Metal part detection system and method based on directional pulsed eddy current

Abstract

The invention discloses a directional pulsed eddy current-based metal part detection system and method, the system comprises a directional pulsed eddy current probe, the directional pulsed eddy current probe is used for scanning a test piece to be tested, and defect information of the test piece to be tested is obtained through an eddy current effect; the defect information comprises the depth, the length and the width of the defect. The induced magnetic field is fully detected through the three mutually orthogonal annular detection coils, and the precision of eddy current detection is improved. According to the method, defect information of a defect position of a tested piece is calibrated, and temperature information is fused to establish a three-dimensional curved surface analysis formula, so that the defect information of the specific defect position of the tested piece can be obtained by importing real-time temperature information and a sensed voltage value.

Description

Technical Field

[0001] The present invention belongs to the technical field of eddy current detection, and in particular relates to a metal component detection system and method based on directional pulsed eddy current. Background Art

[0002] The pulse eddy current detection method has the advantages of many characteristic quantities, wide spectrum, fast detection speed, large detection depth, high sensitivity, and easy quantification. The existing eddy current probes mainly include impedance detection type eddy current probes and field quantity detection type eddy current probes. Impedance detection type eddy current probes generally use cylindrical coils, which are generally used as both excitation units and detection units. The characteristics of the object being tested are determined by measuring the change in the equivalent impedance of the coil. The equivalent impedance of the coil is generally detected through the measurement bridge in the eddy current detection system, and the measurement bridge is not easy to balance under pulse excitation and multi-frequency excitation conditions.

[0003] The equivalent impedance of the coil is generally detected through the measuring bridge in the eddy current testing system. For example, based on the circuit-eddy current loop model of the "equivalence principle", the induced eddy current in the test piece is equivalent to a ring current of finite size, so that the coupling relationship between the excitation coil and the test piece is equivalent to the coupling relationship between the excitation coil and the current loop, and the complex electromagnetic field model is simplified to a circuit model-coil coupling model. The measuring bridge of this model is not easy to balance under pulse excitation and multi-frequency excitation conditions.

[0004] A Chinese patent with a patent publication number of CN103196996A proposes an orthogonal combination probe, each of which includes an excitation coil and two induction coils.

[0005] The induction coil of the probe senses a plane perpendicular to the excitation coil. In other words, assuming that the excitation coil is placed horizontally on the plane, the orthogonal induction coil can achieve defect detection with the same accuracy in both the vertical and horizontal directions of the two-dimensional plane, and ultimately improve the sensitivity to tiny defects on the test piece. However, the induced magnetic field structure is a three-dimensional magnetic field, and the probe in the above scheme cannot determine the vector characteristics of the induced magnetic field, that is, it does not have anti-interference characteristics.

[0006] In addition, during the eddy current detection process, the temperature of the excitation coil will increase after long-term use. The design of eddy current sensors in the prior art has ignored the impact of temperature increase on eddy current detection and has not performed quantitative analysis, which may cause the impedance matching of the detection device to lose its original function and fail to obtain accurate defect information. Summary of the invention

[0007] In view of the shortcomings of the pulsed eddy current detection technology in the prior art, a metal component detection system based on directional pulsed eddy current is provided, wherein the system comprises a directional pulsed eddy current probe, wherein the directional pulsed eddy current probe is used to scan a test piece and obtain defect information of the test piece through the eddy current effect, wherein the defect information comprises the depth, length and width of the defect; The directional pulsed eddy current probe comprises: The excitation coil comprises a plurality of circular rectangular strip metal frames and corresponding wires, wherein the wires are wound around the circular rectangular strip metal frames; a plurality of circular rectangular strip metal frames are coaxially stacked along the central axis direction to strengthen the induced magnetic field in the z-axis direction; The plane where the ring-shaped rectangular metal frame is located is taken as the xoz plane, and the central axis of the ring-shaped rectangular metal frame is taken as the y axis, and a three-dimensional rectangular coordinate system o-xyz is established in space, with o as the origin; An induction coil, the induction coil is arranged in the excitation coil, and includes an x-axis induction coil, a y-axis induction coil and a z-axis induction coil, and is used to receive the induced magnetic field in the x-axis direction, the y-axis direction and the z-axis direction respectively, the induction coil includes a wire, a metal ball and three circular strip metal frames orthogonal to each other in space, the wire is respectively wound around the three circular strip metal frames to form the x-axis induction coil, the y-axis induction coil and the z-axis induction coil, and the induction coil is fixed in a cavity formed by the overlapping of the circular strip metal frames; The length of the excitation coil: the width of the excitation coil: the depth of the excitation coil: the maximum radius of the induction coil = 2.7:2:1.4:1~4:3:2:1; Magnetic field signal sensed by the induction coil It is expressed as: ; ; ; in, are the azimuth and elevation angles of the induced magnetic field, It represents the magnetic field signal in the x-axis direction sensed by the x-axis induction coil. It represents the induced magnetic field in the y-axis direction sensed by the y-axis induction coil. Represents the induced magnetic field in the z-axis direction sensed by the z-axis induction coil; The temperature sensor is connected to the excitation coil signal and is used to detect the temperature of the excitation coil.

[0008] Furthermore, the system also includes: A power amplifier, electrically connected to the excitation coil of the directional pulsed eddy current probe and performing square wave excitation; A signal conditioning circuit is electrically connected to the directional pulsed eddy current probe to obtain the temperature information of the temperature sensor and the voltage value corresponding to the magnetic field strength of the induction coil; An A / D conversion module is connected to the signal conditioning circuit signal; A main controller, connected to the A / D conversion module signal; A DDS signal generator is signal-connected to the main controller and the power amplifier; The power supply module is electrically connected to the temperature sensor, power amplifier, main controller and signal conditioning circuit of the directional pulsed eddy current probe and provides direct current voltage.

[0009] Furthermore, The main controller is an STM32 single-chip microcomputer; The DDS signal generator is an AD9910 chip; The power amplifier uses OPA549 as a driver chip; The signal conditioning circuit is a dual-channel operational amplifier chip LM358; The A / D conversion module is an AD7606 chip; The temperature sensor is an NTC thermistor.

[0010] Furthermore, the present invention also provides a metal component detection method based on directional pulsed eddy current, the method comprising: Step S1: setting the initial excitation coil temperature of the test, and in the absence of a test piece, increasing the excitation coil temperature by a set threshold value, and obtaining the voltage value obtained by the induction coil at each excitation coil temperature, thereby performing calibration and obtaining the calibration temperature; Step S2: Select several standard test pieces, the standard test pieces are preset with several defects of different sizes and shapes, and obtain the voltage values ​​corresponding to the magnetic fields of the defects of the test pieces in different directions in three-dimensional space at the calibration temperature in step S1 through a directional pulsed eddy current probe, so as to complete further calibration and obtain calibration defect information; Step S3: constructing a three-dimensional surface analytical expression according to the calibration temperature in step S1 and the calibration defect information in step S2; Step S4: Enter the actual measurement environment, obtain the real-time excitation coil temperature and the voltage value corresponding to the magnetic field in different directions sensed by the induction coil in three-dimensional space, import it into the three-dimensional surface analytical formula in step S3, and obtain the defect information of the defect of the test piece.

[0011] Furthermore, step S1 is specifically as follows: The initial temperature of the tested excitation coil is set to 22°C, and the temperature of the excitation coil is gradually increased with the set threshold as the temperature increment. During this process, the voltage values ​​of the induction coil in different directions in the three-dimensional space are obtained as reference voltage values, and the corresponding excitation coil temperatures and reference voltage values ​​are calibrated one by one to obtain the calibration temperature.

[0012] Furthermore, step S2 is specifically as follows: A square steel plate is selected as the standard test piece. The material of the square steel plate is steel1008 carbon steel. Several defects with different defect information are processed on the square steel plate. According to the calibration temperature in step S1, the defects on the square steel plate are scanned in sequence by a directional pulsed eddy current probe, and the voltage values ​​sensed by the induction coil in different directions in three-dimensional space are obtained as induced voltage values. Further calibration is completed according to the defect information to obtain the calibrated defect information.

[0013] Furthermore, step S3 is specifically as follows: The induced voltage value obtained by detection is differentiated from the reference voltage value to obtain the voltage difference value of the induction coil in different directions in three-dimensional space; it is expressed as: ; ; ; in, Represents the voltage difference value of the x-axis induction coil in the x-axis direction in three-dimensional space, Represents the voltage difference value of the y-axis induction coil in the y-axis direction in three-dimensional space, As the voltage difference value of the z-axis induction coil in the z-axis direction in three-dimensional space, Indicates the calibration temperature, Indicates the calibrated defect depth, Indicates the calibration defect length, Indicates the calibration defect width; The voltage differential values ​​of n groups of induction coils in different directions in three-dimensional space are obtained, and a three-dimensional surface analytical expression is established based on this, which is: According to the calibration temperature, calibration defect depth, calibration defect length, and calibration defect width, several groups of function curves with defect depth, defect length, and defect width as independent variables and corresponding voltage difference values ​​as dependent variables are fitted by the least square method. The function curves include the function curve in the x-axis direction. , function curve in the y-axis direction And the function curve in the z-axis direction ; The function curve is expressed as: ; in, , and are the least square fitting coefficients of the corresponding arbitrary defect depths in the set mth group of polynomials based on the voltage difference values ​​of the i-th group of induction coils in the x-axis direction, the y-axis direction and the z-axis direction, respectively. , and are the least square fitting coefficients of the corresponding arbitrary defect lengths in the set mth group of polynomials based on the voltage difference values ​​of the i-th group of induction coils in the x-axis direction, the y-axis direction and the z-axis direction, respectively. , and are respectively the least square fitting coefficients of the corresponding arbitrary defect widths in the set mth group of polynomials based on the voltage difference values ​​of the i-th group of induction coils in the x-axis direction, the y-axis direction and the z-axis direction, where m is a preset accuracy value; They represent arbitrary defect depth, arbitrary defect length and arbitrary defect width respectively; The m-th-order terms representing arbitrary defect depth, arbitrary defect length and arbitrary defect width are used for nonlinear fitting. ; The coefficients of each polynomial are obtained by using the least squares surface fitting algorithm based on the rectangular domain; the above function curve is fitted into a continuous and smooth three-dimensional surface to obtain the three-dimensional surface analytical expression , and ; in, For any temperature.

[0014] Furthermore, step S4 is specifically as follows: Enter the actual measurement environment, obtain the current excitation coil temperature through the temperature sensor, obtain the voltage values ​​in different directions in the three-dimensional space corresponding to the defects of the test piece through the induction coil, import the excitation coil temperature and the voltage values ​​in different directions in the three-dimensional space into the three-dimensional surface analytical formula in step S3, and obtain the defect information of the defect under test.

[0015] Furthermore, the process of obtaining the defect information of the defect under test in step S4 is implemented by a CNN convolutional network; specifically, as follows: Step S4A: Obtain the voltage differential value and the corresponding calibration temperature and calibration defect information as training set data, import the real-time induced voltage value and the real-time excitation coil temperature, obtain the voltage differential value through the calibration temperature as the input of the CNN convolutional network, the CNN convolutional network outputs the predicted defect information, and trains the CNN convolutional network through the training set data. During the training process, the learning rate of the CNN convolutional network is optimized through the particle swarm algorithm to obtain the optimized CNN convolutional network.

[0016] Furthermore, in step S4A, the CNN convolutional network is optimized by using a particle swarm algorithm, specifically: Step S4A1: Set the induced voltage difference value to a randomly generated cluster, formalized as a matrix with D columns and N rows , expressed as: ; in, is the number of data points corresponding to the learning rate, is the feature dimension, and denote the lower and upper bounds of the solution space, respectively. A random number between 0 and 1; Step S4A2: Divide the training set data into two parts, one for global exploration and the other for local detection; update the position of the individual through global exploration or local detection; Global exploration is expressed as: ; in, Indicates In the iteration individual location, Indicates In the iteration The individual position of represents a random number vector based on Gaussian distribution, represents term-by-term multiplication, Represents a number randomly selected from [0,1]; represents the current optimal solution, represents an individual randomly selected from a group of elites in the population; Represents a number randomly selected from [0,1]; Indicates the location of an average individual; represents the training set data used for global exploration, Represents the training set data used for local detection; Indicated in The centroid position of the entire group in iterations; During the local detection process, the iterative process of ordinary individuals is expressed as: ; ; in, represents the individual after local detection iteration, is the maximum number of iterations, Represents a dynamic adjustment function; Step S4A3: Obtain the optimal individual position through a preset fitness function and use it as the learning rate of the CNN convolutional network.

[0017] The positive and progressive effects of the present invention are:

[0018] 1) The present invention provides a new type of probe for eddy current detection. The three mutually orthogonal annular detection coils of the probe can fully detect the vector characteristics of the induced magnetic field, thereby improving the accuracy and anti-interference of eddy current detection. Furthermore, the excitation coils are adaptively changed to mutually superimposed rectangular coils, so that the eddy current generated at the defect can have a certain intensity in different directions in three-dimensional space at the position of the induction coil.

[0019] 2) Furthermore, the present invention also provides a method for using the above-mentioned probe, including calibrating the defect information of the defect position, and fusing the temperature information to establish a three-dimensional surface analytical expression, so that the defect information of the defect location of the test piece can be obtained by importing real-time temperature information and perceived voltage values; based on this, the present invention also obtains defect information through a CNN convolutional network based on a particle swarm optimization algorithm, thereby improving the speed of data convergence and the accuracy of classification while making full use of the calibration data. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The schematic diagram of a metal component detection system based on directional pulsed eddy current of the present invention.

[0021] Figure 2 The present invention is a schematic diagram of the probe structure of a metal component detection system based on directional pulsed eddy current.

[0022] Figure 3 : is the distribution diagram of induced eddy current on the inspected surface of a conductor in one example.

[0023] Figure 4 FIG. 1 is a distribution diagram of the eddy current field in which the induced eddy current is perpendicular to the defect in an example.

[0024] Figure 5 FIG. 1 is a diagram showing the eddy current field distribution of an induced eddy current parallel to a defect in an example.

[0025] Figure 6 A diagram of a simulation verification model in another example.

[0026] Figure 7 For Figure 6 Schematic diagram of the magnetic field distribution after excitation of the simulation verification model diagram.

[0027] Figure 8 It is a front view of defect distribution of a steel plate used as a test piece in one example of the present invention.

[0028] Fig. 9 It is a side view of defect distribution of a steel plate used as a test piece in one example of the present invention. DETAILED DESCRIPTION

[0029] The following describes the embodiments of the present invention through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention.

[0030] Reference Figure 1~Figure 2 , a metal component detection system based on directional pulsed eddy current, the system comprising a directional pulsed eddy current probe, the directional pulsed eddy current probe is used to scan a test piece, and obtain defect information of the test piece through eddy current effect, the defect information including the depth, length and width of the defect; The directional pulsed eddy current probe comprises: The excitation coil comprises a plurality of circular rectangular strip metal frames and corresponding wires, wherein the wires are wound around the circular rectangular strip metal frames; a plurality of circular rectangular strip metal frames are coaxially stacked along the central axis direction to strengthen the induced magnetic field in the z-axis direction; The plane where the ring-shaped rectangular metal frame is located is taken as the xoz plane, and the central axis of the ring-shaped rectangular metal frame is taken as the y axis, and a three-dimensional rectangular coordinate system o-xyz is established in space, with o as the origin; An induction coil, the induction coil is arranged in the excitation coil, and includes an x-axis induction coil, a y-axis induction coil and a z-axis induction coil, and is used to receive the induced magnetic field in the x-axis direction, the y-axis direction and the z-axis direction respectively, the induction coil includes a wire, a metal ball and three circular strip metal frames orthogonal to each other in space, the wire is respectively wound around the three circular strip metal frames to form the x-axis induction coil, the y-axis induction coil and the z-axis induction coil, and the induction coil is fixed in a cavity formed by the overlapping of the circular strip metal frames; The length of the excitation coil: the width of the excitation coil: the depth of the excitation coil: the maximum radius of the induction coil = 2.7:2:1.4:1~4:3:2:1; Magnetic field signal sensed by the induction coil It is expressed as: ; ; ; in, are the azimuth and elevation angles of the induced magnetic field, It represents the magnetic field signal in the x-axis direction sensed by the x-axis induction coil. It represents the induced magnetic field in the y-axis direction sensed by the y-axis induction coil. Represents the induced magnetic field in the z-axis direction sensed by the z-axis induction coil; The temperature sensor is connected to the excitation coil signal and is used to detect the temperature of the excitation coil.

[0031] Furthermore, the system also includes: A power amplifier, electrically connected to the excitation coil of the directional pulsed eddy current probe and performing square wave excitation; A signal conditioning circuit is electrically connected to the directional pulsed eddy current probe to obtain the temperature information of the temperature sensor and the voltage value corresponding to the magnetic field strength of the induction coil; An A / D conversion module is connected to the signal conditioning circuit signal; A main controller, connected to the A / D conversion module signal; A DDS signal generator is signal-connected to the main controller and the power amplifier; The power supply module is electrically connected to the temperature sensor, power amplifier, main controller and signal conditioning circuit of the directional pulsed eddy current probe and provides direct current voltage.

[0032] Furthermore, The main controller is an STM32 single-chip microcomputer; The DDS signal generator is an AD9910 chip; The power amplifier uses OPA549 as a driver chip; The signal conditioning circuit is a dual-channel operational amplifier chip LM358; The A / D conversion module is an AD7606 chip; The temperature sensor is an NTC thermistor.

[0033] Furthermore, the present invention also provides a metal component detection method based on directional pulsed eddy current, the method comprising: Step S1: setting the initial excitation coil temperature of the test, and in the absence of a test piece, increasing the excitation coil temperature by a set threshold value, and obtaining the voltage value obtained by the induction coil at each excitation coil temperature, thereby performing calibration and obtaining the calibration temperature; Step S2: Select several standard test pieces, the standard test pieces are preset with several defects of different sizes and shapes, and obtain the voltage values ​​corresponding to the magnetic fields of the defects of the test pieces in different directions in three-dimensional space at the calibration temperature in step S1 through a directional pulsed eddy current probe, so as to complete further calibration and obtain calibration defect information; Step S3: constructing a three-dimensional surface analytical expression according to the calibration temperature in step S1 and the calibration defect information in step S2; Step S4: Enter the actual measurement environment, obtain the real-time excitation coil temperature and the voltage value corresponding to the magnetic field in different directions sensed by the induction coil in three-dimensional space, import it into the three-dimensional surface analytical formula in step S3, and obtain the defect information of the defect of the test piece.

[0034] Furthermore, step S1 is specifically as follows: The initial temperature of the tested excitation coil is set to 22°C, and the temperature of the excitation coil is gradually increased with the set threshold as the temperature increment. During this process, the voltage values ​​of the induction coil in different directions in the three-dimensional space are obtained as reference voltage values, and the corresponding excitation coil temperatures and reference voltage values ​​are calibrated one by one to obtain the calibration temperature.

[0035] Furthermore, step S2 is specifically as follows: A square steel plate is selected as the standard test piece. The material of the square steel plate is steel1008 carbon steel. Several defects with different defect information are processed on the square steel plate. According to the calibration temperature in step S1, the defects on the square steel plate are scanned in sequence by a directional pulsed eddy current probe, and the voltage values ​​sensed by the induction coil in different directions in three-dimensional space are obtained as induced voltage values. Further calibration is completed according to the defect information to obtain the calibrated defect information.

[0036] Furthermore, step S3 is specifically as follows: The induced voltage value obtained by detection is differentiated from the reference voltage value to obtain the voltage difference value of the induction coil in different directions in three-dimensional space; it is expressed as: ; ; ; in, Represents the voltage difference value of the x-axis induction coil in the x-axis direction in three-dimensional space, Represents the voltage difference value of the y-axis induction coil in the y-axis direction in three-dimensional space, As the voltage difference value of the z-axis induction coil in the z-axis direction in three-dimensional space, Indicates the calibration temperature, Indicates the calibrated defect depth, Indicates the calibration defect length, Indicates the calibration defect width; The voltage differential values ​​of n groups of induction coils in different directions in three-dimensional space are obtained, and a three-dimensional surface analytical expression is established based on this, which is: According to the calibration temperature, calibration defect depth, calibration defect length, and calibration defect width, several groups of function curves with defect depth, defect length, and defect width as independent variables and corresponding voltage difference values ​​as dependent variables are fitted by the least square method. The function curves include the function curve in the x-axis direction. , function curve in the y-axis direction And the function curve in the z-axis direction ; The function curve is expressed as: ; in, , and are the least square fitting coefficients of the corresponding arbitrary defect depths in the set mth group of polynomials based on the voltage difference values ​​of the i-th group of induction coils in the x-axis direction, the y-axis direction and the z-axis direction, respectively. , and are the least square fitting coefficients of the corresponding arbitrary defect lengths in the set mth group of polynomials based on the voltage difference values ​​of the i-th group of induction coils in the x-axis direction, the y-axis direction and the z-axis direction, respectively. , and are respectively the least square fitting coefficients of the corresponding arbitrary defect widths in the set mth group of polynomials based on the voltage difference values ​​of the i-th group of induction coils in the x-axis direction, the y-axis direction and the z-axis direction, where m is a preset accuracy value; They represent arbitrary defect depth, arbitrary defect length and arbitrary defect width respectively; The m-th-order terms representing arbitrary defect depth, arbitrary defect length and arbitrary defect width are used for nonlinear fitting. ; The coefficients of each polynomial are obtained by using the least squares surface fitting algorithm based on the rectangular domain; the above function curve is fitted into a continuous and smooth three-dimensional surface to obtain the three-dimensional surface analytical expression , and ; in, For any temperature.

[0037] Furthermore, step S4 is specifically as follows: Enter the actual measurement environment, obtain the current excitation coil temperature through the temperature sensor, obtain the voltage values ​​in different directions in the three-dimensional space corresponding to the defects of the test piece through the induction coil, import the excitation coil temperature and the voltage values ​​in different directions in the three-dimensional space into the three-dimensional surface analytical formula in step S3, and obtain the defect information of the defect under test.

[0038] Furthermore, the process of obtaining the defect information of the defect under test in step S4 is implemented by a CNN convolutional network; specifically, as follows: Step S4A: Obtain the voltage differential value and the corresponding calibration temperature and calibration defect information as training set data, import the real-time induced voltage value and the real-time excitation coil temperature, obtain the voltage differential value through the calibration temperature as the input of the CNN convolutional network, the CNN convolutional network outputs the predicted defect information, and trains the CNN convolutional network through the training set data. During the training process, the learning rate of the CNN convolutional network is optimized through the particle swarm algorithm to obtain the optimized CNN convolutional network.

[0039] Furthermore, in step S4A, the CNN convolutional network is optimized by using a particle swarm algorithm, specifically: Step S4A1: Set the induced voltage difference value to a randomly generated cluster, formalized as a matrix with D columns and N rows , expressed as: ; in, is the number of data points corresponding to the learning rate, is the feature dimension, and denote the lower and upper bounds of the solution space, respectively. A random number between 0 and 1; Step S4A2: Divide the training set data into two parts, one for global exploration and the other for local detection; update the position of the individual through global exploration or local detection; Global exploration is expressed as: ; in, Indicates In the iteration individual location, Indicates In the iteration The individual position of represents a random number vector based on Gaussian distribution, represents term-by-term multiplication, Represents a number randomly selected from [0,1]; represents the current optimal solution, represents an individual randomly selected from a group of elites in the population; Represents a number randomly selected from [0,1]; Indicates the location of an average individual; represents the training set data used for global exploration, Represents the training set data used for local detection; Indicated in The centroid position of the entire group in iterations; During the local detection process, the iterative process of ordinary individuals is expressed as: ; ; in, represents the individual after local detection iteration, is the maximum number of iterations, Represents a dynamic adjustment function; Step S4A3: Obtain the optimal individual position through a preset fitness function and use it as the learning rate of the CNN convolutional network.

[0040] It should be noted that, refer to Figure 3~Figure 5 When the probe in the prior art is not disturbed by other external conditions, the distribution of the induced eddy current is determined by the coil structure and its scanning method.

[0041] For example, when the rectangular coil is placed tangentially above the conductor and scanned, the pulse excitation signal amplitude is set to 2.52V, the frequency is 100Hz, the duty cycle is 50%, and the lift-off height is 2mm, and the induced eddy current of the conductor at a certain moment is analyzed.

[0042] When there are defects in the material being tested, the uniform eddy current field will be blocked by the defects. In actual testing situations, the scanning detection of the probe has two directions: X and Y. Figure 3 As shown, the X direction is parallel to the excitation current in the excitation coil, and the Y direction is parallel to the magnetic induction line generated by the excitation coil. When the eddy current is perpendicular to the defect, the defect has the greatest disturbance on the eddy current field, as shown in Figure 4 As shown; when the eddy current is parallel to the defect, the disturbance of the defect to the eddy current field is minimal, as shown Figure 5 As shown, it can be seen that the probe has a sensitive direction during the scanning detection process, that is, when the defect is perpendicular to the eddy current, the detection sensitivity of the defect is greater.

[0043] Further simulation verification of the above example is carried out, refer to Figure 6 The smaller rectangular block can be regarded as a conductor, and the larger rectangular block can be regarded as the space where the conductor is located; specifically, the excitation signal of the pulse eddy current is set to a periodic square wave pulse current with adjustable amplitude, frequency and duty cycle. The induced eddy current inside the conductor to be measured will only appear near the rising and falling edges of the signal. When establishing the finite element model, an exponentially rising current source is used. The analog excitation signal is expressed as: ; In the formula, is the time constant, represents the initial amplitude of the excitation signal of the pulsed eddy current, is a unit step function, >0 o'clock =1.

[0044] For a typical eddy current problem, the solution region can be Eddy zone and non-eddy current zone Two parts, using The potential function method is used to solve the problem, where A represents the vector magnetic potential, represents the scalar potential. Ignoring the displacement current, considering the uniqueness of the solution, and incorporating the Coulomb gauge condition, it is formalized as: ,in, represents the divergence operator, then the control equations of each part are as follows: In the eddy zone Inside: ; ; In the formula, the solution area ; is the eddy current region, which contains a conductive medium with a conductivity of , the magnetic permeability is .

[0045] In the non-eddy zone Inside: Since there is no conductive medium, the conductivity is zero, the equation in the non-eddy region will be simplified to: ; ; in, is the magnetic permeability of air.

[0046] right Figure 6 The simulation model in the simulation model applies a square wave pulse excitation of 500mA, frequency 976Hz, and duty cycle 50%. The waveform generated on the rising edge is as follows Figure 7 As shown, it can be seen that the directions of the excitation current and the induced current are basically consistent with the theory.

[0047] Furthermore, in one example, referring to Figure 8 and Fig. 9 , step S2 is specifically as follows: Three square steel plates with a length × width × height = 200mm × 200mm × 10mm were selected as standard test pieces. A total of 9 strip defects with a 3×3 distribution at an equal interval of 70mm were processed in the center of the square steel plates. The widths of the 1st, 2nd and 3rd steel plates were 3mm, 6mm and 9mm respectively. The defect length of each steel plate was 10mm, 15mm and 20mm from the first row to the third row, and the depth was 1mm, 3mm and 5mm from the first column to the third column.

[0048] The present invention is described in detail above in conjunction with the embodiments of the drawings, and those skilled in the art can make various variations of the present invention according to the above description. Therefore, certain details in the embodiments should not constitute limitations on the present invention, and the present invention shall be protected by the scope defined by the attached claims.

Claims

1. A metal parts detection system based on directional pulsed eddy current, characterized in that: The system includes a directional pulsed eddy current probe, which is used to scan the test piece and obtain defect information of the test piece through the eddy current effect, wherein the defect information includes the depth, length and width of the defect; The directional pulsed eddy current probe comprises: The excitation coil comprises a plurality of circular rectangular strip metal frames and corresponding wires, wherein the wires are wound around the circular rectangular strip metal frames; a plurality of circular rectangular strip metal frames are coaxially stacked along the central axis direction to strengthen the induced magnetic field in the z-axis direction; The plane where the ring-shaped rectangular metal frame is located is taken as the xoz plane, and the central axis of the ring-shaped rectangular metal frame is taken as the y axis, and a three-dimensional rectangular coordinate system o-xyz is established in space, with o as the origin; An induction coil, the induction coil is arranged in the excitation coil, and includes an x-axis induction coil, a y-axis induction coil and a z-axis induction coil, and is used to receive the induced magnetic field in the x-axis direction, the y-axis direction and the z-axis direction respectively, the induction coil includes a wire, a metal ball and three circular strip metal frames orthogonal to each other in space, the wire is respectively wound around the three circular strip metal frames to form the x-axis induction coil, the y-axis induction coil and the z-axis induction coil, and the induction coil is fixed in a cavity formed by the overlapping of the circular strip metal frames; The length of the excitation coil: the width of the excitation coil: the depth of the excitation coil: the maximum radius of the induction coil = 2.7:2:1.4:1~4:3:2:1; Magnetic field signal sensed by the induction coil It is expressed as: ； ； ； in, are the azimuth and elevation angles of the induced magnetic field, It represents the magnetic field signal in the x-axis direction sensed by the x-axis induction coil. It represents the induced magnetic field in the y-axis direction sensed by the y-axis induction coil. Represents the induced magnetic field in the z-axis direction sensed by the z-axis induction coil; The temperature sensor is connected to the excitation coil signal and is used to detect the temperature of the excitation coil.

2. A metal component detection system based on directional pulsed eddy current according to claim 1, characterized in that: The system further comprises: A power amplifier, electrically connected to the excitation coil of the directional pulsed eddy current probe and performing square wave excitation; A signal conditioning circuit is electrically connected to the directional pulsed eddy current probe to obtain the temperature information of the temperature sensor and the voltage value corresponding to the magnetic field strength of the induction coil; An A / D conversion module is connected to the signal conditioning circuit signal; A main controller, connected to the A / D conversion module signal; A DDS signal generator is signal-connected to the main controller and the power amplifier; The power supply module is electrically connected to the temperature sensor, power amplifier, main controller and signal conditioning circuit of the directional pulsed eddy current probe and provides direct current voltage.

3. A metal component detection system based on directional pulsed eddy current according to claim 2, characterized in that: The main controller is an STM32 single-chip microcomputer; The DDS signal generator is an AD9910 chip; The power amplifier uses OPA549 as a driver chip; The signal conditioning circuit is a dual-channel operational amplifier chip LM358; The A / D conversion module is an AD7606 chip; The temperature sensor is an NTC thermistor.

4. A metal component detection method based on directional pulsed eddy current, used in a metal component detection system based on directional pulsed eddy current as claimed in any one of claims 1 to 3, characterized in that: The method comprises: Step S1: setting the initial excitation coil temperature of the test, and in the absence of a test piece, increasing the excitation coil temperature by a set threshold value, and obtaining the voltage value obtained by the induction coil at each excitation coil temperature, thereby performing calibration and obtaining the calibration temperature; Step S2: Select several standard test pieces, the standard test pieces are preset with several defects of different sizes and shapes, and obtain the voltage values ​​corresponding to the magnetic fields of the defects of the test pieces in different directions in three-dimensional space at the calibration temperature in step S1 through a directional pulsed eddy current probe, so as to complete further calibration and obtain calibration defect information; Step S3: constructing a three-dimensional surface analytical expression according to the calibration temperature in step S1 and the calibration defect information in step S2; Step S4: Enter the actual measurement environment, obtain the real-time excitation coil temperature and the voltage value corresponding to the magnetic field in different directions sensed by the induction coil in three-dimensional space, import it into the three-dimensional surface analytical formula in step S3, and obtain the defect information of the defect of the test piece.

5. A metal component detection method based on directional pulsed eddy current according to claim 4, characterized in that: Step S1 is specifically as follows: The initial temperature of the tested excitation coil is set to 22°C, and the temperature of the excitation coil is gradually increased with the set threshold as the temperature increment. During this process, the voltage values ​​of the induction coil in different directions in the three-dimensional space are obtained as reference voltage values, and the corresponding excitation coil temperatures and reference voltage values ​​are calibrated one by one to obtain the calibration temperature.

6. A metal component detection method based on directional pulsed eddy current according to claim 5, characterized in that: Step S2 is specifically as follows: A square steel plate is selected as the standard test piece. The material of the square steel plate is steel1008 carbon steel. Several defects with different defect information are processed on the square steel plate. According to the calibration temperature in step S1, the defects on the square steel plate are scanned in sequence by a directional pulsed eddy current probe, and the voltage values ​​sensed by the induction coil in different directions in three-dimensional space are obtained as induced voltage values. Further calibration is completed according to the defect information to obtain the calibrated defect information.

7. A metal component detection method based on directional pulsed eddy current according to claim 6, characterized in that: Step S3 is specifically as follows: The induced voltage value obtained by detection is differentiated from the reference voltage value to obtain the voltage difference value of the induction coil in different directions in three-dimensional space; it is expressed as: ； ； ； in, Represents the voltage difference value of the x-axis induction coil in the x-axis direction in three-dimensional space, Represents the voltage difference value of the y-axis induction coil in the y-axis direction in three-dimensional space, As the voltage difference value of the z-axis induction coil in the z-axis direction in three-dimensional space, Indicates the calibration temperature, Indicates the calibrated defect depth, Indicates the calibration defect length, Indicates the calibration defect width; The voltage differential values ​​of n groups of induction coils in different directions in three-dimensional space are obtained, and a three-dimensional surface analytical expression is established based on this, which is: According to the calibration temperature, calibration defect depth, calibration defect length, and calibration defect width, several groups of function curves with defect depth, defect length, and defect width as independent variables and corresponding voltage difference values ​​as dependent variables are fitted by the least square method. The function curves include the function curve in the x-axis direction. , function curve in the y-axis direction And the function curve in the z-axis direction ; The function curve is expressed as: ； in, , and are the least square fitting coefficients of the corresponding arbitrary defect depths in the set mth group of polynomials based on the voltage difference values ​​of the i-th group of induction coils in the x-axis direction, the y-axis direction and the z-axis direction, respectively. , and are the least square fitting coefficients of the corresponding arbitrary defect lengths in the set mth group of polynomials based on the voltage difference values ​​of the i-th group of induction coils in the x-axis direction, the y-axis direction and the z-axis direction, respectively. , and are respectively the least square fitting coefficients of the corresponding arbitrary defect widths in the set mth group of polynomials based on the voltage difference values ​​of the i-th group of induction coils in the x-axis direction, the y-axis direction and the z-axis direction, where m is a preset accuracy value; They represent arbitrary defect depth, arbitrary defect length and arbitrary defect width respectively; The m-th-order terms representing arbitrary defect depth, arbitrary defect length and arbitrary defect width are used for nonlinear fitting. ; The coefficients of each polynomial are obtained by using the least squares surface fitting algorithm based on the rectangular domain; the above function curve is fitted into a continuous and smooth three-dimensional surface to obtain the three-dimensional surface analytical expression , and ; in, For any temperature.

8. A metal component detection method based on directional pulsed eddy current according to claim 7, characterized in that: Step S4 is specifically as follows: Enter the actual measurement environment, obtain the current excitation coil temperature through the temperature sensor, obtain the voltage values ​​in different directions in the three-dimensional space corresponding to the defects of the test piece through the induction coil, import the excitation coil temperature and the voltage values ​​in different directions in the three-dimensional space into the three-dimensional surface analytical formula in step S3, and obtain the defect information of the defect under test.

9. A metal component detection method based on directional pulsed eddy current according to claim 8, characterized in that: The process of obtaining the defect information of the defect under test in step S4 is implemented by a CNN convolutional network; specifically, as follows: Step S4A: Obtain the voltage differential value and the corresponding calibration temperature and calibration defect information as training set data, import the real-time induced voltage value and the real-time excitation coil temperature, obtain the voltage differential value through the calibration temperature as the input of the CNN convolutional network, the CNN convolutional network outputs the predicted defect information, and trains the CNN convolutional network through the training set data. During the training process, the learning rate of the CNN convolutional network is optimized through the particle swarm algorithm to obtain the optimized CNN convolutional network.

10. A metal component detection method based on directional pulsed eddy current according to claim 9, characterized in that: In step S4A, the CNN convolutional network is optimized by the particle swarm algorithm, specifically: Step S4A1: Set the induced voltage difference value to a randomly generated cluster, formalized as a matrix with D columns and N rows , expressed as: ； in, is the number of data points corresponding to the learning rate, is the feature dimension, and denote the lower and upper bounds of the solution space, respectively. A random number between 0 and 1; Step S4A2: Divide the training set data into two parts, one for global exploration and the other for local detection; update the position of the individual through global exploration or local detection; Global exploration is expressed as: ； in, Indicates In the iteration individual location, Indicates In the iteration The individual position of represents a random number vector based on Gaussian distribution, represents term-by-term multiplication, Represents a number randomly selected from [0,1]; represents the current optimal solution, represents an individual randomly selected from a group of elites in the population; Represents a number randomly selected from [0,1]; Indicates the location of an average individual; represents the training set data used for global exploration, Represents the training set data used for local detection; Indicated in The centroid position of the entire group in iterations; During the local detection process, the iterative process of ordinary individuals is expressed as: ； ； in, represents the individual after local detection iteration, is the maximum number of iterations, Represents a dynamic adjustment function; Step S4A3: Obtain the optimal individual position through a preset fitness function and use it as the learning rate of the CNN convolutional network.

Images