1: operating a metal defect detection device in a pipeline with known position and size, and recording a sample data set X of experimental data₁And sample tag set Y₁；

The sample data set comprises data of pipe metal defect angles and defect heights; the sample label set comprises sample labels of the corresponding relation between the actual defect angle and the defect height of the sample data set pipeline metal defect angle data and the defect height data respectively;

step 4.2: using finite elementsThe simulation software establishes a pipeline defect detection simulation model and records a sample data set X of simulation data₂And sample tag set Y₂；

Step 4.3: randomly selecting a sample; by applying to a sample data set X₁And X₂In the method, random sampling with replacement is carried out to construct a certain number of new test sample sets, wherein random sampling with replacement is carried out for M times, and N data are sampled each time, thereby constructing M test sample sets X_m；

Step 4.4: randomly selecting a sample label; from the sample tag set Y₁And Y₂Random sampling without putting back is carried out, and a certain number of new test sample label sets are constructed; randomly sampling k sample labels without putting back, calculating the information gain of the samples, and then selecting an optimal label;

step 4.5: constructing a decision tree; determining node selection of the decision tree by using an information entropy method, namely selecting the node sequence of the decision tree by calculating the information entropy of each node so as to construct the decision tree;

step 4.6: cutting a decision tree; cutting the decision tree established in the step 4.5 by using a cost complexity pruning method;

step 4.7: random forest voting classification is carried out; repeating the steps 4.5-4.6 to construct L decision trees, training the L decision trees aiming at the test samples respectively to obtain L results, and forming a voting result set T by the results_lAnd calculating a voting result T_iAnd T_jEuclidean distance d (T) therebetween_i，T_j) The following formula shows:

wherein, T_iFor the ith voting result, T_jFor the jth voting result, i ≠ 1, …, L, j ≠ 1, …, L, and i ≠ j; deeply cutting the decision tree corresponding to the classification result with the maximum distance, and then classifying again until the distance of the voting result is less than a threshold value d_stopWhen the cutting is stopped,and obtaining a final defect classification result.

Adopt the produced beneficial effect of above-mentioned technical scheme to lie in: the device and the method for detecting the metal defects of the multi-frequency rotating magnetic field based on the phase-locked amplification are realized based on the electromagnetic eddy current principle, so that non-contact coupling detection can be realized; the excitation signal is a three-phase rotating eddy current, so that the defect detection in any direction of the near surface of the metal can be realized, and the detection depth can be adjusted according to the frequency of the excitation signal; defect detection based on dynamic threshold can realize defect identification; the random forest based defect angle identification module can realize the identification of any defect angle.

Drawings

Fig. 1 is a block diagram of a metal defect detection apparatus based on a phase-locked amplified multi-frequency rotating magnetic field according to an embodiment of the present invention;

fig. 2 is a schematic model diagram of a three-phase rotating magnetic field probe according to an embodiment of the present invention;

FIG. 3 is a graph of a sinusoidal excitation current waveform provided by an embodiment of the present invention;

fig. 4 is a schematic diagram of AB-phase rotating magnetic field synthesis provided in the embodiment of the present invention.

FIG. 5 is a circuit schematic of a lock-in amplifier provided by an embodiment of the invention;

fig. 6 is a flowchart illustrating a defect angle identification module based on a random forest according to an embodiment of the present invention to identify a metal defect angle.

Detailed Description

The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.

A metal defect detection device based on a multi-frequency rotating magnetic field amplified by phase locking is shown in figure 1 and comprises a sinusoidal excitation signal generation module, a detection probe, a signal processing module, a defect data identification module and a defect angle identification module based on random forests;

the sinusoidal excitation signal generation module is used for generating a sinusoidal current excitation signal and providing excitation for a detection probe for detecting metal defects; the detection probe comprises a rotating magnetic field detection coil and a TMR giant magneto-resistance sensor; the rotating magnetic field detection coil comprises three rectangular coils which form an angle of 120 degrees with each other, and the three rectangular coils are connected through a triangular magnetic yoke which forms an angle of 120 degrees with each other and are used for generating a three-phase rotating magnetic field, so that a three-phase rotating eddy current is induced on the surface of a test piece to be detected; the TMR giant magnetoresistance sensor is placed in the center of the rotating magnetic field detection coil and used for detecting the magnetic field change caused by the rotating magnetic field detection coil and a test piece to be detected and transmitting a detected voltage signal to the signal processing module; the signal processing module carries out frequency-selecting filtering and analog-to-digital conversion on the detected voltage signal and then inputs the voltage signal to the defect data identification module; the defect data identification module receives the data transmitted by the signal processing module, and performs normalization processing and defect identification processing on the data to obtain defect data and sends the defect data to the defect angle identification module; and the defect angle identification module receives the defect data transmitted by the defect data identification module and identifies the angle of the defect through an angle identification algorithm.

The sinusoidal excitation signal generation module comprises a DDS chip and a voltage-current conversion module; the DDS chip generates a voltage signal, and then the voltage signal is converted into a current signal through a voltage-current conversion module;

the signal processing module comprises a lock-in amplifier module and an AD conversion module; the locking amplifier module is used for carrying out frequency-selective filtering and amplifying on the voltage signal detected by the TMR giant magneto-resistance sensor and then inputting the voltage signal into the AD conversion module; the AD conversion module converts the analog signal into a digital signal and inputs the converted digital signal into the defect data identification module.

The defect data identification module processes data output by the signal processing module by adopting a normalized dynamic threshold value method based on the extensible processing platform ZYNQ, normalizes the pipeline data, calculates a dynamic threshold value for identifying the metal defect characteristics of the pipeline, filters abnormal data while obtaining the defect data characteristics, and inputs the identified defect characteristic data to the defect angle identification module based on the random forest.

The random forest based angle defect recognition module is used for collecting D through training samples on an upper computer_aConstructing a defect angle identification model, and testing the sample set D_bVerifying the accuracy of the identification model, and performing iterative feedback on the identification model to obtain an optimal angle identification model; through the established optimal defect angle identification model, the characteristic vector X of the detection signal is equal to { B ] in the defect angle analysis process_angle，B_heighObtaining a corresponding label vector Y ═ angle, h }, wherein angle is the angle of the defect, h is the depth of the defect, and B is the depth of the defect_angleAs a difference in the magnetic field of the X-axis component of the defect, B_heighIs the defect Z-axis component magnetic field difference.

In this embodiment, the sinusoidal excitation signal generation module is composed of a DDS chip AD9851 and a voltage-current conversion module, the three-phase rotating magnetic field detection probe is formed by winding a 0.4mm copper wire, the TMR sensor adopts a TMR1302S, the chip model used by the lock-in amplifier is AD620, the AD conversion module is AD633JRZ, and the chip model used by the extensible processing platform ZYNQ based on the defect identification module is XC7Z020-2CLG 484I.

The sinusoidal current excitation signal generation module is composed of a DDS chip AD9851 and a voltage-current conversion module. The control word programming is carried out on the AD9851 chip to realize the control of the amplitude, the frequency and the phase of the voltage signal, and the voltage signal is generated as follows:

V(t)＝sin(2*π*100*t)+sin(2*π*300*t)+sin(2*π*500*t)

then, the voltage signal is converted into a current signal, i.e., i (t) ═ sin (2 × pi × 100 × t) + sin (2 × pi × 300 × t) + sin (2 × pi × 500 × t), by a voltage-current conversion module built by the operational amplifier LM358, and the output waveform is shown in fig. 3; the signal is used for driving a three-phase rotating magnetic field probe at the later stage.

The three-phase rotating magnetic field detection probe consists of three rectangular enameled wire coils which form an angle of 120 degrees with each other, wherein the number of turns of each coil is 500. The wire diameter of the enameled wire is 0.4 mm. The length of the wound rectangular enameled wire coil is 20mm, the width of the wound rectangular enameled wire coil is 10mm, and after the probe coil passes through the current signal, an electromagnetic field is generated to perform induction on the pipeline.

The method for generating the three-phase rotating eddy current by the rotating magnetic field detection coil comprises the following steps: the three-phase coil is marked with ABC clockwise. The ABC three phases adopt a time-sharing multiplexing scanning mode, namely AB two-phase scanning is firstly carried out, then BC two-phase scanning is carried out, and finally CA two-phase scanning is carried out, so that the scanning of the whole circumference of the three-phase probe is completed. A rectangular coordinate system is established at the part of the AB two-phase scanning probe, as shown in fig. 4, in the two 120-degree angles of AB, a step scanning mode with one step length every 30 degrees is adopted, and the included angles between the direction of the resultant magnetic field and the Y axis are respectively 0 degree, 30 degrees, 60 degrees, 90 degrees and 120 degrees. Wherein

Is (0, 1),

has a direction vector of

By adjusting

The direction of the resultant magnetic field is controlled by the mode length of (2) to calculate

And

the required modular length is derived by the following steps:

(1) is provided with

(2) When the resultant current direction is 30 DEG to the Y axis, the resultant vector is

(3) Due to the fact that

Using the method of undetermined coefficientsb＝1，

Namely, it is

(4) Because a and b are

Is the amplitude of the excitation current, so in this case to generate a magnetic field at 30 ° to the Y axis, one sets

I_b＝1*sin(2*π*100*t)+1*sin(2*π*300*t)+1*sin(2*π*500*t)。

By analogy, the relationship between the magnetic field generated at each angle between the two phases AB and the amplitudes of the two phases AB is shown in table 1:

the resultant magnetic field derivation for the remaining BC, CA phases is the same as described above.

The TMR giant magneto-resistance sensor is placed at the center of the coil, and the height from the surface of the test piece is 2 mm. When the detection probe is placed above the test piece, the TMR giant magneto-resistance sensor captures the change of the magnetic field signal, and the acquired analog signal is input into the signal conditioning module.

The signal conditioning circuit mainly comprises a lock-in amplifier module built by AD620 and an AD conversion module built by AD633JRZ as shown in FIG. 5. The output signal of the TMR giant magneto-resistance sensor is used as the signal input end of the lock-in amplifier module, and the signal generated by the sine excitation signal generation module is used as the reference signal end of the lock-in amplifier module. The signal output by the lock-in amplifier module is a filtered signal, then the filtered signal is input into the AD conversion module, the signal is converted from an analog value into a digital value, and then the digital value is sent to the ZYNQ-based defect identification module.

The metal defect detection method of the multi-frequency rotating magnetic field based on phase-locked amplification comprises the following steps:

step 1, generating a sinusoidal current excitation signal through a sinusoidal excitation signal generation module, and generating a three-phase rotating magnetic field through a rotating magnetic field detection coil in a detection probe so as to induce a three-phase rotating eddy current on the surface of a test piece to be detected; the TMR giant magnetoresistance sensor in the detection probe detects the magnetic field change caused by the rotating magnetic field detection coil and the test piece to be detected, and transmits the detected voltage signal to the signal processing module;

step 2, the signal processing module carries out frequency-selective filtering and analog-to-digital conversion on the detected voltage signal and inputs the voltage signal to the defect data identification module;

step 3, processing the data output by the AD conversion module by an extensible processing platform ZYNQ based on the defect data identification module by adopting a normalized dynamic threshold method, normalizing the metal data of the pipeline, calculating a dynamic threshold for identifying the defect characteristics of the pipeline, and filtering abnormal data while obtaining the defect data characteristics, wherein the specific method comprises the following steps:

step 3.1: establishing initial threshold lambda of pipe metal defect angle and defect height_angle，λ_heigh；

Step 3.2: the acquired X-axis magnetic field data and Z-axis magnetic field data are normalized by depending on the background magnetic field, and the following formula is shown:

wherein, B_{x_back}For detecting the x-component of the magnetic field in the absence of defects, B_xFor detecting the X-axis component of the magnetic field when there is a defect, B_{z_back}For detecting the Z-component of the magnetic field in the absence of defects, B_zDetecting a Z-axis component of the magnetic field when the defect is detected;

step 3.3: will come intoNormalized B_angle，B_heighAnd a threshold lambda_angle，λ_heighBy comparison, when B_angle≥λ_angleWhen it is, B_angleMarking the data as valid data for expressing the angle of the metal defect of the pipeline; all the same thing as B_heigh≥λ_heighWhen it is, B_heighMarking the data as valid data for expressing the depth of the metal defect of the pipeline;

step 4, the defect angle recognition module recognizes the angle of the defect through an angle recognition algorithm to obtain a final defect classification result, as shown in fig. 6, the specific method is as follows:

step 4.1: operating a metal defect detection device in a pipeline with known position and size, and recording a sample data set X of experimental data₁And sample tag set Y₁(ii) a The sample data set comprises data of pipe metal defect angles and defect heights; the sample label set comprises sample labels of the corresponding relation between the actual defect angle and the defect height of the sample data set pipeline metal defect angle data and the defect height data respectively;

step 4.2: establishing a pipeline defect detection simulation model by using finite element simulation software, and recording a sample data set X of simulation data₂And sample tag set Y₂；

Step 4.3: randomly selecting a sample; by applying to a sample data set X₁And X₂In the method, random sampling with replacement is carried out to construct a certain number of new test sample sets, wherein random sampling with replacement is carried out for M times, and N data are sampled each time, thereby constructing M test sample sets X_m；

Step 4.4: randomly selecting a sample label; from the sample tag set Y₁And Y₂Random sampling without putting back is carried out, and a certain number of new test sample label sets are constructed; randomly sampling k sample labels without putting back, calculating the information gain of the samples, and then selecting an optimal label;

step 4.5: constructing a decision tree; determining node selection of the decision tree by using an information entropy method, namely selecting the node sequence of the decision tree by calculating the information entropy of each node so as to construct the decision tree;

step 4.6: cutting a decision tree; cutting the decision tree established in the step 4.5 by using a cost complexity pruning method;

step 4.7: random forest voting classification is carried out; repeating the steps 4.5-4.6 to construct L decision trees, training the L decision trees aiming at the test samples respectively to obtain L results, and forming a voting result set T by the results_lAnd calculating a voting result T_iAnd T_jEuclidean distance d (T) therebetween_i，T_j) The following formula shows:

wherein, T_iFor the ith voting result, T_jFor the jth voting result, i ≠ 1, …, L, j ≠ 1, …, L, and i ≠ j; deeply cutting the decision tree corresponding to the classification result with the maximum distance, and then classifying again until the distance of the voting result is less than a threshold value d_stopAnd stopping cutting to obtain a final defect classification result.

Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit of the corresponding technical solutions and scope of the present invention as defined in the appended claims.