CN114359193A - Defect classification method and system based on ultrasonic phased array imaging - Google Patents
Defect classification method and system based on ultrasonic phased array imaging Download PDFInfo
- Publication number
- CN114359193A CN114359193A CN202111593864.5A CN202111593864A CN114359193A CN 114359193 A CN114359193 A CN 114359193A CN 202111593864 A CN202111593864 A CN 202111593864A CN 114359193 A CN114359193 A CN 114359193A
- Authority
- CN
- China
- Prior art keywords
- image
- simulation
- defect
- data
- ultrasonic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)
Abstract
The invention provides a defect classification method and a system based on ultrasonic phased array imaging, which belong to the field of nondestructive testing, and the method comprises the following steps: acquiring ultrasonic full matrix data by adopting ultrasonic phased array imaging on a sample piece to be detected; carrying out full focusing processing on the ultrasonic full matrix data, carrying out color coding according to the signal amplitude, and obtaining an image of a sample piece to be detected; preprocessing an image of a sample to be detected, inputting the preprocessed image into a classification prediction model, and acquiring defect classification of the sample to be detected; the method for training the classification prediction model comprises the following steps: acquiring simulation data by simulating ultrasonic phased array imaging by using a finite element simulation method of defect ultrasonic scattering data; carrying out full focusing processing on the simulation data to obtain a simulation image; carrying out data enhancement after the simulation image is preprocessed; carrying out image feature extraction on the simulation image by adopting a convolutional neural network; and inputting the image characteristics into a full connection layer, and training a classification prediction model by taking defect classification as output. The invention improves the precision of defect classification.
Description
Technical Field
The invention belongs to the field of nondestructive testing, and particularly relates to a defect classification method and system based on ultrasonic phased array imaging.
Background
The development and application of the ultrasonic nondestructive testing technology are established on the basis of the interaction of ultrasonic waves and a tested object; when the ultrasonic wave with good guidance quality encounters a defect in the propagation process, the propagation direction or the characteristics of the ultrasonic wave change, and the detection and characterization of the workpiece defect can be realized by researching the reflection, refraction and scattering. Compared with other nondestructive testing methods, the ultrasonic nondestructive testing has unique advantages; the method is widely applicable to nondestructive evaluation of metal, nonmetal and composite materials; the penetration capability is strong, the detection depth is large, and the positioning capability on the internal defects of the workpiece is good; high sensitivity, low cost and no harm to human body.
In the existing ultrasonic defect detection technology, the detection aiming at the pore type defect is lacked, namely, a plurality of small pore parametric data with small size and close distance are lacked. In addition, the conventional B-mode ultrasonic imaging has low resolution and poor imaging precision, and has no resolving power for a plurality of small defects (the size is less than 0.8 wavelength) at a short distance. In common classification methods, dimension reduction algorithms such as Linear Discriminant Analysis (LDA) and Quadratic Discriminant Analysis (QDA) can simplify high-dimensional data to process a large-scale data set, but the significance of the result is difficult to understand; the naive Bayes method is classified based on probability theory, so the method is limited by Bayes theorem and conditional independence hypothesis, but the conditional independence hypothesis is usually not established in practical application, thereby influencing the classification effect of the Bayes method.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a defect classification method and a system based on ultrasonic phased array imaging, and aims to adopt full-focus imaging to the defects of a detected object, perform defect classification after feature extraction by using a convolutional neural network, and improve the precision of defect classification.
To achieve the above object, in one aspect, the present invention provides a defect classification based on ultrasonic phased array imaging, comprising the following steps:
acquiring ultrasonic full matrix data by adopting ultrasonic phased array imaging on a sample piece to be detected;
after the ultrasonic full matrix data is subjected to full focusing processing, color coding is carried out according to the signal amplitude value, and an image of a sample piece to be detected is obtained;
preprocessing an image of a sample to be tested, inputting the preprocessed image into a trained classification prediction model, and obtaining defect classification of the sample to be tested;
the classification prediction model is obtained by adding a full connection layer to the output layer of the convolutional neural network;
the method for training the classification prediction model comprises the following steps:
setting simulation parameters by using a finite element simulation method of defect ultrasonic scattering data, and acquiring simulation data through simulation ultrasonic phased array imaging;
carrying out full focusing processing on the simulation data, carrying out color coding according to the signal amplitude, and obtaining a simulation image;
carrying out data enhancement on the preprocessed simulation image to obtain an enhanced simulation image;
carrying out image feature extraction on the enhanced simulation image by adopting the trained convolutional neural network;
and inputting the image characteristics and the corresponding class labels into a full connection layer, and training a classification prediction model.
Further preferably, the simulation parameters include sample material, ultrasonic detection frequency, sample defect distribution, sample defect number, sample defect size and phased array probe parameters;
the phased array probe parameters include probe size, array element number, center frequency, bandwidth and array element spacing.
Further preferably, the method for preprocessing the simulation image comprises:
cutting an edge image irrelevant to defect classification according to the defect distribution position, and reserving a defect area;
dividing the cut simulation image into different folders according to categories to finish the data set manufacturing;
and performing label setting on the data set by adopting thermal coding.
Further preferably, the method for enhancing the data of the simulation image in the data set comprises:
carrying out normalization processing on the simulation image in the data set;
and (3) performing rotation and/or horizontal translation and/or vertical translation and/or random horizontal overturning and/or filling in a nearest mode on the simulation image after the normalization processing.
Further preferably, the method for preprocessing the image of the sample to be tested comprises:
interactively cutting the image of the sample piece to be detected, and reserving a defect area to be classified;
filling the upper edge, the lower edge, the left edge and the right edge of the image with the defect area in the height and width directions respectively by three RGB channels; wherein the filled color is consistent with the image color of the non-defective region;
and combining the three channel images to ensure that the size of the filled image is consistent with the size of the input image of the convolutional neural network.
Further preferably, the convolutional neural network is a vgg (visual Geometry Group network)16 convolutional neural network with the top layer removed, and includes 13 convolutional layers and 5 pooling layers.
Further preferably, the activation function of the last fully-connected layer of the classification prediction model adopts a softmax activation function, specifically:
wherein the Softmax function transforms the vector (a)1，a2，…，an) Mapping as a vector (S)1，S2，…，Sn) Wherein n is the number of categories; a isjRepresents the input value of the jth output node;
is a normalized coefficient; sjShowing the output after Softmax calculation.
In another aspect, the present invention provides a defect classification system based on ultrasonic phased array imaging, including:
the ultrasonic phased array probe is used for imaging a sample piece to be detected by adopting an ultrasonic phased array to acquire ultrasonic full matrix data;
the full-focusing module is used for carrying out full-focusing processing on the ultrasonic full-matrix data and the simulation data, carrying out color coding according to the signal amplitude, and respectively obtaining an image and a simulation image of the sample piece to be detected;
the simulation module is used for setting simulation parameters by using a finite element simulation method of the defect ultrasonic scattering data and acquiring simulation data by simulating ultrasonic phased array imaging;
the image preprocessing module is used for preprocessing the image and the simulation image of the sample piece to be detected;
the classification prediction module is used for inputting the preprocessed image of the sample piece to be detected into the trained column prediction model to obtain the defect classification of the sample piece to be detected;
the data enhancement module is used for enhancing the data of the preprocessed simulation image to obtain an enhanced simulation image;
the classification prediction model is obtained by adding a full connection layer to the output layer of the convolutional neural network;
the training method of the classification prediction module comprises the following steps:
carrying out image feature extraction on the enhanced simulation image by adopting the trained convolutional neural network;
and inputting the image characteristics and the corresponding class labels into a full connection layer, and training a classification prediction model.
Further preferably, the activation function of the last fully-connected layer of the classification prediction model adopts a softmax activation function, and the softmax activation function is as follows:
wherein the Softmax function transforms the vector (a)1，a2，…，an) Mapping as a vector (S)1，S2，…，Sn) Wherein n is the number of categories; a isjRepresents the input value of the jth output node;
is a normalized coefficient; sjShowing the output after Softmax calculation.
Further preferably, the method for preprocessing the image of the sample to be tested comprises:
interactively cutting the image of the sample piece to be detected, and reserving a defect area to be classified;
filling the upper edge, the lower edge, the left edge and the right edge of the image with the defect area in the height and width directions respectively by three RGB channels; wherein the filled color is consistent with the image color of the non-defective area;
merging the three channel images to ensure that the size of the filled image is consistent with the size of an input image of the convolutional neural network;
the method for preprocessing the simulation image comprises the following steps:
cutting an edge image irrelevant to defect classification according to the defect distribution position of the simulation image, and reserving a defect area;
dividing the cut simulation image into different folders according to categories to finish the data set manufacturing;
and performing label setting on the data set by adopting thermal coding.
Further preferably, the simulation parameters include sample material, ultrasonic detection frequency, sample defect distribution, sample defect number, sample defect size and phased array probe parameters;
the phased array probe parameters comprise probe size, array element number, center frequency, bandwidth and array element spacing.
Further preferably, the method for enhancing the data of the simulated image in the data set comprises:
normalizing the simulation image in the data set;
and (3) performing rotation and/or horizontal translation and/or vertical translation and/or random horizontal overturning and/or filling in a nearest mode on the simulation image after the normalization processing.
Further preferably, the convolutional neural network is a VGG16 convolutional neural network with the top layer removed, and comprises 13 convolutional layers and 5 pooling layers.
Generally, compared with the prior art, the above technical solution conceived by the present invention has the following beneficial effects:
when the classification prediction model is trained, any defect is simulated by using a defect ultrasonic scattering data finite element simulation method, and a simulation image is obtained after full focusing processing is adopted after simulation ultrasonic phased array imaging, wherein the simulation image is used for constructing a training set of the classification prediction model, and the finite element simulation model considers multiple scattering of ultrasonic waves among a plurality of defects, so that the constructed simulation image can accurately reflect the defect problem, and the trained classification prediction model has higher precision on defect classification.
The invention adopts an advanced data post-processing method to carry out imaging, namely, a full focusing method is adopted to carry out delay superposition processing on the ultrasonic full matrix data acquired by the ultrasonic phased array so as to obtain the internal imaging result of the object to be measured, and all acquired signals can be fully utilized, so that the imaging resolution is far higher than that of the traditional ultrasonic B scanning imaging.
The classification prediction model adds a full connection layer to the convolutional neural network output layer, extracts image characteristics by using the trained convolutional neural network, and greatly shortens the calculation and training time of the classification prediction model because only network parameters of a classification part in the classification prediction model need to be trained. The classification prediction model provided by the invention has a self-updating function, namely, the model can be trained and updated when new data is added.
Drawings
FIG. 1 is a flowchart of a defect classification method based on ultrasonic phased array imaging according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a random distribution of defect locations provided by an embodiment of the present invention;
FIG. 3 is a schematic diagram of an ultrasonic full focus imaging system provided by an embodiment of the present invention;
FIG. 4 is a schematic diagram of a classification prediction model provided by an embodiment of the present invention;
fig. 5 is a schematic diagram of the precision variation of the training process provided by the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As shown in fig. 1, the present invention provides a defect classification method based on ultrasonic phased array imaging, and the whole implementation process is as follows: designing four types of circular hole defects by taking the size and the density of the circular hole as defect classification parameters and processing an actual sample; acquiring simulation data and experimental data by combining parameters such as detection frequency, bandwidth percentage, the number of ultrasonic phased array elements and the like; carrying out full-focus imaging processing on the simulation data and the experimental data; after corresponding data set division and preprocessing are carried out on the image, finally, a convolution neural network and a full-connection neural network are utilized to classify the imaging result;
examples
As shown in fig. 1, in one aspect, the present embodiment provides a defect classification method based on ultrasonic phased array imaging, including the following steps:
the method comprises the following steps: setting simulation parameters during simulation, and simulating ultrasonic phased array imaging to obtain simulation data; :
the simulation parameters comprise: selecting a material to be detected, detecting frequency by ultrasonic waves, and detecting size and density of the circular hole;
the material selected in this example is aluminum; setting the frequency of an ultrasonic phased array probe to be 10MHz in simulation; the air hole defects with different parameters in four groups are obtained through shape modeling of the air hole defects, and the method specifically comprises the following steps: the radius of the first type of defect round hole is 0.5mm, and the number of defects is 3; the radius of the second type of defect circular hole is 0.45mm, and the number of defects is 5; the radius of the third type defect circular hole is 0.4mm, and the number of defects is 7; the radius of the fourth type defect circular hole is 0.35mm, and the number of defects is 11; the distribution of the defects adopts random distribution, as shown in FIG. 2; storing the simulation data into a simul _ x.mat file;
step two: designing parameters of an experimental sample during experiment, and processing the experimental sample;
the sample parameters include: the size and density of the circular hole, the detection frequency and the bandwidth percentage; processing a test sample by using a linear cutting mode;
in the embodiment, a 6061 aluminum material is used for manufacturing a sample, a wire cut electrical discharge machining mode is adopted for machining a cylindrical hole to simulate defects, the design method of the size and the random distribution position of the round hole is consistent with the design method in the first step, and the position of the round hole is distributed within the range of 300 +/-5 mm from the detection contact surface;
step three: acquiring ultrasonic full matrix data, namely experimental data, from an experimental sample by adopting an ultrasonic phased array probe;
in the embodiment, an ultrasonic phased array probe with 64 array elements linearly arranged is adopted, the spacing between the array elements is 0.3mm, the frequency is 10MHz, and the length of the array elements is 5 mm; an Explorer open type ultrasonic phased array platform which is introduced by American AOS company is selected and can control 64 independent transmitting-receiving channels corresponding to the number of probe array elements in parallel; sequentially exciting each array element by adopting rectangular pulses, wherein all the array elements in each transmission are used as receivers, so that full matrix data of time domain signals of each transmitting-receiving array element group are obtained;
step four: performing full-focusing processing on the acquired simulation data and experimental data, performing color coding according to the amplitude value, and displaying and storing corresponding images (simulation images and experimental images), as shown in fig. 3;
in the embodiment, the array elements are sequentially excited, the excitation delay time of each array element is calculated according to the geometric acoustic principle, and the focusing of the wave beam at a designated position point is realized by delaying and superposing echo signals received by each channel of the phased array;
in the data processing process, acceleration calculation is carried out in two modes, wherein one mode is parallel processing by using a graphic accelerator; secondly, simplifying the superposition of each pixel point by adopting a triangular matrix to replace the original full matrix data according to the sound field reciprocity theorem; the final signal amplitude at the focus point is as follows:
wherein S isrefThe amplitude of a signal received on a sound wave propagation path passing through the focusing point; t is tpAnd tqRespectively the time required by the sound wave from the transmitting array element to the focusing point and the time required by the sound wave from the focusing point to the receiving array element; deltapqThe weighting coefficient is used for controlling the number of times of each data in the full matrix to participate in calculation so as to meet the reciprocity theorem; when p is q, δpq1 is ═ 1; otherwise, δ pq0; the resulting image is saved in classX _ xxx. png format;
step five: preprocessing the simulation image, including automatically cutting and reserving a defect area, classifying according to the defect design and carrying out single hot coding;
in the embodiment, edge images irrelevant to defect classification are automatically cut and removed according to defect distribution positions, defect areas are reserved, and the defect areas are divided according to categories and stored in a folder to finish data set manufacturing; performing label setting by adopting one-hot coding, wherein the label setting is (1,0,0,0), (0,1,0,0), (0,0,1,0) and (0,0,0, 1);
step six: carrying out data enhancement on the simulation image in the data set;
in the embodiment, firstly, the simulation image is normalized, and then data enhancement is performed by adopting methods of rotating by 0-40 degrees, horizontally translating by 0-0.2 image widths, vertically translating by 0-0.2 image heights, randomly horizontally turning a general image, filling in a nearest mode and the like;
step seven: extracting the simulation image features after data enhancement by using the trained convolutional neural network;
in this embodiment, as shown in fig. 4, the trained convolutional neural network is a VGG16 convolutional neural network; performing feature extraction on the simulation image after data enhancement by using the pre-training VGG16 convolutional neural network with the top layer removed for training a classification prediction model;
the VGG16 convolutional neural network after the top layer is removed comprises 13 convolutional layers, namely conv3-XXX, and five pooling layers, namely maxpool; wherein, the convolution layer adopts convolution kernel with the size of 3 × 3; taking 20 training samples as a group, generating a picture group input _ batch and a label group labelsjbatch by using a picture generator in a circulating way, and inputting the picture group input _ batch and the label group labelsjbatch into a VGG16 convolutional neural network for feature extraction;
step eight: and (5) establishing a defect image classification prediction model by using a convolutional neural network, and using the simulated image features obtained in the step seven for classification prediction model training, as shown in fig. 5.
In this embodiment, the classification prediction model is a DNN neural network model, and includes a four-layer structure, except for the last fully-connected layer, the activation functions of other fully-connected layers use relu, and the formula is as follows:
and the activation function of the last layer of full connection layer adopts a softmax activation function, and the formula is as follows:
the Softmax function transforms the vector (a)1，a2，…，an) Mapping as a vector (S)1，S2，…，Sn) Wherein n is the number of categories; a isjRepresents the input value of the jth output node;
is a normalized coefficient; sjThe output result after the calculation of Softmax is shown; the Softmax function is used for probability calculation of multi-classification tasks; dropout was set to 0.5, and a portion of neurons were randomly inactivated to prevent overfitting; the data set is scrambled in a random index arrangement mode, 30% of the data set is taken as a verification set, and the accuracy and the loss of the coordinated _ cross control corresponding to the training set and the verification set are printed after each round of training is completed;
step nine: preprocessing an experimental image;
in the embodiment, a mouse is adopted for interactively cutting and reserving a defect area to be classified; filling the upper edge, the lower edge, the left edge and the right edge of the experimental image with equal height or equal width respectively by three RGB channels; merging the three channel images to ensure that the filled experimental image is consistent with the input image size of the VGG16 convolutional neural network in the seventh step; the fill color is consistent with the image color at the defect-free location, which is [0, 143] in this example;
step ten: and predicting the preprocessed experimental image in the ninth step by using a classification prediction model.
In another aspect, the present invention provides a defect classification system based on ultrasonic phased array imaging, including:
the ultrasonic phased array probe is used for imaging a sample piece to be detected by adopting an ultrasonic phased array to acquire ultrasonic full matrix data;
the full-focusing module is used for carrying out full-focusing processing on the ultrasonic full-matrix data and the simulation data, carrying out color coding according to the signal amplitude, and respectively obtaining an image and a simulation image of the sample piece to be detected;
the simulation module is used for setting simulation parameters by using a finite element simulation method of the defect ultrasonic scattering data and acquiring simulation data by simulating ultrasonic phased array imaging;
the image preprocessing module is used for preprocessing the image and the simulation image of the sample piece to be detected;
the classification prediction module is used for inputting the preprocessed image of the sample piece to be detected into the trained column prediction model to obtain the defect classification of the sample piece to be detected;
the data enhancement module is used for enhancing the data of the preprocessed simulation image to obtain an enhanced simulation image;
the classification prediction model is obtained by adding a full connection layer to the output layer of the convolutional neural network;
the training method of the classification prediction module comprises the following steps:
carrying out image feature extraction on the enhanced simulation image by adopting the trained convolutional neural network;
and inputting the image characteristics and the corresponding class labels into a full connection layer, and training a classification prediction model.
Further preferably, the activation function of the last fully-connected layer of the classification prediction model adopts a softmax activation function, and the softmax activation function is as follows:
wherein the Softmax function transforms the vector (a)1，a2，…，an) Is mapped asVector (S)1，S2，…，Sn) Wherein n is the number of categories; a isjRepresents the input value of the jth output node;
is a normalized coefficient; sjShowing the output after Softmax calculation.
Further preferably, the method for preprocessing the image of the sample to be tested comprises:
interactively cutting the image of the sample piece to be detected, and reserving a defect area to be classified;
filling the upper edge, the lower edge, the left edge and the right edge of the image with the defect area in the height and width directions respectively by three RGB channels; wherein the filled color is consistent with the image color of the non-defective area;
merging the three channel images to ensure that the size of the filled image is consistent with the size of an input image of the convolutional neural network;
the method for preprocessing the simulation image comprises the following steps:
cutting an edge image irrelevant to defect classification according to the defect distribution position of the simulation image, and reserving a defect area;
dividing the cut simulation image into different folders according to categories to finish the data set manufacturing;
and performing label setting on the data set by adopting thermal coding.
Further preferably, the simulation parameters include sample material, ultrasonic detection frequency, sample defect distribution, sample defect number, sample defect size and phased array probe parameters;
the phased array probe parameters comprise probe size, array element number, center frequency, bandwidth and array element spacing.
Further preferably, the method for enhancing the data of the simulated image in the data set comprises:
normalizing the simulation image in the data set;
and (3) performing rotation and/or horizontal translation and/or vertical translation and/or random horizontal overturning and/or filling in a nearest mode on the simulation image after the normalization processing.
Further preferably, the convolutional neural network is a VGG16 convolutional neural network with the top layer removed, and comprises 13 convolutional layers and 5 pooling layers.
In summary, compared with the prior art, the invention has the following advantages:
when the classification prediction model is trained, any defect is simulated by using a defect ultrasonic scattering data finite element simulation method, and a simulation image is obtained after full focusing processing is adopted after simulation ultrasonic phased array imaging, wherein the simulation image is used for constructing a training set of the classification prediction model, and the finite element simulation model considers multiple scattering of ultrasonic waves among a plurality of defects, so that the constructed simulation image can accurately reflect the defect problem, and the trained classification prediction model has higher precision on defect classification.
The invention adopts an advanced data post-processing method to carry out imaging, namely, a full focusing method is adopted to carry out delay superposition processing on the ultrasonic full matrix data acquired by the ultrasonic phased array so as to obtain the internal imaging result of the object to be measured, and all acquired signals can be fully utilized, so that the imaging resolution is far higher than that of the traditional ultrasonic B scanning imaging.
The classification prediction model adds a full connection layer to the convolutional neural network output layer, extracts image characteristics by using the trained convolutional neural network, and greatly shortens the calculation and training time of the classification prediction model because only network parameters of a classification part in the classification prediction model need to be trained. The classification prediction model provided by the invention has a self-updating function, namely, the model can be trained and updated when new data is added.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (10)
1. A defect classification method based on ultrasonic phased array imaging is characterized by comprising the following steps:
acquiring ultrasonic full matrix data by adopting ultrasonic phased array imaging on a sample piece to be detected;
after the ultrasonic full matrix data is subjected to full focusing processing, color coding is carried out according to the signal amplitude value, and an image of a sample piece to be detected is obtained;
preprocessing the image of the sample to be tested, inputting the preprocessed image into a trained classification prediction model, and obtaining the defect classification of the sample to be tested;
the classification prediction model is obtained by adding a full connection layer to the output layer of the convolutional neural network;
the method for training the classification prediction model comprises the following steps:
setting simulation parameters by using a finite element simulation method of defect ultrasonic scattering data, and acquiring simulation data through simulation ultrasonic phased array imaging;
after full focusing processing is carried out on the simulation data, color coding is carried out according to the signal amplitude, and then a simulation image is obtained;
carrying out data enhancement on the preprocessed simulation image to obtain an enhanced simulation image;
carrying out image feature extraction on the enhanced simulation image by adopting the trained convolutional neural network;
and inputting the image characteristics and the corresponding class labels into a full connection layer, and training a classification prediction model.
2. The defect classification method of claim 1, wherein the simulation parameters include sample material, ultrasonic inspection frequency, sample defect distribution, sample defect number, sample defect size, and phased array probe parameters;
the phased array probe parameters comprise probe size, array element number, center frequency, bandwidth and array element spacing.
3. The defect classification method according to claim 1 or 2, wherein the method for preprocessing the simulation image is as follows:
cutting an edge image irrelevant to defect classification according to the defect distribution position of the simulation image, and reserving a defect area;
dividing the cut simulation image into different folders according to categories to finish the data set manufacturing;
and performing label setting on the data set by adopting thermal coding.
4. The defect classification method according to claim 1 or 2, wherein the method for preprocessing the image of the sample to be tested comprises:
interactively cutting the image of the sample piece to be detected, and reserving a defect area to be classified;
filling the upper edge, the lower edge, the left edge and the right edge of the image with the defect area in the height and width directions respectively by three RGB channels; wherein the filled color is consistent with the image color of the non-defective area;
and combining the three channel images to ensure that the size of the filled image is consistent with the size of the input image of the convolutional neural network.
5. The defect classification method of claim 4, wherein the method for enhancing the simulation image in the data set comprises:
normalizing the simulation image in the data set;
and (3) performing rotation and/or horizontal translation and/or vertical translation and/or random horizontal overturning and/or filling in a nearest mode on the simulation image after the normalization processing.
6. The defect classification method of claim 1, wherein the convolutional neural network is a VGG16 convolutional neural network with the top layer removed, and comprises 13 convolutional layers and 5 pooling layers.
7. The defect classification method according to claim 1 or 6, characterized in that the activation function of the last fully-connected layer of the classification prediction model adopts a softmax activation function, and the softmax activation function is:
wherein the Softmax function transforms the vector (a)1，a2，…，an) Mapping as a vector (S)1，S2，…，Sn) Wherein n is the number of categories; a isjRepresents the input value of the jth output node;
is a normalized coefficient; sjShowing the output after Softmax calculation.
8. A defect classification system based on ultrasonic phased array imaging, comprising:
the ultrasonic phased array probe is used for imaging a sample piece to be detected by adopting an ultrasonic phased array to acquire ultrasonic full matrix data;
the full-focusing module is used for carrying out full-focusing processing on the ultrasonic full-matrix data and the simulation data, carrying out color coding according to the signal amplitude, and respectively obtaining an image and a simulation image of the sample piece to be detected;
the simulation module is used for setting simulation parameters by using a finite element simulation method of the defect ultrasonic scattering data and acquiring simulation data by simulating ultrasonic phased array imaging;
the image preprocessing module is used for preprocessing the image and the simulation image of the sample piece to be detected;
the data enhancement module is used for enhancing the data of the preprocessed simulation image to obtain an enhanced simulation image;
the classification prediction module is used for inputting the preprocessed image of the sample piece to be detected into the trained column prediction model to obtain the defect classification of the sample piece to be detected;
the classification prediction model is obtained by adding a full connection layer to the output layer of the convolutional neural network;
the training method of the classification prediction module comprises the following steps:
carrying out image feature extraction on the enhanced simulation image by adopting the trained convolutional neural network;
and inputting the image characteristics and the corresponding class labels into a full connection layer, and training a classification prediction model.
9. The defect classification system of claim 8, wherein the last fully-connected layer activation function of the classification prediction model is a softmax activation function, and the softmax activation function is:
wherein the Softmax function transforms the vector (a)1，a2，…，an) Mapping as a vector (S)1，S2，…，Sn) Wherein n is the number of categories; a isjRepresents the input value of the jth output node;
is a normalized coefficient; sjShowing the output after Softmax calculation.
10. The defect classification system of claim 8 or 9, wherein the method for preprocessing the image of the sample to be tested comprises:
interactively cutting the image of the sample piece to be detected, and reserving a defect area to be classified;
filling the upper edge, the lower edge, the left edge and the right edge of the image with the defect area in the height and width directions respectively by three RGB channels; wherein the filled color is consistent with the image color of the non-defective area;
merging the three channel images to ensure that the size of the filled image is consistent with the size of an input image of the convolutional neural network;
the method for preprocessing the simulation image comprises the following steps:
cutting an edge image irrelevant to defect classification according to the defect distribution position of the simulation image, and reserving a defect area;
dividing the cut simulation image into different folders according to categories to finish the data set manufacturing; and performing label setting on the data set by adopting thermal coding.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111593864.5A CN114359193B (en) | 2021-12-23 | 2021-12-23 | Defect classification method and system based on ultrasonic phased array imaging |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111593864.5A CN114359193B (en) | 2021-12-23 | 2021-12-23 | Defect classification method and system based on ultrasonic phased array imaging |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114359193A true CN114359193A (en) | 2022-04-15 |
| CN114359193B CN114359193B (en) | 2023-06-30 |
Family
ID=81100810
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111593864.5A Active CN114359193B (en) | 2021-12-23 | 2021-12-23 | Defect classification method and system based on ultrasonic phased array imaging |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114359193B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114783011A (en) * | 2022-06-22 | 2022-07-22 | 广东惠丰达电气设备有限公司 | Ultrasonic imaging identification method for internal defects of GIS |
| CN115587291A (en) * | 2022-09-26 | 2023-01-10 | 华中科技大学 | Denoising characterization method and system based on crack ultrasonic scattering matrix |
| CN115774014A (en) * | 2022-08-12 | 2023-03-10 | 四川大学 | Welding seam defect detection system and method based on vision and ultrasound |
| CN115830012A (en) * | 2023-02-08 | 2023-03-21 | 诺比侃人工智能科技(成都)股份有限公司 | Method for detecting and analyzing contact net clue damage data |
| CN116388884A (en) * | 2023-06-05 | 2023-07-04 | 浙江大学 | Method, system and device for designing anti-eavesdrop ultrasonic interference sample |
| CN116908314A (en) * | 2023-09-08 | 2023-10-20 | 中国电力科学研究院有限公司 | Ultrasonic detection method and system for lead sealing defect of cable accessory |
| CN117974910A (en) * | 2024-03-29 | 2024-05-03 | 南京航空航天大学苏州研究院 | Image simulation, pseudo-random defect data set generation and micro-nano defect detection method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107451997A (en) * | 2017-07-31 | 2017-12-08 | 南昌航空大学 | A kind of automatic identifying method of the welding line ultrasonic TOFD D scanning defect types based on deep learning |
| CN111007151A (en) * | 2019-12-30 | 2020-04-14 | 华东理工大学 | Ultrasonic phased array rapid full-focusing imaging detection method based on defect pre-positioning |
| CN111060601A (en) * | 2019-12-27 | 2020-04-24 | 武汉武船计量试验有限公司 | Weld ultrasonic phased array detection data intelligent analysis method based on deep learning |
| US20210334587A1 (en) * | 2018-09-04 | 2021-10-28 | Boe Technology Group Co., Ltd. | Method and apparatus for training a convolutional neural network to detect defects |
-
2021
- 2021-12-23 CN CN202111593864.5A patent/CN114359193B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107451997A (en) * | 2017-07-31 | 2017-12-08 | 南昌航空大学 | A kind of automatic identifying method of the welding line ultrasonic TOFD D scanning defect types based on deep learning |
| US20210334587A1 (en) * | 2018-09-04 | 2021-10-28 | Boe Technology Group Co., Ltd. | Method and apparatus for training a convolutional neural network to detect defects |
| CN111060601A (en) * | 2019-12-27 | 2020-04-24 | 武汉武船计量试验有限公司 | Weld ultrasonic phased array detection data intelligent analysis method based on deep learning |
| CN111007151A (en) * | 2019-12-30 | 2020-04-14 | 华东理工大学 | Ultrasonic phased array rapid full-focusing imaging detection method based on defect pre-positioning |
Non-Patent Citations (1)
| Title |
|---|
| 申庆: "加筋板中基于深度学习的超声导波损伤识别评估研究" * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114783011A (en) * | 2022-06-22 | 2022-07-22 | 广东惠丰达电气设备有限公司 | Ultrasonic imaging identification method for internal defects of GIS |
| CN114783011B (en) * | 2022-06-22 | 2022-09-06 | 广东惠丰达电气设备有限公司 | Ultrasonic imaging identification method for internal defects of GIS |
| CN115774014A (en) * | 2022-08-12 | 2023-03-10 | 四川大学 | Welding seam defect detection system and method based on vision and ultrasound |
| CN115587291A (en) * | 2022-09-26 | 2023-01-10 | 华中科技大学 | Denoising characterization method and system based on crack ultrasonic scattering matrix |
| CN115830012A (en) * | 2023-02-08 | 2023-03-21 | 诺比侃人工智能科技(成都)股份有限公司 | Method for detecting and analyzing contact net clue damage data |
| CN116388884A (en) * | 2023-06-05 | 2023-07-04 | 浙江大学 | Method, system and device for designing anti-eavesdrop ultrasonic interference sample |
| CN116388884B (en) * | 2023-06-05 | 2023-10-20 | 浙江大学 | Method, system and device for designing anti-eavesdrop ultrasonic interference sample |
| CN116908314A (en) * | 2023-09-08 | 2023-10-20 | 中国电力科学研究院有限公司 | Ultrasonic detection method and system for lead sealing defect of cable accessory |
| CN117974910A (en) * | 2024-03-29 | 2024-05-03 | 南京航空航天大学苏州研究院 | Image simulation, pseudo-random defect data set generation and micro-nano defect detection method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114359193B (en) | 2023-06-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114359193B (en) | Defect classification method and system based on ultrasonic phased array imaging | |
| EP3382386B1 (en) | Defect detection using ultrasound scan data | |
| Posilović et al. | Flaw detection from ultrasonic images using YOLO and SSD | |
| Siljama et al. | Automated flaw detection in multi-channel phased array ultrasonic data using machine learning | |
| Gantala et al. | Automated defect recognition for welds using simulation assisted TFM imaging with artificial intelligence | |
| CN107219305A (en) | A kind of total focus imaging detection method based on annular array transducer | |
| Cheng et al. | IRT-GAN: A generative adversarial network with a multi-headed fusion strategy for automated defect detection in composites using infrared thermography | |
| CN113252779B (en) | Method for improving defect inversion imaging quality | |
| Bai et al. | Ultrasonic defect characterization using the scattering matrix: A performance comparison study of Bayesian inversion and machine learning schemas | |
| McKnight et al. | A comparison of methods for generating synthetic training data for domain adaption of deep learning models in ultrasonic non-destructive evaluation | |
| Wang et al. | The aircraft skin crack inspection based on different-source sensors and support vector machines | |
| Bevan et al. | Automated detection and characterisation of defects from multiview ultrasonic imaging | |
| Zhao et al. | Automated quantification of small defects in ultrasonic phased array imaging using AWGA-gcForest algorithm | |
| Chen et al. | Ellipse of uncertainty based algorithm for quantitative evaluation of defect localization using Lamb waves | |
| Osman | Automated evaluation of three dimensional ultrasonic datasets | |
| Zhang et al. | Super-resolution reconstruction of ultrasonic Lamb wave TFM image via deep learning | |
| Bordekar et al. | eXplainable artificial intelligence for automatic defect detection in additively manufactured parts using CT scan analysis | |
| Guo et al. | Ultrasonic characterization of small defects based on Res-ViT and unsupervised domain adaptation | |
| Aldrin et al. | The need and approach for characterization-US air force perspectives on materials state awareness | |
| Sutcliffe et al. | Automatic defect recognition of single-v welds using full matrix capture data, computer vision and multi-layer perceptron artificial neural networks | |
| McKnight et al. | GANs and alternative methods of synthetic noise generation for domain adaption of defect classification of Non-destructive ultrasonic testing | |
| CN110779477A (en) | Acoustic method for identifying shape of object in real time | |
| Xiang et al. | A reliability study on automated defect assessment in optical pulsed thermography | |
| Tunukovic et al. | A study of machine learning object detection performance for phased array ultrasonic testing of carbon fibre reinforced plastics | |
| Garton et al. | UTSim: overview and application |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |