A method for quantifying defect sizes and detecting motion conditions of thin plate materials in rapid motion

Abstract

The present invention proposes a method for quantifying the defect size of thin plate materials in rapid motion and detecting the motion conditions. The method includes obtaining two-axis motion-induced eddy current signals of defects in experimental thin plate materials with different motion speeds, lifting distances and defect sizes, wherein the two-axis motion-induced eddy current signals include parallel and vertical motion-induced eddy current signals, so as to obtain multiple characteristic parameters and then construct a sample data set; constructing a quantitative detection model based on a neural network, wherein the parallel characteristic parameter input module in the quantitative detection model is connected to the fusion module via a first quantitative detection module, and the vertical characteristic parameter input module is connected to the fusion module via a second quantitative detection module, and the outputs of the two quantitative detection modules both include motion speed, lifting distance and defect size; training the model to obtain a target quantitative detection model; obtaining the target two-axis motion-induced eddy current signals of the defects of the target thin plate material to be quantified, and using the target quantitative detection model to output the defect size, motion speed and lifting distance of the target thin plate material.

Description

Technical Field

[0001] The present invention relates to the technical field of thin plate material defect detection in rapid motion, and in particular to a method for quantifying defect sizes of thin plate materials in rapid motion and detecting motion conditions. Background Art

[0002] Sheet materials are often processed into a variety of thin structural components and are widely used in many fields. During the rapid production of these structural components, defects such as bubbles, cracks, and various non-metallic inclusions are prone to occur within the sheet materials. Even the smallest defects can seriously affect product quality and performance.

[0003] Currently, defect detection for thin sheet materials mostly relies on magnetic flux leakage testing (MFL). However, this method is significantly affected by interference signals in the environment and is limited by the metallic properties of the sheet material, resulting in a high false alarm rate. Eddy current testing also exists, but the detection signal is easily distorted for moving objects, so detection accuracy needs to be improved. Furthermore, the precise location of detected defects requires consideration of the sheet material's motion, but existing technologies lack such capabilities.

[0004] In summary, accurate detection of thin plate material defects in a high-speed production process and quantification of defect size and motion conditions are of great significance for product quality control and accurate defect positioning. Summary of the Invention

[0005] The present invention aims to solve one of the technical problems in the related art at least to a certain extent.

[0006] To this end, the first purpose of the present invention is to propose a method for quantifying the defect size of thin plate materials in rapid motion and detecting the motion conditions, so as to accurately quantify the defect size and accurately detect the motion conditions during the rapid motion process of thin plate material production.

[0007] The second object of the present invention is to provide a system for quantifying defect sizes of thin plate materials in rapid motion and detecting motion conditions.

[0008] A third object of the present invention is to provide an electronic device.

[0009] A fourth object of the present invention is to provide a computer-readable storage medium.

[0010] To achieve the above objectives, the first aspect of the present invention provides a method for quantifying defect sizes of thin plate materials during rapid motion and detecting motion conditions, comprising:

[0011] Obtaining two-axis motional eddy current signals of defects in the test sheet material at different motion speeds, different lift-off distances, and different defect sizes, wherein the two-axis motional eddy current signals include a parallel motional eddy current signal and a vertical motional eddy current signal;

[0012] performing feature extraction on the parallel motion-induced eddy current signal and the vertical motion-induced eddy current signal to obtain a plurality of parallel feature parameters and a plurality of vertical feature parameters, respectively, and constructing a sample data set based on the plurality of parallel feature parameters and the plurality of vertical feature parameters, as well as corresponding motion speeds, lift-off distances, and defect sizes;

[0013] A quantitative detection model is constructed based on a neural network, wherein the quantitative detection model includes a parallel feature parameter input module, a first quantitative detection module, a vertical feature parameter input module, a second quantitative detection module, a fusion module, and an output module. The parallel feature parameter input module is connected to the fusion module via the first quantitative detection module, and the vertical feature parameter input module is connected to the fusion module via the second quantitative detection module. The fusion module is also connected to the output module. The outputs of the first quantitative detection module and the second quantitative detection module both include movement speed, lift-off distance, and defect size.

[0014] Using the sample data set to train the first quantization detection module and the second quantization detection module to obtain a target quantization detection model;

[0015] Obtaining target two-axis motional eddy current signals of defects in a target thin plate material to be quantified, and then obtaining corresponding multiple target parallel characteristic parameters and multiple target perpendicular characteristic parameters;

[0016] The plurality of target parallel characteristic parameters and the plurality of target vertical characteristic parameters are input into a target quantitative detection model to output the defect size, movement speed and lifting distance of the target thin plate material.

[0017] In the method of the first aspect of the present invention, the first quantification detection module includes a first defect quantification module and a first motion condition detection module, and the parallel characteristic parameter input module is connected to the fusion module via the first defect quantification module and the first motion condition detection module respectively; the second quantification detection module includes a second defect quantification module and a second motion condition detection module, and the vertical characteristic parameter input module is connected to the fusion module via the second defect quantification module and the second motion condition detection module respectively; the outputs of the first defect quantification module and the second defect quantification module both include defect size; the outputs of the first motion condition detection module and the second motion condition detection module both include motion speed and lifting distance.

[0018] In the method of the first aspect of the present invention, both the first defect quantization module and the second defect quantization module adopt a convolutional network and a bidirectional long short-term memory neural network.

[0019] In the method of the first aspect of the present invention, both the first motion condition detection module and the second motion condition detection module adopt convolutional networks.

[0020] In the method of the first aspect of the present invention, the fusion module adopts a fully connected layer.

[0021] In the method of the first aspect of the present invention, the multiple parallel characteristic parameters include the amplitude, peak-to-peak value, pulse width, signal area, signal energy, pulse deflection obtained by fast Fourier transform, the ratio of signal area to signal peak-to-peak value, and the ratio of signal area to pulse width extracted from the parallel motion-induced eddy current signal; the multiple vertical characteristic parameters include the amplitude, peak-to-peak value, pulse width, signal area, signal energy, pulse deflection obtained by fast Fourier transform, the ratio of signal area to signal peak-to-peak value, and the ratio of signal area to pulse width extracted from the vertical motion-induced eddy current signal.

[0022] In the method of the first aspect of the present invention, two-axis motional eddy current signals of defects in the experimental thin plate material at different movement speeds, different lifting distances, and different defect sizes are obtained through finite element software modeling simulation or actual measurement.

[0023] To achieve the above-mentioned objectives, the second aspect of the present invention provides a system for quantifying defect sizes of thin plate materials in rapid motion and detecting motion conditions, comprising:

[0024] A sample data set construction unit is used to obtain two-axis motion-induced eddy current signals of defects in experimental thin plate materials under different motion speeds, different lift-off distances, and different defect sizes, wherein the two-axis motion-induced eddy current signals include parallel motion-induced eddy current signals and vertical motion-induced eddy current signals; feature extraction is performed on the parallel motion-induced eddy current signals and the vertical motion-induced eddy current signals to obtain a plurality of parallel feature parameters and a plurality of vertical feature parameters, and a sample data set is constructed based on the plurality of parallel feature parameters and the plurality of vertical feature parameters, as well as the corresponding motion speeds, lift-off distances, and defect sizes;

[0025] a modeling unit for constructing a quantitative detection model based on a neural network, the quantitative detection model comprising a parallel feature parameter input module, a first quantitative detection module, a vertical feature parameter input module, a second quantitative detection module, a fusion module, and an output module, wherein the parallel feature parameter input module is connected to the fusion module via the first quantitative detection module, the vertical feature parameter input module is connected to the fusion module via the second quantitative detection module, and the fusion module is further connected to the output module, wherein the outputs of the first quantitative detection module and the second quantitative detection module both include motion speed, lift-off distance, and defect size; and the first and second quantitative detection modules are trained using the sample data set to obtain a target quantitative detection model;

[0026] An acquisition unit is used to acquire target two-axis motional eddy current signals of defects in a target sheet material to be quantified, and then obtain corresponding multiple target parallel characteristic parameters and multiple target perpendicular characteristic parameters;

[0027] The quantitative detection unit is used to input the multiple target parallel characteristic parameters and the multiple target vertical characteristic parameters into the target quantitative detection model to output the defect size, movement speed and lifting distance of the target thin plate material.

[0028] To achieve the above-mentioned purpose, the third aspect of the present invention proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method proposed in the first aspect of the present invention.

[0029] To achieve the above-mentioned purpose, the fourth aspect of the present invention proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the method proposed in the first aspect of the present invention.

[0030] The present invention provides a method, system, electronic device and storage medium for quantifying defect sizes and detecting motion conditions of thin plate materials in rapid motion. The method obtains two-axis motion-induced eddy current signals of defects of experimental thin plate materials under different motion speeds, different lifting distances and different defect sizes. The two-axis motion-induced eddy current signals include parallel motion-induced eddy current signals and vertical motion-induced eddy current signals. Feature extraction is performed on the parallel motion-induced eddy current signals and the vertical motion-induced eddy current signals respectively to obtain a variety of parallel feature parameters and a variety of vertical feature parameters. A sample data set is constructed based on the multiple parallel feature parameters and the multiple vertical feature parameters, as well as the corresponding motion speeds, lifting distances and defect sizes. A quantitative detection model is constructed based on a neural network. The quantitative detection model includes a parallel feature parameter input module, a first quantitative detection module, a vertical feature parameter input module, a second quantitative detection module, and a second quantitative detection module. A detection module, a fusion module and an output module, a parallel feature parameter input module is connected to the fusion module via a first quantitative detection module, a vertical feature parameter input module is connected to the fusion module via a second quantitative detection module, and the fusion module is also connected to the output module, the outputs of the first quantitative detection module and the second quantitative detection module both include movement speed, lifting distance and defect size; the first quantitative detection module and the second quantitative detection module are trained using a sample data set to obtain a target quantitative detection model; the target two-axis motional eddy current signal of the defect of the target thin plate material to be quantified is obtained, and then a corresponding plurality of target parallel feature parameters and a plurality of target vertical feature parameters are obtained; the plurality of target parallel feature parameters and the plurality of target vertical feature parameters are input into the target quantitative detection model to output the defect size, movement speed and lifting distance of the target thin plate material.In this case, the kinematic eddy current signals in two directions, the parallel kinematic eddy current signals and the vertical kinematic eddy current signals, are comprehensively considered, and feature extraction is performed on the kinematic eddy current signals in the two directions to obtain corresponding multiple parallel feature parameters and multiple vertical feature parameters, and then a sample data set is constructed. A quantitative detection model is also constructed based on a neural network. The quantitative detection model includes a parallel feature parameter input module, a first quantitative detection module, a vertical feature parameter input module, a second quantitative detection module, a fusion module and an output module, wherein the parallel feature parameter input module is connected to the first quantitative detection module, the first quantitative detection module processes and analyzes the parallel feature parameters to obtain the corresponding motion speed, lifting distance and defect size, and the vertical feature parameter input module It is connected to the second quantification detection module, which processes and analyzes the vertical characteristic parameters to obtain the corresponding movement speed, lifting distance and defect size. The fusion module fuses the outputs of the first quantification detection module and the second quantification detection module to obtain the final movement speed, lifting distance and defect size. The quantification detection model is trained using the sample data set to obtain a target quantification detection model. The target quantification detection model can accurately quantify the defect size and accurately detect the movement condition. In actual scenarios, the target quantification detection model is used for the defect size quantification and movement condition detection of the target thin plate material to be quantified, so that the defect size can be accurately quantified and the movement condition can be accurately detected during the rapid movement of the thin plate material production.

[0031] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0033] Figure 1 A schematic flow chart of a method for quantifying defect sizes of thin plate materials in rapid motion and detecting motion conditions provided by an embodiment of the present invention;

[0034] Figure 2 A schematic diagram of a parallel motion-induced eddy current signal provided by an embodiment of the present invention;

[0035] Figure 3 A schematic diagram of a quantitative detection model provided by an embodiment of the present invention;

[0036] Figure 4 A schematic diagram of the training of the quantization detection model provided by an embodiment of the present invention;

[0037] Figure 5This is a block diagram of a system for quantifying defect sizes of thin plate materials in rapid motion and detecting motion conditions provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0039] The following describes a method and system for quantifying defect sizes of thin plate materials and detecting motion conditions during rapid motion according to an embodiment of the present invention with reference to the accompanying drawings.

[0040] The embodiment of the present invention provides a method for quantifying defect sizes of thin plate materials during rapid motion and detecting motion conditions, so as to accurately quantify defect sizes and accurately detect motion conditions during the rapid motion process of thin plate material production.

[0041] Figure 1 A schematic flow chart of a method for quantifying defect sizes of thin plate materials and detecting motion conditions during rapid motion provided by an embodiment of the present invention.

[0042] like Figure 1 As shown, the method for quantifying defect sizes of thin plate materials and detecting motion conditions during rapid motion includes the following steps:

[0043] Step S101 , obtaining two-axis motion-induced eddy current signals of defects in a test sheet material at different motion speeds, different lift-off distances, and different defect sizes, wherein the two-axis motion-induced eddy current signals include a parallel motion-induced eddy current signal and a vertical motion-induced eddy current signal.

[0044] In step S101, finite element software modeling simulation or actual measurement can be used to subject experimental thin plate materials with different defect sizes to different movement speeds and different lifting distances to obtain two-axis motion-induced eddy current signals of defects in experimental thin plate materials with different defect sizes at different movement speeds and different lifting distances.

[0045] For example, through actual measurements, we obtained the parallel and vertical motion-induced eddy current signals for through-hole defects of 100μm, 200μm, 300μm, 400μm, 500μm, 700μm, and 800μm on 0.1mm thick copper foil at a motion speed of 15m / s and a lift-off distance of 0.5mm. By resetting the motion speed and lift-off distance, we obtained a set of two-axis motion-induced eddy current signals for through-hole defects of 100μm, 200μm, 300μm, 400μm, 500μm, 700μm, and 800μm on 0.1mm thick copper foil at different motion speeds and lift-off distances.

[0046] For easy understanding, the movement speed in step S101 refers to the movement speed of the thin plate material. The lift-off distance refers to the distance between the thin plate material and the probe. The defect size includes the defect length, defect width, and defect depth of the thin plate material.

[0047] In step S101, the two-axis kinematic eddy current signals include a parallel kinematic eddy current signal and a perpendicular kinematic eddy current signal. The parallel kinematic eddy current signal is the parallel component, which refers to the direction of motion of the sheet material. The perpendicular kinematic eddy current signal is the perpendicular component, which refers to the direction perpendicular to the sheet material.

[0048] In step S102, feature extraction is performed on the parallel motion-induced eddy current signal and the vertical motion-induced eddy current signal to obtain a plurality of parallel feature parameters and a plurality of vertical feature parameters, and a sample data set is constructed based on the plurality of parallel feature parameters and the plurality of vertical feature parameters, as well as the corresponding motion speed, lifting distance and defect size.

[0049] In step S102, a variety of parallel characteristic parameters include but are not limited to the amplitude Up, peak-to-peak value Up-p, pulse width Dv-v, signal area Sa, signal energy E, pulse skewness I obtained by fast Fourier transform, and ratio of signal area to signal peak-to-peak value extracted from the parallel motion eddy current signal. Ratio of signal area to pulse width

[0050] Figure 2 This is a schematic diagram of a parallel motion-induced eddy current signal provided by an embodiment of the present invention. Figure 2 The parallel motion eddy current signal generated by the defect of the experimental sheet material is shown in FIG. Figure 2 Only some types of characteristic parameters are shown, for example, Up is the amplitude, Up-p is the peak-to-peak value, Dv-v is the pulse width, and Sa is the signal area.

[0051] In step S102, a variety of vertical characteristic parameters include but are not limited to the amplitude Up, peak-to-peak value Up-p, pulse width Dv-v, signal area Sa, signal energy E, pulse skewness I obtained by fast Fourier transform, and the ratio of signal area to signal peak-to-peak value extracted from the vertical motional eddy current signal. Ratio of signal area to pulse width

[0052] The amplitude is the difference between the peak value of the defect signal (i.e., the parallel motion eddy current signal or the vertical motion eddy current signal) and the baseline. The coordinate when the voltage is zero is taken as the baseline. The calculation formula is:

[0053] U p =min{p i}, i=0,1,2,…,N

[0054] Where N is the number of peaks. i is the i-th peak.

[0055] The peak-to-peak value is the absolute value of the difference between the larger peak value and the valley value of the defect signal. The calculation formula is:

[0056] U p-p =|max{p i}-min{p i}|

[0057] The pulse width is the time interval between the larger peak and the smaller peak in the defect signal in the time domain (it can also be the number of sampling points). The data point numbers at the larger peak and the smaller peak are m and n respectively, and the calculation formula is:

[0058] D v-v =|mn|

[0059] The calculation formula for the signal area is:

[0060]

[0061] The calculation formula of signal energy is:

[0062]

[0063] Where, Mean{p i} is the peak mean.

[0064] The calculation formula for pulse skewness is:

[0065]

[0066] Among them, μ tran is the average pulse amplitude obtained by fast Fourier transform. pTi represents the i-th frequency component after fast Fourier transform; M represents the total number of frequency components.

[0067] The formula for calculating the ratio of signal area to signal peak-to-peak value is:

[0068]

[0069] The ratio of signal area to pulse width is calculated as:

[0070]

[0071] Where, T s is the sampling period.

[0072] In step S102 , a sample data set is constructed based on a plurality of parallel characteristic parameters and a plurality of vertical characteristic parameters, as well as corresponding movement speeds, lift-off distances, and defect sizes.

[0073] Specifically, the sample dataset constructed in step S102 of this embodiment includes a first sample dataset and a second sample dataset. The first sample dataset includes multiple parallel feature parameters, as well as corresponding motion speeds, lift-off distances, and defect sizes. The multiple parallel feature parameters in the first sample dataset serve as input data for model training. The second sample dataset includes multiple vertical feature parameters, as well as corresponding motion speeds, lift-off distances, and defect sizes. The multiple vertical feature parameters in the second sample dataset serve as input data for model training.

[0074] In some embodiments, the first sample dataset includes a first quantized sample dataset and a first inspection sample dataset. The first quantized sample dataset includes multiple parallel feature parameters and corresponding defect sizes. The first inspection sample dataset includes multiple parallel feature parameters and corresponding motion speeds and lift-off distances. The second sample dataset includes a second quantized sample dataset and a second inspection sample dataset. The second quantized sample dataset includes multiple vertical feature parameters and corresponding defect sizes. The second inspection sample dataset includes multiple vertical feature parameters and corresponding motion speeds and lift-off distances.

[0075] Step S103, constructing a quantitative detection model based on a neural network, the quantitative detection model includes a parallel feature parameter input module, a first quantitative detection module, a vertical feature parameter input module, a second quantitative detection module, a fusion module and an output module, the parallel feature parameter input module is connected to the fusion module via the first quantitative detection module, the vertical feature parameter input module is connected to the fusion module via the second quantitative detection module, the fusion module is also connected to the output module, and the outputs of the first quantitative detection module and the second quantitative detection module both include movement speed, lifting distance and defect size.

[0076] In step S103, the parallel feature parameter input module is used to input a plurality of parallel feature parameters. The first quantitative detection module is used to process and analyze the plurality of parallel feature parameters to obtain corresponding movement speeds, lift-off distances, and defect sizes. The vertical feature parameter input module is used to input a plurality of vertical feature parameters. The second quantitative detection module is used to process and analyze the plurality of vertical feature parameters to obtain corresponding movement speeds, lift-off distances, and defect sizes. The fusion module is used to fuse the outputs of the first quantitative detection module and the second quantitative detection module to obtain the final movement speed, lift-off distance, and defect size. The output module is used to output the final movement speed, lift-off distance, and defect size.

[0077] In some embodiments, the first quantification detection module in step S103 includes a first defect quantification module and a first motion condition detection module, and the parallel characteristic parameter input module is connected to the fusion module via the first defect quantification module and the first motion condition detection module respectively; the second quantification detection module includes a second defect quantification module and a second motion condition detection module, and the vertical characteristic parameter input module is connected to the fusion module via the second defect quantification module and the second motion condition detection module respectively; the outputs of the first defect quantification module and the second defect quantification module both include defect size; the outputs of the first motion condition detection module and the second motion condition detection module both include motion speed and lift-off distance.

[0078] In some embodiments, the first defect quantization module and the second defect quantization module both use a convolutional network and a bidirectional long short-term memory neural network.

[0079] In some embodiments, the first motion condition detection module and the second motion condition detection module both use convolutional networks.

[0080] In some embodiments, the fusion module employs fully connected layers.

[0081] Figure 3 A schematic diagram of a quantitative detection model provided by an embodiment of the present invention. Figure 3 As shown, the parallel feature parameter input module is the first input layer, which is used to receive n types of parallel feature parameters, namely x1, x2, x3, ..., x n-2 、x n-1 、x n For example, when n is 8, the eight parallel characteristic parameters can be, for example, the amplitude Up, peak-to-peak value Up-p, pulse width Dv-v, signal area Sa, signal energy E, pulse skewness I obtained by fast Fourier transform, and the ratio of signal area to signal peak-to-peak value extracted from the parallel motion eddy current signal. Ratio of signal area to pulse width

[0082]

[0083] The first motion condition detection module is a first convolutional network, which can be a convolutional neural network (CNN). The input of the first convolutional network is connected to the first input layer. The output of the first convolutional network includes motion speed and liftoff distance.

[0084] The first quantization detection module includes the first convolutional layer and the first BILSTM (bidirectional long short-term memory) layer. The first convolutional layer can utilize a convolutional neural network (CNN). Using CNN and BILSTM can fully explore the relationship between the temporal and spatial features of the input data and the defect size. The output of the first BILSTM layer is the defect size.

[0085] The vertical feature parameter input module is the second input layer, which is used to receive n types of vertical feature parameters, namely z1, z2, z3, ..., z n-2 、z n-1 、z n For example, when n is 8, the eight vertical characteristic parameters can be, for example, the amplitude Up, peak-to-peak value Up-p, pulse width Dv-v, signal area Sa, signal energy E, pulse skewness I obtained by fast Fourier transform, and the ratio of signal area to signal peak-to-peak value extracted from the vertical motional eddy current signal. Ratio of signal area to pulse width

[0086] The second defect quantification module includes a second convolutional layer and a second BILSTM (bidirectional long short-term memory) layer. The second convolutional layer can use a convolutional neural network (CNN). The output of the second BILSTM layer is the defect size.

[0087] The second motion condition detection module is a second convolutional network, which can be a convolutional neural network (CNN). The input of the second convolutional network is connected to the second input layer. The output of the second convolutional network includes motion speed and liftoff distance.

[0088] The fusion module uses a fully connected layer. It is connected to the output of the first convolutional network, the output of the first BILSTM layer, the output of the second BILSTM layer, and the output of the second convolutional network. The fully connected layer fuses the output of the first convolutional network with the output of the second convolutional network, and also fuses the output of the first BILSTM layer with the output of the second BILSTM layer, thereby obtaining the final motion speed, lift-off distance, and defect size.

[0089] The output layer is used to output the final motion speed, lift-off distance and defect size.

[0090] Step S104 : using the sample data set to train the first quantization detection module and the second quantization detection module to obtain a target quantization detection model.

[0091] In step S104, the first and second sample data sets are used to train the first and second quantization detection modules, respectively, to obtain trained first and second quantization detection modules. The parallel feature parameter input module, the trained first quantization detection module, the perpendicular feature parameter input module, the trained second quantization detection module, the fusion module, and the output module constitute a target quantization detection model.

[0092] In step S104, the first sample data set may be further divided into a first sample training set and a first sample test set. The first sample training set may be used to train the first quantization detection module, and the first sample test set may be used to test the first quantization detection module, thereby making the model results more accurate. The training of the second quantization detection module may refer to the training of the first quantization detection module.

[0093] In some embodiments, a first sample validation set may be further divided from the first sample data set, and the tested first quantitative detection module may be further validated using the first sample validation set to further improve the accuracy of the model results.

[0094] Figure 4 Schematic diagram of the training of the quantization detection model provided by the embodiment of the present invention. Taking the training of the first quantization detection module as an example, Figure 4 As shown, the first sample data set is divided into a first sample training set and a first sample test set, the initial parameters of the first quantization detection module are set, the first sample training set is used to train the first quantization detection module to update the network parameters in the first quantization detection module, and then the first sample test set is used to test the trained first quantization detection module, thereby obtaining a trained first quantization detection module.

[0095] In step S104, when the first quantization detection module includes a first defect quantization module and a first motion condition detection module, the first defect quantization module is trained using the first quantization sample data set, and the first motion condition detection module is trained using the first detection sample data set. When the second quantization detection module includes a second defect quantization module and a second motion condition detection module, the second defect quantization module is trained using the second quantization sample data set, and the second motion condition detection module is trained using the second detection sample data set.

[0096] Step S105 , obtaining target two-axis motional eddy current signals of defects of the target thin plate material to be quantified, and then obtaining corresponding multiple target parallel characteristic parameters and multiple target perpendicular characteristic parameters.

[0097] In step S105, in actual application, target two-axis motional eddy current signals of defects in the target sheet material to be quantified are obtained. The target two-axis motional eddy current signals include target parallel motion eddy current signals and target vertical motion eddy current signals. Feature extraction is performed on the target parallel motion eddy current signals and target vertical motion eddy current signals, respectively, to obtain a plurality of target parallel feature parameters and a plurality of target vertical feature parameters. The types and number of target parallel feature parameters are consistent with the types and number of parallel feature parameters selected during training. The types and number of target vertical feature parameters are consistent with the types and number of vertical feature parameters selected during training.

[0098] Step S106 , inputting a plurality of target parallel characteristic parameters and a plurality of target vertical characteristic parameters into a target quantitative detection model to output defect size, movement speed and lifting distance of the target thin plate material.

[0099] In step S106, the multiple target parallel characteristic parameters and the multiple target perpendicular characteristic parameters are input into the target quantitative detection model, and the defect size of the target sheet material, the movement speed of the target sheet material, and the lift-off distance of the target sheet material are output. The defect size of the target sheet material includes the length, width, and depth of the defect of the target sheet material.

[0100] In other embodiments of the present invention, the quantitative detection model can be optimized to improve efficiency while ensuring the accuracy of the quantitative detection model. The optimized quantitative detection model includes a first input module, a first defect quantization module, a second input module, a first motion condition detection module, a third input module, a second defect quantization module, a fourth input module, a second motion condition detection module, a fusion module and an output module. The first input module is connected to the fusion module via the first defect quantization module, the second input module is connected to the fusion module via the first motion condition detection module, the third input module is connected to the fusion module via the second defect quantization module, and the fourth input module is connected to the fusion module via the second motion condition detection module. The input parameters of the first input module and the second input module are parallel characteristic parameters.

[0101] The parallel characteristic parameters of the first input module can be the eight parallel characteristic parameters in step S102, and the parallel characteristic parameters of the second input module can be the signal reference value U extracted from the parallel motion eddy current signal. f , signal slope and energy divergence The input parameters of the third input module and the fourth input module are vertical characteristic parameters, wherein the vertical characteristic parameters of the third input module can be the eight vertical characteristic parameters in step S102, and the vertical characteristic parameters of the fourth input module can be the signal reference value U extracted from the vertical motional eddy current signal. f, signal slope and energy divergence

[0102] A corresponding sample dataset is constructed for this optimized quantitative detection model to train the model, which is then applied to actual defect size quantification and motion condition detection of target sheet materials. It should be noted that in actual applications, the acquired feature parameters are consistent with the type and quantity of input data used to construct the corresponding sample dataset.

[0103] Compared with the quantization detection model before optimization, the optimized quantization detection model of the present invention has less input data when performing motion condition detection, which saves calculation time and improves efficiency. f , signal slope and energy divergence The correlation with the motion conditions is high, so although the number of input parameters is small, the accuracy of the quantitative detection model is still guaranteed.

[0104] In order to implement the above embodiment, the present invention further proposes a system for quantifying defect sizes of thin plate materials in rapid motion and detecting motion conditions.

[0105] Figure 5 This is a block diagram of a system for quantifying defect sizes of thin plate materials in rapid motion and detecting motion conditions provided by an embodiment of the present invention.

[0106] like Figure 5 As shown, the system for quantifying defect size of thin plate materials in rapid motion and detecting motion conditions includes a sample data set construction unit 11, a modeling unit 12, an acquisition unit 13, and a quantification detection unit 14, wherein:

[0107] The sample data set construction unit 11 is used to obtain two-axis motion-induced eddy current signals of defects in the experimental thin plate material under different motion speeds, different lift-off distances, and different defect sizes, where the two-axis motion-induced eddy current signals include parallel motion-induced eddy current signals and vertical motion-induced eddy current signals; feature extraction is performed on the parallel motion-induced eddy current signals and the vertical motion-induced eddy current signals to obtain a plurality of parallel feature parameters and a plurality of vertical feature parameters, and a sample data set is constructed based on the plurality of parallel feature parameters and the plurality of vertical feature parameters, as well as the corresponding motion speeds, lift-off distances, and defect sizes;

[0108] A modeling unit 12 is configured to construct a quantitative detection model based on a neural network. The quantitative detection model includes a parallel feature parameter input module, a first quantitative detection module, a vertical feature parameter input module, a second quantitative detection module, a fusion module, and an output module. The parallel feature parameter input module is connected to the fusion module via the first quantitative detection module, and the vertical feature parameter input module is connected to the fusion module via the second quantitative detection module. The fusion module is also connected to the output module. The outputs of the first quantitative detection module and the second quantitative detection module both include motion speed, lift-off distance, and defect size. The first quantitative detection module and the second quantitative detection module are trained using a sample data set to obtain a target quantitative detection model.

[0109] An acquisition unit 13 is used to acquire target two-axis motional eddy current signals of defects in a target sheet material to be quantified, and then obtain corresponding multiple target parallel characteristic parameters and multiple target perpendicular characteristic parameters;

[0110] The quantitative detection unit 14 is used to input multiple target parallel characteristic parameters and multiple target vertical characteristic parameters into the target quantitative detection model to output the defect size, movement speed and lifting distance of the target thin plate material.

[0111] Furthermore, in a possible implementation of an embodiment of the present invention, the sample data set construction unit 11 obtains two-axis motion-induced eddy current signals of defects in the experimental thin plate material under different movement speeds, different lifting distances, and different defect sizes through finite element software modeling simulation or actual measurement.

[0112] Furthermore, in a possible implementation of an embodiment of the present invention, the multiple parallel characteristic parameters in the sample data set construction unit 11 include the amplitude, peak-to-peak value, pulse width, signal area, signal energy, pulse deflection obtained by fast Fourier transform, the ratio of signal area to signal peak-to-peak value, and the ratio of signal area to pulse width extracted from the parallel motion-induced eddy current signal; the multiple vertical characteristic parameters include the amplitude, peak-to-peak value, pulse width, signal area, signal energy, pulse deflection obtained by fast Fourier transform, the ratio of signal area to signal peak-to-peak value, and the ratio of signal area to pulse width extracted from the vertical motion-induced eddy current signal.

[0113] Furthermore, in a possible implementation of an embodiment of the present invention, the first quantification detection module in the modeling unit 12 includes a first defect quantification module and a first motion condition detection module, and the parallel feature parameter input module is connected to the fusion module via the first defect quantification module and the first motion condition detection module respectively; the second quantification detection module includes a second defect quantification module and a second motion condition detection module, and the vertical feature parameter input module is connected to the fusion module via the second defect quantification module and the second motion condition detection module respectively; the outputs of the first defect quantification module and the second defect quantification module both include defect size; the outputs of the first motion condition detection module and the second motion condition detection module both include motion speed and lifting distance.

[0114] Furthermore, in a possible implementation of the embodiment of the present invention, the first defect quantification module and the second defect quantification module in the modeling unit 12 both adopt a convolutional network and a bidirectional long short-term memory neural network.

[0115] Furthermore, in a possible implementation of the embodiment of the present invention, both the first motion condition detection module and the second motion condition detection module in the modeling unit 12 adopt convolutional networks.

[0116] Furthermore, in a possible implementation of the embodiment of the present invention, the fusion module in the modeling unit 12 adopts a fully connected layer.

[0117] It should be noted that the above explanation of the embodiment of the method for quantifying the defect size of thin plate materials in rapid motion and detecting the motion condition is also applicable to the system for quantifying the defect size of thin plate materials in rapid motion and detecting the motion condition of this embodiment, and will not be repeated here.

[0118] In an embodiment of the present invention, two-axis motion-induced eddy current signals of defects in experimental thin plate materials under different motion speeds, different lifting distances, and different defect sizes are obtained, and the two-axis motion-induced eddy current signals include parallel motion-induced eddy current signals and vertical motion-induced eddy current signals; feature extraction is performed on the parallel motion-induced eddy current signals and the vertical motion-induced eddy current signals respectively to obtain a variety of parallel feature parameters and a variety of vertical feature parameters, and a sample data set is constructed based on the multiple parallel feature parameters and the multiple vertical feature parameters, as well as the corresponding motion speeds, lifting distances, and defect sizes; a quantitative detection model is constructed based on a neural network, and the quantitative detection model includes a parallel feature parameter input module, a first quantitative detection module, a vertical feature parameter input module, a second quantitative detection module, a fusion module, and an output module, and the parallel feature parameters are obtained. The characteristic parameter input module is connected to the fusion module via the first quantitative detection module, and the vertical characteristic parameter input module is connected to the fusion module via the second quantitative detection module. The fusion module is also connected to the output module. The outputs of the first quantitative detection module and the second quantitative detection module both include movement speed, lifting distance and defect size; the first quantitative detection module and the second quantitative detection module are trained using the sample data set to obtain a target quantitative detection model; the target two-axis motional eddy current signal of the defect of the target thin plate material to be quantified is obtained, and then the corresponding multiple target parallel characteristic parameters and multiple target vertical characteristic parameters are obtained; the multiple target parallel characteristic parameters and multiple target vertical characteristic parameters are input into the target quantitative detection model to output the defect size, movement speed and lifting distance of the target thin plate material.In this case, the kinematic eddy current signals in two directions, the parallel kinematic eddy current signals and the vertical kinematic eddy current signals, are comprehensively considered, and feature extraction is performed on the kinematic eddy current signals in the two directions to obtain corresponding multiple parallel feature parameters and multiple vertical feature parameters, and then a sample data set is constructed. A quantitative detection model is also constructed based on a neural network. The quantitative detection model includes a parallel feature parameter input module, a first quantitative detection module, a vertical feature parameter input module, a second quantitative detection module, a fusion module and an output module, wherein the parallel feature parameter input module is connected to the first quantitative detection module, the first quantitative detection module processes and analyzes the parallel feature parameters to obtain the corresponding motion speed, lifting distance and defect size, and the vertical feature parameter input module It is connected to the second quantification detection module, which processes and analyzes the vertical characteristic parameters to obtain the corresponding movement speed, lifting distance and defect size. The fusion module fuses the outputs of the first quantification detection module and the second quantification detection module to obtain the final movement speed, lifting distance and defect size. The quantification detection model is trained using the sample data set to obtain a target quantification detection model. The target quantification detection model can accurately quantify the defect size and accurately detect the movement condition. In actual scenarios, the target quantification detection model is used for the defect size quantification and movement condition detection of the target thin plate material to be quantified, so that the defect size can be accurately quantified and the movement condition can be accurately detected during the rapid movement of the thin plate material production.

[0119] The method and system of the present invention extract the characteristics of the two-axis kinematic eddy current signal and establish a neural network model based on the motion speed, lift-off distance, length, width and depth of the defect, thereby achieving accurate quantification of the defect size of different thin plate materials and detection of the motion speed and lift-off distance. Considering that the defect size alone cannot achieve accurate positioning of the defect, and the positioning of the defect is also related to the motion conditions, the present invention takes into account the accurate quantification of the defect size and the detection of the motion speed and lift-off distance, so that the subsequent comprehensive defect size, motion speed and lift-off distance can achieve accurate positioning of the defect. The method and system of the present invention are suitable for the rapid production process of thin plate materials and can be widely used in the field of non-destructive detection of thin plate material defects. It has great engineering significance and good application prospects.

[0120] In order to implement the above embodiments, the present invention also proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.

[0121] In order to implement the above embodiments, the present invention further provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided in the above embodiments.

[0122] In order to implement the above embodiments, the present invention further provides a computer program product, including a computer program, which implements the methods provided in the above embodiments when executed by a processor.

[0123] In the descriptions of the foregoing embodiments, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and features of different embodiments or examples, unless they are mutually inconsistent.

[0124] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0125] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0126] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary, and then storing it in a computer memory.

[0127] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0128] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0129] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing module, or each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.

[0130] The storage medium mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and are not to be construed as limiting the present invention. Persons skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for quantifying defect sizes of thin plate materials during rapid motion and detecting motion conditions, characterized in that: include: Obtaining two-axis motional eddy current signals of defects in the test sheet material at different motion speeds, different lift-off distances, and different defect sizes, wherein the two-axis motional eddy current signals include a parallel motional eddy current signal and a vertical motional eddy current signal; performing feature extraction on the parallel motion-induced eddy current signal and the vertical motion-induced eddy current signal to obtain a plurality of parallel feature parameters and a plurality of vertical feature parameters, respectively, and constructing a sample data set based on the plurality of parallel feature parameters and the plurality of vertical feature parameters, as well as corresponding motion speeds, lift-off distances, and defect sizes; A quantitative detection model is constructed based on a neural network, wherein the quantitative detection model includes a parallel feature parameter input module, a first quantitative detection module, a vertical feature parameter input module, a second quantitative detection module, a fusion module, and an output module. The parallel feature parameter input module is connected to the fusion module via the first quantitative detection module, and the vertical feature parameter input module is connected to the fusion module via the second quantitative detection module. The fusion module is also connected to the output module. The outputs of the first quantitative detection module and the second quantitative detection module both include movement speed, lift-off distance, and defect size. Using the sample data set to train the first quantization detection module and the second quantization detection module to obtain a target quantization detection model; Obtaining target two-axis motional eddy current signals of defects in a target thin plate material to be quantified, and then obtaining corresponding multiple target parallel characteristic parameters and multiple target perpendicular characteristic parameters; The plurality of target parallel characteristic parameters and the plurality of target vertical characteristic parameters are input into a target quantitative detection model to output the defect size, movement speed and lifting distance of the target thin plate material.

2. The method for quantifying defect sizes and detecting motion conditions of thin plate materials in rapid motion according to claim 1, characterized in that: The first quantification detection module includes a first defect quantification module and a first motion condition detection module, and the parallel characteristic parameter input module is connected to the fusion module via the first defect quantification module and the first motion condition detection module respectively; the second quantification detection module includes a second defect quantification module and a second motion condition detection module, and the vertical characteristic parameter input module is connected to the fusion module via the second defect quantification module and the second motion condition detection module respectively; the outputs of the first defect quantification module and the second defect quantification module both include defect size; the outputs of the first motion condition detection module and the second motion condition detection module both include motion speed and lifting distance.

3. The method for quantifying defect sizes and detecting motion conditions of thin plate materials in rapid motion according to claim 2, characterized in that: The first defect quantification module and the second defect quantification module both adopt convolutional networks and bidirectional long short-term memory neural networks.

4. The method for quantifying defect sizes and detecting motion conditions of thin plate materials in rapid motion according to claim 2, characterized in that: The first motion condition detection module and the second motion condition detection module both adopt convolutional networks.

5. The method for quantifying defect sizes of thin plate materials and detecting motion conditions during rapid motion according to claim 1, characterized in that: The fusion module adopts a fully connected layer.

6. The method for quantifying defect sizes of thin plate materials and detecting motion conditions during rapid motion according to claim 1, characterized in that: The multiple parallel characteristic parameters include the amplitude, peak-to-peak value, pulse width, signal area, signal energy, pulse deflection obtained by fast Fourier transform, the ratio of signal area to signal peak-to-peak value, and the ratio of signal area to pulse width extracted from the parallel motion-induced eddy current signal; the multiple vertical characteristic parameters include the amplitude, peak-to-peak value, pulse width, signal area, signal energy, pulse deflection obtained by fast Fourier transform, the ratio of signal area to signal peak-to-peak value, and the ratio of signal area to pulse width extracted from the vertical motion-induced eddy current signal.

7. The method for quantifying defect sizes and detecting motion conditions of thin plate materials in rapid motion according to claim 1, characterized in that: The two-axis motion-induced eddy current signals of defects in the experimental thin plate material under different movement speeds, different lifting distances, and different defect sizes are obtained through finite element software modeling and simulation or actual measurement.

8. A system for quantifying defect sizes of thin plate materials during rapid motion and detecting motion conditions, characterized in that: include: A sample data set construction unit is used to obtain two-axis motion-induced eddy current signals of defects in experimental thin plate materials under different motion speeds, different lift-off distances, and different defect sizes, wherein the two-axis motion-induced eddy current signals include parallel motion-induced eddy current signals and vertical motion-induced eddy current signals; feature extraction is performed on the parallel motion-induced eddy current signals and the vertical motion-induced eddy current signals to obtain a plurality of parallel feature parameters and a plurality of vertical feature parameters, and a sample data set is constructed based on the plurality of parallel feature parameters and the plurality of vertical feature parameters, as well as the corresponding motion speeds, lift-off distances, and defect sizes; a modeling unit for constructing a quantitative detection model based on a neural network, the quantitative detection model comprising a parallel feature parameter input module, a first quantitative detection module, a vertical feature parameter input module, a second quantitative detection module, a fusion module, and an output module, wherein the parallel feature parameter input module is connected to the fusion module via the first quantitative detection module, the vertical feature parameter input module is connected to the fusion module via the second quantitative detection module, and the fusion module is further connected to the output module, wherein the outputs of the first quantitative detection module and the second quantitative detection module both include motion speed, lift-off distance, and defect size; and the first and second quantitative detection modules are trained using the sample data set to obtain a target quantitative detection model; An acquisition unit is used to acquire target two-axis motional eddy current signals of defects in a target sheet material to be quantified, and then obtain corresponding multiple target parallel characteristic parameters and multiple target perpendicular characteristic parameters; The quantitative detection unit is used to input the multiple target parallel characteristic parameters and the multiple target vertical characteristic parameters into the target quantitative detection model to output the defect size, movement speed and lifting distance of the target thin plate material.

9. An electronic device, characterized in that: include: a processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 7 when executed by a processor.

Images