Phased array ultrasonic nondestructive testing device and method fused with machine learning

Abstract

The invention discloses a phased array ultrasonic nondestructive testing device and method fused with machine learning, and the device comprises at least one ultrasonic phased array probe which is used for collecting flaw detection data and executing a control instruction; and the control end is used for receiving and analyzing the flaw detection data and sending a control instruction to the ultrasonic phased array probe. The detection method comprises the steps of ultrasonic phased array probe initialization, position information initialization, detection starting and intelligent echo data processing. According to the invention, the specific structure of the internal damage of the detected piece can be presented by intelligently utilizing the multi-dimensional point cloud picture, the ultrasonic transceiving information of the ultrasonic phased array can be quickly processed, and meanwhile, the analysis of flaw detection behaviors can be simulated, so that the clear and visual internal perspective of the detected piece can be presented.

Description

Technical Field

[0001] The present invention relates to the technical field of nondestructive flaw detection, and in particular to a phased array ultrasonic nondestructive testing device and a testing method integrating machine learning. Background Art

[0002] With the booming development of artificial intelligence, in the field of non-destructive ultrasonic testing, the current presentation methods of ultrasonic flaw detection instruments on the market remain in the form of two-dimensional profile presentation, oscilloscope presentation and intuitive data presentation, defects require subjective judgment by humans and the position marking is complicated. However, with the improvement of computer capabilities and the decrease in the number of qualified non-destructive testing personnel, the industry has an increasing demand for automated non-destructive testing data analysis.

[0003] Nondestructive ultrasonic testing is a process in which ultrasonic waves are reflected when they propagate through defects or impurities in metal castings or welds. The echo data is analyzed and the location of the defects or impurities is calculated using the speed of sound to determine whether the casting meets the standards.

[0004] Machine learning is used for big data analysis and prediction. Convolutional Neural Network (CNN) is a deep learning model designed for processing data with grid structures, and is particularly good at processing image data. CNN automatically learns local features in images, such as edges, textures, etc., through convolutional layers, and gradually builds higher-level abstract representations. Another type of recurrent neural network (RNN), namely recurrent neural network (RNN), is a deep learning model for processing sequence data. It is particularly suitable for time series analysis, natural language processing, speech recognition and other fields, which usually involve processing time-dependent data.

[0005] The current mainstream piezoelectric material, lead zirconate titanate (PTZ) material, is already relatively mature. Piezoelectric materials can generate electric charge when subjected to mechanical stress, or deform when an electric field is applied. This effect is called the piezoelectric effect, which includes the direct piezoelectric effect and the inverse piezoelectric effect, thereby achieving the effect of ultrasonic emission and reception.

[0006] The nondestructive ultrasonic flaw detection device in the prior art is not only complex in structure and expensive, but also cannot perform three-dimensional reconstruction of the detected data, resulting in low detection efficiency and failure to meet actual flaw detection requirements. Summary of the invention

[0007] The purpose of the present invention is to provide a technical solution of a phased array ultrasonic nondestructive testing device and a testing method integrating machine learning to address the deficiencies in the prior art. The testing device has a simple structure and can not only intelligently use a multi-dimensional point cloud image to present the specific structure of the internal damage of the test piece, but also can quickly process the ultrasonic transceiver information of the ultrasonic phased array, and at the same time can simulate the analysis of the flaw detection behavior, thereby presenting a clear and intuitive internal perspective view of the test piece. The testing method has simple steps, and controls the ultrasonic phased array transmission wave and analyzes the ultrasonic phased array echo data and performs three-dimensional reconstruction by combining the phased array with machine learning, thereby realizing intelligent nondestructive ultrasonic flaw detection and improving the detection efficiency.

[0008] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0009] A phased array ultrasonic nondestructive testing device integrating machine learning, characterized in that:

[0010] At least one ultrasonic phased array probe, used to collect flaw detection data and execute control instructions;

[0011] The control end is used to receive and analyze the flaw detection data and send control instructions to the ultrasonic phased array probe.

[0012] The detection device has a simple structure. It can not only intelligently use multi-dimensional point cloud images to present the specific structure of the internal damage of the test piece, but also quickly process the ultrasonic transmission and reception information of the ultrasonic phased array. At the same time, it can simulate the analysis of the flaw detection behavior, thereby presenting a clear and intuitive internal perspective view of the test piece.

[0013] Furthermore, the ultrasonic phased array probe includes a control unit, an image displacement sensing mechanism, a first PCB circuit board and a phased array assembly. The control unit is respectively connected to the image displacement sensing mechanism and the phased array assembly through the first PCB circuit board, and the displacement and rotation angle of the ultrasonic phased array probe during sliding detection on the test piece are detected through the image displacement sensing mechanism and the phased array assembly, and the probe position offset and angle offset caused by manual operation when the three-dimensional model is constructed are eliminated.

[0014] Furthermore, the control unit includes an MCU, a communication module, a DRAM, a photoelectric sensor and a laser emitter module, a data iteration image processor, a digital control timer, several regional data processors and several control components. The MCU communicates with the control end through the communication module, the DRAM, the regional data processor and the digital control timer are all electrically connected to the MCU, the MCU controls several regional data processors, one regional data processor controls several control components, and the photoelectric sensor and laser emitter module are electrically connected to the MCU through the data iteration image processor.

[0015] Furthermore, each control component includes a digital control counter high-frequency crystal oscillator module, a high-frequency and high-voltage electric pulse generating circuit, a plurality of array elements, a limiting circuit, a gain amplifier circuit, a high-speed signal acquisition chip and a high-speed AD conversion chip, a high-speed access micro DRAM and an array element signal transmission master control circuit. The high-speed signal acquisition chip and the high-speed AD conversion chip and the digital control counter high-frequency crystal oscillator module are all electrically connected to the array element signal transmission master control circuit. The digital control counter high-frequency crystal oscillator module is respectively connected to each array element through the high-frequency and high-voltage electric pulse generating circuit. The array element is connected to the gain amplifier circuit through the limiting circuit. The gain amplifier circuit is connected to the digital control counter through the high-speed signal acquisition chip and the high-speed AD conversion chip. The high-speed signal acquisition chip and the high-speed AD conversion chip are connected to the MCU through the high-speed access micro DRAM.

[0016] Furthermore, the image displacement sensing mechanism includes a fixed ring and a protective cover symmetrically distributed along the fixed ring, an image displacement sensor and an image displacement circuit board are arranged in the protective cover, and the image displacement sensor is connected to the first PCB circuit board through the image displacement circuit board.

[0017] Furthermore, the phased array assembly includes a phased array plate body, a second PCB circuit board and a plurality of array elements. The phased array plate body is connected to the first PCB circuit board through the second PCB circuit board. The phased array plate body is provided with a plurality of array element holes, which are blind holes, and the array elements are arranged in the array element holes.

[0018] Furthermore, the array element includes a phased array probe and a piezoelectric crystal. The piezoelectric crystal is arranged at one end of the phased array probe. The phased array plate is made of high damping material to prevent mutual interference between two adjacent phased array probes.

[0019] Furthermore, the radius of the phased array probe is 0.5-0.8 mm, and the spacing between the center lines of two adjacent phased array probes is 1-1.3 mm.

[0020] Furthermore, the ultrasonic phased array probe also includes a shell, the control unit and the first PCB circuit board are both arranged in the shell, and reserved holes and through holes are respectively provided at both ends of the shell, and the image displacement sensing mechanism and the phased array component are both located on a side close to the through hole.

[0021] As mentioned above, a detection method of a phased array ultrasonic nondestructive testing device integrating machine learning is characterized by comprising the following steps:

[0022] S1. Initialization of ultrasonic phased array probe

[0023] 1) Initialize all modules through MCU and perform status detection on each module;

[0024] 2) MCU obtains the parameters of the device under test from the control end through the communication module, and obtains the parameters of the material parameter matrix P material= [v(ρ, T), a, b, h] and the equation Get the acquisition time t, and use the MCU to control the high-precision digital control timer to set the upper limit of the signal acquisition time, where v is the propagation speed of ultrasound, ρ is the density of the material, T is the temperature of the material, and a, b, and h are the length, width, height, or thickness of the measured object, respectively;

[0025] 3) MCU turns on the image displacement sensor and sends an initialization completion signal to the control end. Initialization is completed and waits for the next instruction;

[0026] S2. Position information initialization

[0027] The MCU turns on the image displacement sensor, prompts the stationary ultrasonic phased array probe through the control terminal, and determines the current position and angle of the ultrasonic phased array probe as the starting point and the starting azimuth, and locates the origin of the current coordinate system in the data iteration image processor;

[0028] S3. Start detection

[0029] 1) The control end sends a signal to the ultrasonic phased array probe through the communication module and receives the detection signal matrix: Where a is the detection mode, is the azimuth angle, θ is the acoustic wave depression angle, r is the focus position radius, and the detection signal matrix S_detection is stored in the MCU cache;

[0030] 2) MCU converts the material parameter matrix P material The detection signal matrix S_detection is transmitted to all regional data processors, and the regional data processors calculate the delay time of the array element transmission in parallel;

[0031] 3) Substitute the delay time result matrix T into the equation: i =x*T to obtain the matrix A=[a1, a2, a3...ai...an], which is the setting value of the high-frequency crystal oscillator module of the digital control counter;

[0032] 4) The echo signal is transmitted through the limiting circuit and the gain amplifier circuit. The high-speed signal acquisition chip and high-speed AD conversion chip of the array element i start to collect the echo signal and store the data in the i-th high-speed access micro DRAM corresponding to the array element in real time. The echo signal is stored as a sequential signal matrix S_echo i =[1, 2, 3, 3, 2, 6...], when the access is completed, the digital control timer module sends a stop acquisition signal to the high-speed signal acquisition chip and the high-speed AD conversion chip;

[0033] 5) MCU sends a signal that all high-speed access micro DRAM data has been stored. MCU traverses all high-speed access micro DRAMs and stores the data in DRAM. MCU reads the data stored in each array element and combines it with the position information of the image displacement sensor to form a large data matrix list: S_eclo_matrix = [S_echo l ,S_echo2,S_echo3,...,S_echo i , ..., S_echo n , ...];

[0034] 6) When the ultrasonic echo is received, the data is stored in the S_eclo_matrix matrix, and the MCU calls the position information in the DRAM to form the final matrix of the echo data: S_all = [D_all, S_echo_matrix], and this matrix is ​​sent back to the control end as the result data of this time;

[0035] 7) When an area is explored, the control end commands to stop the exploration. When it stops, it makes a judgment between two rounds of ultrasonic exploration to achieve stop reset;

[0036] S4, intelligent echo data processing

[0037] 1) The control end uses a machine learning model to analyze the result data, introduces the ultrasonic energy level attenuation curve for secondary simulation judgment, and then uses the dynamic simulation process of the sound pressure attenuation equation to judge the echo waveform;

[0038] 2) The simulation operation is completed by the ultrasonic phased array probe, and the obtained sound pressure result is compared with the waveform collected by the ultrasonic phased array probe. If the comparison is consistent with the expectation, point data will be directly generated in the multidimensional data set. If it is not consistent with the expectation, point data will be generated in the area of ​​the data set by the time t method;

[0039] 3) Eliminate problematic waveforms, substitute the collected echo data into the machine learning model, and analyze many waveforms with different collection positions together to simulate the waveform changes before and after the detection position when moving the ultrasonic phased array probe during the detection process;

[0040] 4) The control end models the test piece with low adjustable transparency, and simulates the ultrasonic wave to demonstrate its detection range and direction in real time with high adjustable transparency, thus realizing three-dimensional visual non-destructive ultrasonic testing.

[0041] This detection method has simple steps. By combining the phased array with machine learning, the ultrasonic phased array transmits waves, analyzes the ultrasonic phased array echo data, and performs three-dimensional reconstruction, it realizes intelligent non-destructive ultrasonic flaw detection and improves detection efficiency.

[0042] The present invention has the following beneficial effects due to the adoption of the above technical solution:

[0043] 1. The detection device of the present invention has a simple structure. It can not only intelligently use a multi-dimensional point cloud image to present the specific structure of the internal damage of the test piece, but also can quickly process the ultrasonic transmission and reception information of the ultrasonic phased array, and at the same time can simulate the analysis of the flaw detection behavior, thereby presenting a clear and intuitive internal perspective view of the test piece.

[0044] 2. The detection method of the present invention has simple steps. It combines the phased array with machine learning to control the ultrasonic phased array transmission wave and analyze the ultrasonic phased array echo data and perform three-dimensional reconstruction, thereby realizing intelligent non-destructive ultrasonic flaw detection and improving detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The present invention will be further described below in conjunction with the accompanying drawings:

[0046] Figure 1 It is a rendering of a phased array ultrasonic nondestructive testing device in a phased array ultrasonic nondestructive testing device and a testing method integrated with machine learning in the present invention;

[0047] Figure 2 This is a rendering of the ultrasonic phased array probe in the present invention;

[0048] Figure 3 for Figure 2 The effect diagram of the middle A direction;

[0049] Figure 4 It is a schematic diagram of the bottom structure of the housing in the present invention;

[0050] Figure 5 It is a schematic diagram of the connection between the control unit, the first PCB circuit board and the image displacement sensing mechanism in the present invention;

[0051] Figure 6 It is a schematic diagram of the connection between the phased array plate body, the first PCB circuit board and the image displacement circuit board in the present invention;

[0052] Figure 7 It is a structural schematic diagram of the image displacement sensing mechanism in the present invention;

[0053] Figure 8 is a circuit block diagram of the control unit in the present invention;

[0054] Fig. 9 is a flow chart of the detection method of the present invention;

[0055] Fig.10 is a curve diagram of the acoustic wave energy attenuation and the reverse transmission energy diversion in the present invention;

[0056] Fig.11It is a schematic diagram of the structure of the phased array assembly in the present invention.

[0057] In the figure: 1-ultrasonic phased array probe; 2-control end; 3-housing; 4-reserved hole; 5-image displacement sensing mechanism; 6-phased array probe; 7-through hole; 8-first PCB circuit board; 9-control unit; 10-phased array board; 11-array element hole; 12-image displacement circuit board; 13-fixing ring; 14-protective cover; 15-image displacement sensor; 16-second PCB circuit board; 17-piezoelectric crystal. DETAILED DESCRIPTION

[0058] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0059] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only embodiments of a part of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0060] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions.

[0061] like Figures 1 to 8 , Fig.11 As shown, a phased array ultrasonic nondestructive testing device integrating machine learning is provided in the present invention, comprising at least one ultrasonic phased array probe 1 and a control terminal 2, wherein the ultrasonic phased array probe 1 is used to collect flaw detection data and execute control instructions.

[0062] This application takes an ultrasonic phased array probe 1 as an example.

[0063] The ultrasonic phased array probe 1 includes a control unit 9, an image displacement sensing mechanism 5, a first PCB circuit board 8 and a phased array component. The control unit 9 is respectively connected to the image displacement sensing mechanism 5 and the phased array component through the first PCB circuit board 8. The displacement and rotation angle of the ultrasonic phased array probe 1 when sliding on the test piece are detected through the image displacement sensing mechanism 5 and the phased array component, and the probe position offset and angle offset caused by manual operation when the three-dimensional model is constructed are eliminated.

[0064] The control unit 9 includes an MCU, a communication module, a DRAM, a photoelectric sensor and a laser emitter module, a data iteration image processor, a digital control timer, several regional data processors and several control components. The MCU communicates with the control end 2 through the communication module. The DRAM, the regional data processor and the digital control timer are all electrically connected to the MCU. The MCU controls several regional data processors, and one regional data processor controls several control components. The photoelectric sensor and the laser emitter module are electrically connected to the MCU through the data iteration image processor.

[0065] Each control component includes a digital control counter high-frequency crystal oscillator module, a high-frequency and high-voltage electric pulse generating circuit, a number of array elements, a limiting circuit, a gain amplifier circuit, a high-speed signal acquisition chip and a high-speed AD conversion chip, a high-speed access micro DRAM and an array element signal transmission master control circuit. The high-speed signal acquisition chip and the high-speed AD conversion chip and the digital control counter high-frequency crystal oscillator module are all electrically connected to the array element signal transmission master control circuit. The digital control counter high-frequency crystal oscillator module is respectively connected to each array element through the high-frequency and high-voltage electric pulse generating circuit. The array element is connected to the gain amplifier circuit through the limiting circuit. The gain amplifier circuit is connected to the digital control counter through the high-speed signal acquisition chip and the high-speed AD conversion chip. The high-speed signal acquisition chip and the high-speed AD conversion chip are connected to the MCU through the high-speed access micro DRAM.

[0066] The image displacement sensing mechanism 5 includes a fixing ring 13 and a protective cover 14 symmetrically distributed along the fixing ring 13. An image displacement sensor 15 and an image displacement circuit board 12 are arranged in the protective cover 14. The image displacement sensor 15 is connected to the first PCB circuit board 8 through the image displacement circuit board 12. The image displacement sensor 15 is used for analyzing and receiving the movement data of the ultrasonic phased array probe 1 during the detection process.

[0067] The phased array assembly includes a phased array plate body 10, a second PCB circuit board 16 and a plurality of array elements. The phased array plate body 10 is connected to the first PCB circuit board 8 through the second PCB circuit board 16. The phased array plate body 10 is provided with a plurality of array element holes 11. The array element holes 11 are blind holes, and the array elements are arranged in the array element holes 11.

[0068] In order to ensure that the ultrasonic wave propagates in the phased array probe 6 without causing clutter after being excited, and to realize an ideal spherical wave according to the Huygens principle, the radius of the phased array probe 6 needs to be smaller than the wavelength of the sound wave.

[0069] To calculate the wavelength range of ultrasound, use the wavelength formula:

[0070] λ=vf

[0071] Among them, λ is the wavelength, v is the wave speed, and f is the frequency. Generally, the ultrasonic frequency range used in non-destructive ultrasound is 0.5MHz to 10MHz, and the wave speed range of sound waves in metal castings is about 2000m / s to 6000m / s.

[0072] First, the wavelength at the lowest frequency of 0.5MHz and the highest wave speed of 6000m / s:

[0073] λmax=6000m / s*0.5×10 6 Hz=12mm

[0074] Then, the wavelength at the highest frequency of 10MHz and the lowest wave speed of 2000m / s:

[0075] λmin=2000m / s*10×10 6 Hz=0.2mm

[0076] Therefore, the wavelength range of ultrasonic waves is 0.2mm~12mm.

[0077] In practical applications, since the propagation speed of ultrasound in different metal casting materials is different, the frequency to be used in detection is determined by using f=λ / v after determining the wave speed in the test piece and the probe diameter.

[0078] The radius of the phased array probe 6 is 0.5-0.8 mm, and the distance between the center lines of two adjacent phased array probes 6 is 1-1.3 mm.

[0079] The array element includes a phased array probe 6 and a piezoelectric crystal 17, the piezoelectric crystal 17 is arranged at one end of the phased array probe 6, and the phased array plate 10 is made of high damping material to prevent mutual interference between two adjacent phased array probes 6. High damping materials can convert mechanical energy into heat energy and reduce the reflection, transmission and scattering of ultrasonic waves. For example, common ultrasonic wave absorbing materials include polymer-based materials (such as rubber, polyurethane foam), porous materials (such as open-cell foam, honeycomb structure materials), composite materials (such as carbon fiber, graphene, metal powder added to the matrix material), micro / nanostructure materials and smart materials.

[0080] The piezoelectric crystal 17 is in the shape of a pancake, and the polarization direction of the piezoelectric crystal 17 is parallel to the busbar direction of the phased array probe 6, and the direction of the applied electric field is parallel to the busbar direction of the phased array probe 6. The material of the piezoelectric crystal 17 is generally a piezoelectric material such as lead zirconate titanate (PZT) with high piezoelectric coefficient and good mechanical properties.

[0081] The piezoelectric materials are closely arranged in a matrix to make the ultrasonic emitting surface of the ultrasonic phased array probe 1 as a whole close to a circle, which is different from the uneven wavefront sound pressure caused by the diagonal delay during lateral transmission of the rectangular phased array, which may cause blind spots when receiving echoes.

[0082] When using ultrasonic waves with a frequency of 0.5 to 0 MHz, when the ultrasonic waves are transmitted to the object under test through the matching layer, since the diameter of the matching layer is smaller than the wavelength of the ultrasonic waves, according to the Huygens principle, the ultrasonic waves form spherical waves after passing through the matching layer and being transmitted to the object under test, thereby achieving the emission of the directional interference wavefront of the phased array. The matching layer is located in the rest of the part except the piezoelectric crystal 17, and requires a low damping coefficient material.

[0083] The ultrasonic phased array probe 1 further includes a housing 3, a control unit 9 and a first PCB circuit board 8 are both disposed in the housing 3, a reserved hole 4 and a through hole 7 are respectively disposed at both ends of the housing 3, and the image displacement sensing mechanism 5 and the phased array assembly are both located on a side close to the through hole 7. The housing 3 is made of a hard material such as plastic or metal.

[0084] The control terminal 2 is used to receive and analyze the flaw detection data and send control instructions to the ultrasonic phased array probe 1.

[0085] The detection device has a simple structure. It can not only intelligently use multi-dimensional point cloud images to present the specific structure of the internal damage of the test piece, but also quickly process the ultrasonic transmission and reception information of the ultrasonic phased array. At the same time, it can simulate the analysis of the flaw detection behavior, thereby presenting a clear and intuitive internal perspective view of the test piece.

[0086] like Fig. 9 As shown, a detection method of a phased array ultrasonic nondestructive testing device integrating machine learning is provided in the present invention, comprising the following steps:

[0087] S1. Initialization of ultrasonic phased array probe 1

[0088] 1) Initialize all modules through MCU and perform status detection on each module;

[0089] 2) MCU obtains the parameters of the device under test from the control terminal 2 through the communication module, and obtains the parameters of the device under test through the material parameter matrix P material = [v(ρ, T), a, b, h] and the equation Get the acquisition time t, and use the MCU to control the high-precision digital control timer to set the upper limit of the signal acquisition time, where v is the propagation speed of ultrasound, ρ is the density of the material, T is the temperature of the material, and a, b, and h are the length, width, height, or thickness of the measured object, respectively;

[0090] 3) MCU turns on the image displacement sensor 15, starts sending an initialization completion signal to the control terminal 2, and the initialization is completed, waiting for the next instruction;

[0091] The image displacement sensor 15 emits laser light and captures reflected light. It contains a laser transmitter inside, which emits a beam of infrared laser light. This beam of infrared laser light is focused into a very small laser spot through a special lens (lens). When the laser spot is irradiated onto a plane or other surface, reflection occurs, and the reflected light enters the photoelectric sensor through the lens. The photoelectric sensor is a photosensitive element. Its main function is to receive the reflected light signal and convert it into an electrical signal. The photoelectric sensor contains a high-resolution image sensor and a digital signal processor. The image sensor is responsible for capturing the light reflected from the surface and converting it into a digital signal. The digital signal processor processes and analyzes the image captured by the image sensor. The digital signal processor solves the image displacement through the existing image displacement processing algorithm to determine its moving distance relative to the surface and the direction matrix D_original = [Δx1, Δy1, Δx2, Δy2]. By iterating this process, the position change and rotation angle change of the probe can be obtained. By substituting into the equation:

[0092]

[0093] The displacement Δx, Δy of the probe center and the rotation angle φ of the probe can be obtained, where r represents the midpoint of the electrical connection line detected by the two image sensors respectively.

[0094] The data is passed to the data iteration image processor for iteration, so that the data iteration image processor stores and updates the current probe position and defines the parameters:

[0095] D_location = [x, y, Φ]

[0096] It represents the three-dimensional coordinate system information with the starting point and starting azimuth defined during position initialization as the origin.

[0097] Finally, the data is integrated into:

[0098] D all =[D_probe, D_location]=[Δx, Δy, ΔΦ, x, y, Φ]

[0099] It transmits the position information of the ultrasonic phased array probe 1 to the control terminal 2 in real time through the MCU controlling the communication module, and the signal is always connected to the control terminal 2.

[0100] S2. Position information initialization

[0101] The MCU turns on the image displacement sensor 15, prompts the stationary ultrasonic phased array probe 1 through the control terminal 2, and determines the current position and angle of the ultrasonic phased array probe 1 as the starting point and the starting azimuth, and locates the origin of the current coordinate system in the data iteration image processor;

[0102] S3. Start detection

[0103] 1) The control terminal 2 sends a signal to the ultrasonic phased array probe 1 through the communication module and receives the detection signal matrix: Where a is the detection mode, which is a binary value. When it is 0, it is the plane wave mode, and when it is 1, it is the focused wave mode. is the azimuth angle, θ is the acoustic wave depression angle, r is the focus position radius, and the detection signal matrix S_detection is stored in the MCU cache;

[0104] 2) MCU converts the material parameter matrix P material The detection signal matrix S_detection is transmitted to all regional data processors, and the regional data processors calculate the delay time of the array element transmission in parallel;

[0105] When calculating the ultrasonic array element transmission delay in the regional data processor, the material parameter matrix P material Substitute the detection signal matrix S_detection into the universal equation for calculating the delay time of each array element:

[0106]

[0107] , where each array element has coordinates (x, y), and the relative delay time matrix T = [t1, t2, t3, ..., ti, ..., tn] required for each array element corresponding to the coordinates is obtained.

[0108] In order to ensure the running speed, the probe does not use iterative methods such as min and max to find the maximum function algorithm, but instead adopts a memory call strategy to make a reverse extension line of the emission azimuth angle and define the relationship between the angle and the initial detection array element, which means that between two certain angles, an array element on the edge must be emitted first.

[0109] The data of the azimuth angle φ and the corresponding array element coordinates are stored in the MCU. After the MCU assigns the task to the regional data processor, the azimuth angle is substituted into it for calculation. After the calculation is completed, the coordinates of the first transmitting array element are obtained, and the time required for delay is read to obtain the delay time [ti] of the first transmitting array element. [ti] should be the minimum value and a negative value, calculated as: [tmin], and the delay time of all array elements is subtracted from the minimum value:

[0110] T-[tmin]=[tl-tmin, t2-tmin, t3-tmin,..., ti-tmin,..., tn-tmin]

[0111] In order to achieve the effect of array element delay control transmission.

[0112] 3) Assume that the frequency of the "crystal oscillator" in the circuit is x GHz (1 GHz = 109 Hz), and substitute the delay time result matrix T into the equation: i =x*T to obtain the matrix A=[a1, a2, a3...ai...an], which is the setting value of the high-frequency crystal oscillator module of the digital control counter;

[0113] Each value of the matrix is ​​assigned to the digital control counter high-frequency crystal oscillator module of the corresponding array element by using the regional data processor. This value represents the crystal oscillator needs to vibrate a i In this case, the timing is completed with 1 nanosecond as the basic unit. All counters share the same crystal oscillator, and a serpentine line needs to be set to ensure the consistency of time.

[0114] Here we start with an array element as an example. Let the array element be code-named i. The crystal oscillator count value calculated by the regional data processor corresponding to the array element is a. i , the counter value is set to a through the digital control counter. After the setting is completed, the level of the command line leading to the inter-element signal transmission master control circuit to execute the "complete count setting" command line is pulled high. The inter-element signal transmission master control circuit uses the AND gate to integrate the processing signals of all array elements. When the final end is raised to a high level, that is, the "start timing" command line is at a high level, the "start timing" command line uses a serpentine line to prevent delay, so that the high-frequency crystal oscillator modules of all array elements start timing at the same time; at the same time, the "start acquisition" command line is at a high level, and the "start acquisition" command line uses a serpentine line to prevent delay, so that the high-speed signal acquisition chips and high-speed AD conversion chips of all array elements start to collect at the same time; at the same time, the digital control timer is turned on to start collecting echoes. When the high-frequency crystal oscillator module of the digital control counter reaches the count value, the level leading to the high-frequency high-voltage electric pulse generating circuit is pulled high to start exciting the high-frequency high-voltage electric pulse. The high-voltage DC module continues to supply energy, and after its excitation, ultrasonic pulses are emitted through the piezoelectric material.

[0115] 4) The echo signal is transmitted through the limiting circuit and the gain amplifier circuit. The high-speed signal acquisition chip and high-speed AD conversion chip of the array element i start to collect the echo signal and store the data in the i-th high-speed access micro DRAM corresponding to the array element in real time. The echo signal is stored as a sequential signal matrix S_echo i =[1,2,3,3,2,6…], when the access is completed, the digital control timer module sends a stop acquisition signal to the high-speed signal acquisition chip and the high-speed AD conversion chip;

[0116] 5) MCU sends a signal that all high-speed access micro DRAM data have been stored. MCU traverses all high-speed access micro DRAMs and stores the data in DRAM. MCU reads the data stored in each array element and combines it with the position information of the image displacement sensor 15 to form a large data matrix list: S_echo_matrix = [S_echo1, S_echo2, S_echo3, ..., S_echo i , ..., S_echo n , ...];

[0117] 6) When the ultrasonic echo is received, the data is stored in the S_echo_matrix matrix, and the MCU calls the position information in the DRAM to form the final matrix of the echo data: S_all = [D_all, S_echo_matrix], and this matrix is ​​sent back to the control terminal 2 as the result data;

[0118] 7) When the detection of an area is completed, the control terminal 2 commands to stop the detection. When stopping, a judgment is made between two rounds of ultrasonic detection to achieve stop reset;

[0119] S4, Intelligent echo data processing

[0120] 1) The control end 2 uses a machine learning model to analyze the result data, introduces the ultrasonic energy level attenuation curve for secondary simulation judgment, and then uses the dynamic simulation process of the sound pressure attenuation equation to judge the echo waveform;

[0121] This simulation is used to identify invalid data by removing unlikely clutter or high-amplitude peaks caused by improper operation.

[0122] The sound pressure attenuation equation will judge the possible cause of each possible waveform: the energy level of the sound wave will attenuate when propagating in the material, but the sound speed remains constant in the same material. The time from the starting wave to the first possible peak is t. First, the speed parameter is substituted into the attenuation energy level equation specific to the material to calculate the sound pressure at the moment before the reflection at t / 2. A reflected wave energy level shunt is applied at t / 2 to obtain the sound pressure at the moment after the reflection, which is then substituted into the attenuation energy level equation. At this time, the propagation distance parameter of the attenuation energy level equation continues to increase, but the starting sound pressure is calculated using the sound pressure at the moment after the reflection.

[0123] The attenuation equation of the sound pressure and distance of ultrasonic waves propagating in a hemispherical shape can be expressed as:

[0124]

[0125] Where: P(x) is the sound pressure at a distance x from the sound source; P0 is the initial sound pressure at the sound source; α is the attenuation coefficient of the medium, in units of NP / m (Nepers per meter); x is the distance from the sound source; e is the base of the natural logarithm (e≈2.718).

[0126] The medium attenuation coefficient α can be further decomposed into the absorption attenuation coefficient α a and the scattering attenuation coefficient α s :

[0127] α=α a +α s

[0128] For metal media, the absorption attenuation coefficient α a It can be expressed as:

[0129] α a =c l *f

[0130] Where: α a is the absorption attenuation coefficient; f is the frequency of ultrasound; c1 is a constant, usually called the "absorption coefficient" or "frequency dependence constant of the absorption attenuation coefficient", which can generally be measured by experiments and is related to the nature of the metal material.

[0131] Scattering attenuation coefficient α s :

[0132]

[0133] Where d is the grain diameter of the medium; λ is the wavelength; F is the anisotropy coefficient; c2, c3, c4 are constants of the metal material.

[0134] When ultrasound propagates from one medium to another, it will be reflected at the interface. The magnitude of the reflected energy depends on the difference in acoustic impedance between the two media. Acoustic impedance Z is defined as the product of the density ρ of the medium and the speed of sound c, that is, Z = ρc. The sound pressure reflection coefficient R can be expressed as:

[0135]

[0136] Where Z1 and Z2 are the acoustic impedances of the two media respectively. r With the incident energy E i The relationship is:

[0137] E r =R 2 *E i

[0138] The sound wave energy attenuation and back-projection energy diversion curves are as follows: Fig.10As shown, the curve is obtained by computer calculation combining various parameters, and compared and analyzed with the real echo data to realize intelligent echo clutter processing.

[0139] 2) The ultrasonic phased array probe 1 is used to complete the simulation operation, and the obtained sound pressure result is compared with the waveform collected by the ultrasonic phased array probe 1. If the comparison is consistent with the expectation, point data will be directly generated in the multidimensional data set. If it is not consistent with the expectation, point data will be generated in the area of ​​the data set by the time t method;

[0140] 3) Eliminate problematic waveforms, substitute the collected echo data into the machine learning model, and analyze a number of waveforms with different collection positions together to simulate the waveform changes before and after the detection position by moving the ultrasonic phased array probe 1 during the detection process;

[0141] 4) The control end 2 models the test piece with low adjustable transparency, and simulates the ultrasonic wave to demonstrate its detection range and direction in real time with high adjustable transparency, thereby realizing three-dimensional visual non-destructive ultrasonic testing.

[0142] The detection method is analyzed from three intelligent data analysis perspectives:

[0143] 1. Intelligent analysis of data between array elements

[0144] The defect size is generally smaller than the diameter of the cylindrical wave emitted by the phased array. At this time, a spherical echo will be formed, and the phased array can receive the spherical echo without distortion. When performing a single signal analysis, the radius of the spherical echo can be obtained by comparing the time difference of the echoes received between different array elements. In this way, the specific location and approximate size of the defect can be anchored in combination with the propagation time of the wave, thereby improving data accuracy.

[0145] 2. Intelligent analysis of energy propagation attenuation data

[0146] The position and echo data matrix S_echo_matrix and the detection signal matrix S_detection are combined with the acoustic energy attenuation and anti-transmission energy diversion curves exclusive to this detection to achieve the effect of intelligent clutter filtering. The acoustic pressure of the echo when the x medium reflection surface appears at different positions is iteratively deduced, combined with the interference energy level superposition, and substituted into the echo time for inspection. Each echo is analyzed and the defect model is constructed in the three-dimensional diagram to improve data accuracy.

[0147] 3. Intelligent analysis of data from different observation points at the same location

[0148] When the echo data reaches a certain scale, since the phased array interference wave can be turned, the same area can be explored from different positions and the results will not be covered. And because the exploration frequency is high, the amount of data is large, which is convenient for data coordination, eliminating special conditions, and preventing the impact on accuracy. Improve data accuracy.

[0149] This detection method has simple steps. By combining the phased array with machine learning, the ultrasonic phased array transmits waves, analyzes the ultrasonic phased array echo data, and performs three-dimensional reconstruction, it realizes intelligent non-destructive ultrasonic flaw detection and improves detection efficiency.

[0150] The above are only specific embodiments of the present invention, but the technical features of the present invention are not limited thereto. Any simple changes, equivalent replacements or modifications made based on the present invention to achieve basically the same technical effects are all included in the protection scope of the present invention.

Claims

1. A phased array ultrasonic nondestructive testing device integrating machine learning, characterized in that: include At least one ultrasonic phased array probe, used to collect flaw detection data and execute control instructions; and a control end, for receiving and analyzing the flaw detection data and sending control instructions to the ultrasonic phased array probe.

2. The phased array ultrasonic nondestructive testing device integrating machine learning according to claim 1 is characterized in that: The ultrasonic phased array probe includes a control unit, an image displacement sensing mechanism, a first PCB circuit board and a phased array component. The control unit is respectively connected to the image displacement sensing mechanism and the phased array component through the first PCB circuit board. The displacement and rotation angle of the ultrasonic phased array probe during sliding detection on the test piece are detected through the image displacement sensing mechanism and the phased array component, and the probe position offset and angle offset caused by manual operation when the three-dimensional model is constructed are eliminated.

3. The phased array ultrasonic nondestructive testing device integrating machine learning according to claim 2 is characterized in that: The control unit includes an MCU, a communication module, a DRAM, a photoelectric sensor and a laser emitter module, a data iteration image processor, a digital control timer, several regional data processors and several control components. The MCU communicates with the control end through the communication module. The DRAM, the regional data processor and the digital control timer are all electrically connected to the MCU. The MCU controls several regional data processors, and one regional data processor controls several control components. The photoelectric sensor and laser emitter module are electrically connected to the MCU through the data iteration image processor.

4. The phased array ultrasonic nondestructive testing device integrating machine learning according to claim 3 is characterized in that: Each of the control components includes a digital control counter high-frequency crystal oscillator module, a high-frequency and high-voltage electric pulse generating circuit, a plurality of array elements, a limiting circuit, a gain amplifier circuit, a high-speed signal acquisition chip and a high-speed AD conversion chip, a high-speed access micro DRAM and an inter-array element signal transmission master control circuit. The high-speed signal acquisition chip and the high-speed AD conversion chip and the digital control counter high-frequency crystal oscillator module are all electrically connected to the inter-array element signal transmission master control circuit. The digital control counter high-frequency crystal oscillator module is respectively connected to each of the array elements through the high-frequency and high-voltage electric pulse generating circuit. The array element is connected to the gain amplifier circuit through the limiting circuit. The gain amplifier circuit is connected to the digital control counter through the high-speed signal acquisition chip and the high-speed AD conversion chip. The high-speed signal acquisition chip and the high-speed AD conversion chip are connected to the MCU through the high-speed access micro DRAM.

5. The phased array ultrasonic nondestructive testing device integrating machine learning according to claim 4 is characterized in that: The image displacement sensing mechanism includes a fixing ring and a protective cover symmetrically distributed along the fixing ring. An image displacement sensor and an image displacement circuit board are arranged in the protective cover. The image displacement sensor is connected to the first PCB circuit board through the image displacement circuit board.

6. The phased array ultrasonic nondestructive testing device integrating machine learning according to claim 5 is characterized in that: The phased array assembly includes a phased array plate, a second PCB circuit board and a plurality of array elements. The phased array plate is connected to the first PCB circuit board through the second PCB circuit board. The phased array plate is provided with a plurality of array element holes, which are blind holes, and the array elements are arranged in the array element holes.

7. The phased array ultrasonic nondestructive testing device integrating machine learning according to claim 6 is characterized in that: The array element comprises a phased array probe and a piezoelectric crystal. The piezoelectric crystal is arranged at one end of the phased array probe. The phased array plate is made of high damping material to prevent mutual interference between two adjacent phased array probes.

8. The phased array ultrasonic nondestructive testing device integrating machine learning according to claim 7 is characterized in that: The radius of the phased array probe is 0.5-0.8 mm, and the distance between the center lines of two adjacent phased array probes is 1-1.3 mm.

9. The phased array ultrasonic nondestructive testing device integrating machine learning according to claim 2, characterized in that: The ultrasonic phased array probe also includes a shell, the control unit and the first PCB circuit board are both arranged in the shell, two ends of the shell are respectively provided with a reserved hole and a through hole, and the image displacement sensing mechanism and the phased array component are both located on a side close to the through hole.

10. A detection method for a phased array ultrasonic nondestructive testing device integrating machine learning according to any one of claims 1 to 9, characterized in that The steps include: S1. Initialization of ultrasonic phased array probe 1) Initialize all modules through MCU and perform status detection on each module; 2) MCU obtains the parameters of the device under test from the control end through the communication module, and obtains the parameters of the material parameter matrix P material = [v(ρ, T), a, b, h] and the equation Get the acquisition time t, and use the MCU to control the high-precision digital control timer to set the upper limit of the signal acquisition time, where v is the propagation speed of ultrasound, ρ is the density of the material, T is the temperature of the material, and a, b, and h are the length, width, height, or thickness of the measured object, respectively; 3) MCU turns on the image displacement sensor and sends an initialization completion signal to the control end. Initialization is completed and waits for the next instruction; S2. Position information initialization The MCU turns on the image displacement sensor, prompts the stationary ultrasonic phased array probe through the control terminal, and determines the current position and angle of the ultrasonic phased array probe as the starting point and the starting azimuth, and locates the origin of the current coordinate system in the data iteration image processor; S3. Start detection 1) The control end sends a signal to the ultrasonic phased array probe through the communication module and receives the detection signal matrix: Where a is the detection mode, is the azimuth angle, θ is the acoustic wave depression angle, r is the focus position radius, and the detection signal matrix S_detection is stored in the MCU cache; 2) MCU converts the material parameter matrix P material The detection signal matrix S_detection is transmitted to all regional data processors, and the regional data processors calculate the delay time of the array element transmission in parallel; 3) Substitute the delay time result matrix T into the equation: i =x*T to obtain the matrix A=[a1, a2, a3...ai...an], which is the setting value of the high-frequency crystal oscillator module of the digital control counter; 4) The echo signal is transmitted through the limiting circuit and the gain amplifier circuit. The high-speed signal acquisition chip and high-speed AD conversion chip of the array element i start to collect the echo signal and store the data in the i-th high-speed access micro DRAM corresponding to the array element in real time. The echo signal is stored as a sequential signal matrix S_echo i =[1, 2, 3, 3, 2, 6...], when the access is completed, the digital control timer module sends a stop acquisition signal to the high-speed signal acquisition chip and the high-speed AD conversion chip; 5) MCU sends a signal that all high-speed access micro DRAM data have been stored. MCU traverses all high-speed access micro DRAMs and stores the data in DRAM. MCU reads the data stored in each array element and combines it with the position information of the image displacement sensor to form a large data matrix list: S_echo_matrix = [S_echo1, S_echo2, S_echo3, ..., S_echo i , ..., S_echo n , ...]; 6) When the ultrasonic echo is received, the data is stored in the S_echo_matrix matrix, and the MCU calls the position information in the DRAM to form the final matrix of the echo data: S_all = [D_all, S_echo_matrix], and this matrix is ​​sent back to the control end as the result data of this time; 7) When an area is explored, the control end commands to stop the exploration. When it stops, it makes a judgment between two rounds of ultrasonic exploration to achieve stop reset; S4, intelligent echo data processing 1) The control end uses a machine learning model to analyze the result data, introduces the ultrasonic energy level attenuation curve for secondary simulation judgment, and then uses the dynamic simulation process of the sound pressure attenuation equation to judge the echo waveform; 2) The simulation operation is completed by the ultrasonic phased array probe, and the obtained sound pressure result is compared with the waveform collected by the ultrasonic phased array probe. If the comparison is consistent with the expectation, point data will be directly generated in the multidimensional data set. If it is not consistent with the expectation, point data will be generated in the area of ​​the data set by the time t method; 3) Eliminate problematic waveforms, substitute the collected echo data into the machine learning model, and analyze many waveforms with different collection positions together to simulate the waveform changes before and after the detection position when moving the ultrasonic phased array probe during the detection process; 4) The control end models the test piece with low adjustable transparency, and simulates the ultrasonic wave to demonstrate its detection range and direction in real time with high adjustable transparency, realizing three-dimensional visual non-destructive ultrasonic testing.

Images