CN114926462B - Intelligent detection method and system for metal material surface defects - Google Patents

Intelligent detection method and system for metal material surface defects Download PDF

Info

Publication number
CN114926462B
CN114926462B CN202210848144.7A CN202210848144A CN114926462B CN 114926462 B CN114926462 B CN 114926462B CN 202210848144 A CN202210848144 A CN 202210848144A CN 114926462 B CN114926462 B CN 114926462B
Authority
CN
China
Prior art keywords
defect detection
surface defect
detection result
metal material
historical
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.)
Active
Application number
CN202210848144.7A
Other languages
Chinese (zh)
Other versions
CN114926462A (en
Inventor
李佩
张骁
杨春启
储钱良
沈衡
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Suzhou Xianglou New Material Co ltd
Original Assignee
Suzhou Xianglou New Material Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Suzhou Xianglou New Material Co ltd filed Critical Suzhou Xianglou New Material Co ltd
Priority to CN202210848144.7A priority Critical patent/CN114926462B/en
Publication of CN114926462A publication Critical patent/CN114926462A/en
Application granted granted Critical
Publication of CN114926462B publication Critical patent/CN114926462B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • G06T7/0004Industrial image inspection
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • G06N3/084Backpropagation, e.g. using gradient descent
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00Arrangements for image or video recognition or understanding
    • G06V10/70Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/762Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00Arrangements for image or video recognition or understanding
    • G06V10/70Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20081Training; Learning
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20084Artificial neural networks [ANN]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30108Industrial image inspection
    • G06T2207/30136Metal

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Multimedia (AREA)
  • Quality & Reliability (AREA)
  • Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)

Abstract

The invention provides an intelligent detection method and system for metal material surface defects, and relates to the technical field of metal material detection, wherein the method comprises the following steps: acquiring a surface image set of a metal material to be detected; constructing a surface defect detection model; inputting the surface image set into a surface defect detection model to obtain a first surface defect detection result; transmitting a frequency mixing ultrasonic signal to the metal material to be detected and acquiring an ultrasonic output signal; judging whether a preset signal exists in the ultrasonic output signal or not; if so, extracting to obtain a preset signal, inputting the preset signal into the surface crack analysis model to obtain a second surface defect detection result, and inputting the second surface defect detection result into the surface defect evaluation space in combination with the first surface defect detection result to obtain a metal material surface defect detection result. The invention solves the technical problems of poor detection accuracy of the surface of the metal material and insufficient intelligence of the analysis method in the prior art, and achieves the technical effects of improving the surface detection effect and the intelligent degree.

Description

Intelligent detection method and system for metal material surface defects
Technical Field
The invention relates to the technical field of metal material detection, in particular to an intelligent detection method and system for metal material surface defects.
Background
The surface quality detection of the metal material or the workpiece made of the metal material plays a very important role in controlling the quality of the metal material product. The surface detection of the metal material mainly comprises the detection of the material, appearance, shape and surface defects of the surface.
At present, the surface of the metal material is generally detected by a conventional detection means, such as sampling detection, and whether the relevant standard is met or not is judged to complete detection analysis.
The detection method for detecting the surface of the metal material in the prior art has low intelligent degree and single analysis method, and has the technical problems of low detection efficiency and poor accuracy of the surface of the metal material and insufficient intelligence of the analysis method.
Disclosure of Invention
The application provides an intelligent detection method and system for metal material surface defects, which are used for solving the technical problems that in the prior art, the metal material surface detection efficiency is low, the accuracy is poor, and an analysis method is not intelligent enough.
In view of the above problems, the present application provides an intelligent detection method and system for surface defects of a metal material.
In a first aspect of the present application, there is provided a method for intelligently detecting surface defects of a metal material, the method comprising: acquiring a plurality of surface images of a metal material to be detected to obtain a surface image set; constructing a surface defect detection model according to the size characteristics and the type characteristics of the metal material to be detected; inputting the surface image set into the surface defect detection model to obtain a first surface defect detection result; emitting a mixing ultrasonic signal to the metal material to be detected, and detecting and acquiring an ultrasonic output signal of the mixing ultrasonic signal after passing through the metal material to be detected; judging whether a preset signal of a preset type exists in the ultrasonic output signal, if not, outputting the first surface defect detection result to obtain a metal material surface defect detection result; if yes, extracting to obtain the preset signal, inputting the preset signal into a surface crack analysis model, and obtaining a second surface defect detection result; and inputting the first surface defect detection result and the second surface defect detection result into a surface defect evaluation space to obtain a metal material surface defect detection result.
In a second aspect of the present application, there is provided an intelligent detection system for surface defects of a metal material, the system comprising: the surface image acquisition module is used for acquiring a plurality of surface images of the metal material to be detected to obtain a surface image set; the defect detection model building module is used for building a surface defect detection model according to the size characteristics and the type characteristics of the metal material to be detected; the first surface defect detection module is used for inputting the surface image set into the surface defect detection model to obtain a first surface defect detection result; the ultrasonic detection module is used for transmitting a mixing ultrasonic signal to the metal material to be detected, and detecting and acquiring an ultrasonic output signal of the mixing ultrasonic signal after passing through the metal material to be detected; the preset signal judgment module is used for judging whether preset signals of preset types exist in the ultrasonic output signals or not, and if not, outputting the first surface defect detection result to obtain a metal material surface defect detection result; the second surface defect detection module is used for extracting and obtaining the preset signal if the first surface defect detection module is used for detecting the second surface defect, and inputting the preset signal into the surface crack analysis model to obtain a second surface defect detection result; and the detection result output module is used for inputting the first surface defect detection result and the second surface defect detection result into a surface defect evaluation space to obtain a metal material surface defect detection result.
One or more technical solutions provided in the present application have at least the following technical effects or advantages:
the technical scheme that this application provided acquires a plurality of surface images of waiting to detect metal material through the collection, and construct the surface defect detection model, carry out surface defect detection to a plurality of surface images, accomplish the surface defect detection of first part, and based on mixing ultrasonic detection, with the ultrasonic signal input of different frequencies wait to detect metal material, and gather the ultrasonic output signal after waiting to detect metal material propagation, judge whether there is preset signal wherein, if exist, then input the preset signal that obtains into the surface crack analysis model, accomplish the surface defect detection of the second part of surface crack, finally based on surface defect evaluation space, accomplish the analysis of the detection of surface defect. According to the method, the metal material surface quality defect detection is divided into two parts, namely cracks and other defects, by constructing a specific metal material surface quality defect detection and analysis method, the detection and analysis of the metal material surface defects are carried out by adopting a method in machine learning and combining a method for analyzing the surface cracks by frequency mixing ultrasonic waves, the detection intelligence degree is high, the efficiency is high, the accuracy is high, a method for analyzing and evaluating the metal material surface defect detection results is established, the metal material surface defect detection results can be analyzed and evaluated uniformly, intelligently and accurately according to the detection results, and the technical effects of improving the efficiency, the accuracy and the intelligence degree of the metal material surface detection and analysis are achieved.
Drawings
FIG. 1 is a schematic flow chart of an intelligent detection method for surface defects of a metal material according to the present application;
FIG. 2 is a schematic flow chart illustrating a second surface defect detection result obtained by the intelligent detection method for metal material surface defects provided by the present application;
fig. 3 is a schematic flow chart illustrating a detection result of a surface defect of a metal material obtained by the intelligent detection method for a surface defect of a metal material provided by the present application;
fig. 4 is a schematic structural diagram of an intelligent detection system for surface defects of a metal material according to the present application.
Description of reference numerals: the device comprises a surface image acquisition module 11, a defect detection model construction module 12, a first surface defect detection module 13, an ultrasonic detection module 14, a preset signal judgment module 15, a second surface defect detection module 16 and a detection result output module 17.
Detailed Description
The application provides an intelligent detection method and system for metal material surface defects, and aims to solve the technical problems that in the prior art, the metal material surface detection efficiency is low, the accuracy is poor, and an analysis method is not intelligent enough.
Example one
As shown in fig. 1, the present application provides a method for intelligently detecting surface defects of a metal material, the method comprising:
s100: acquiring a plurality of surface images of a metal material to be detected to obtain a surface image set;
in the embodiment of the application, the metal material to be detected can be any element metal material with any mark or a workpiece made of metal material, such as a metal material plate, square steel, round steel and the like.
And acquiring a surface image of the metal material, and using the acquired surface image as a data base for detecting the surface quality of the metal material based on an image processing technology.
Step S100 in the method provided in the embodiment of the present application includes:
s110: setting and obtaining a plurality of acquisition angles for detecting the surface image;
s120: and acquiring a plurality of surface images of the metal material to be detected according to the plurality of acquisition angles to obtain the surface image set.
Specifically, when acquiring and acquiring a surface image of a metal material to be detected, a plurality of angles for acquiring the surface image are set to acquire and acquire an omnidirectional surface image of the metal material to be detected, such as a surface image in a front view direction, a side view direction, and the like. And acquiring surface images of the metal material to be detected at a plurality of angles in the same direction.
The surface image of the metal material to be detected with higher resolution can be acquired by adopting an industrial camera.
Based on a plurality of angles, acquiring and acquiring surface images of the metal material to be detected to obtain a surface image set, wherein the surface image set is used as a data basis for detecting the surface defects of the metal material based on image processing. The surface images of the metal material to be detected are collected based on the set multiple angles, so that the method is more comprehensive and the accuracy of surface defect detection can be improved.
S200: constructing a surface defect detection model according to the size characteristics and the type characteristics of the metal material to be detected;
specifically, based on the size characteristics and the type characteristics of the metal material to be detected, surface images of other metal materials identical to the metal material to be detected are acquired, a model for surface defect detection is constructed, and intelligent image processing surface detection is performed.
The size characteristics comprise information such as size specification of the metal material to be detected, and the type characteristics comprise information such as metal grade and alloy element content of the metal material to be detected.
Step S200 in the method provided in the embodiment of the present application includes:
s210: setting and obtaining a preset time range;
s220: acquiring a surface image of the metal material for surface defect detection within the preset time range in the historical time to obtain a historical surface image set;
s230: traversing the historical surface image set according to the size characteristic and the type characteristic to obtain a historical surface image of the metal material which is the same as the metal material to be detected as a sample surface image set;
s240: and constructing the surface defect detection model according to the sample surface image set.
Specifically, a preset time range for acquiring the surface image of the other metal material is set, the preset time range being a time range having an arbitrary time length, wherein, in order to increase the data amount of acquiring the surface image of the other metal material, a longer time range is preferably set.
And acquiring surface images of other metal materials for surface defect detection within a preset time range in historical time, wherein the surface images can be acquired based on surface defect detection log records. And the surface images of other metal materials are the same as the surface image set acquired based on a plurality of angles, so that a historical surface image set is acquired.
Traversing the historical surface image set according to the size characteristic and the type characteristic of the metal material to be detected, acquiring a historical surface image of the metal material with the same size characteristic and the same type characteristic as those of the metal material to be detected as a sample surface image according to the size information and the type information of the corresponding metal material stored in the historical surface image set, and acquiring a sample surface image set.
And constructing a surface defect detection model for detecting the surface defects of the metal material according to the sample surface image set. Step S240 in the method provided in the embodiment of the present application includes:
s241: marking surface defects of the sample surface images in the sample surface image set to obtain a defect image set;
s242: setting a defect detection result information set according to the defect image set;
s243: dividing and identifying the defect image set and the defect detection result information set to obtain training data, verification data and test data;
s244: constructing the surface defect detection model based on a convolutional neural network;
s245: and carrying out supervision training, verification and testing on the surface defect detection model by adopting the training data, the verification data and the testing data until the accuracy of the surface defect detection model reaches a preset requirement, and obtaining the surface defect detection model.
Specifically, for the collected sample surface image set, based on a metal material surface inspection means in the prior art, such as manual inspection, whether a surface defect exists in the sample surface image is detected, and the sample surface image with the surface defect is selected to obtain a defect image set. Wherein the surface defects may include: including macroscopic metal material surface defects such as wounds, ears, scars, stains, pocks, subcutaneous bubbles, and the like.
And setting a corresponding detection result as construction data for constructing a surface defect detection model according to the sample surface image with the surface defect in the defect image set. Illustratively, for the scratch image in the sample surface image with scratch surface defects, the corresponding detection result is set as the scratch, so that the corresponding detection result is set for other surface defects and the detection result without surface defects, and a defect detection result information set is obtained. The defect detection results in the defect detection result information set correspond to the defect images in the defect image set.
And identifying the defect image set and the defect detection result information set, and dividing according to a preset rule to obtain training data, verification data and test data for constructing a surface defect detection model.
Based on the convolutional neural network model in deep learning, a network structure is designed according to data in the defect image set and the defect detection result information set, and a surface defect detection model is constructed, wherein the surface defect detection model comprises a plurality of convolutional layers, pooling layers and full-connection layers.
The training data, the verification data and the test data are adopted to carry out supervision training, verification and testing on the surface defect detection model, wherein the supervision training is carried out on the surface defect detection model by adopting the supervision data one by one until the output result of the model is converged or the accuracy reaches the preset requirement, the model is verified and tested, the verification and the testing can be carried out based on a cross verification mode, and the problem of overfitting of the model is avoided.
According to the embodiment of the application, the surface images of other metal materials with the same size and type are collected, the defect surface image set is obtained through screening, the corresponding defect detection result set is set, the surface defect detection model is constructed and obtained, the surface defect detection can be carried out accurately according to the input metal material surface image, the detection efficiency is high, the universality is high, the intelligent degree is high, and the accuracy and the efficiency of the metal material surface detection are improved.
S300: inputting the surface image set into the surface defect detection model to obtain a first surface defect detection result;
based on the trained surface defect detection model, inputting a set of surface images of the metal material to be detected into the surface defect detection model, analyzing and detecting the surface images to obtain an output result, wherein the output result comprises the defect detection result of each surface image, and completing intelligent detection of macroscopic defects in the surface of the metal material to be detected to obtain a first surface defect detection result.
S400: emitting a mixing ultrasonic signal to the metal material to be detected, and detecting and acquiring an ultrasonic output signal of the mixing ultrasonic signal after passing through the metal material to be detected;
specifically, after the macroscopic defect detection of the surface of the metal material to be detected is completed and the first surface defect detection result is obtained, cracks which are difficult to distinguish by naked eyes on the surface of the metal material to be detected also need to be detected.
In the embodiment of the application, the frequency mixing ultrasonic detection is adopted to detect the cracks on the surface of the metal material to be detected. Specifically, a mixed-frequency ultrasonic signal, illustratively, two ultrasonic signals of different frequencies, is emitted toward the surface of the metal material to be detected. And acquiring an ultrasonic signal of the frequency mixing ultrasonic signal transmitted through the surface of the metal material to be detected to obtain an ultrasonic output signal. When the frequency mixing ultrasonic detection is carried out, the position of detection on the surface of the metal material to be detected can be selected, and the interference of other defects is avoided.
S500: judging whether a preset signal of a preset type exists in the ultrasonic output signal, if not, outputting the first surface defect detection result to obtain a metal material surface defect detection result;
if nonlinear cracks exist on the surface of the metal material to be detected, the frequency mixing ultrasonic signals are affected, and signals with other frequencies are generated. Therefore, whether cracks exist on the surface of the metal material to be detected is judged by judging whether preset signals of preset types exist in the ultrasonic output signals.
Step S500 in the method provided in the embodiment of the present application includes:
s510: judging whether a sum frequency component exists in the ultrasonic output signal or not to obtain a first judgment result;
s520: judging whether the ultrasonic output signal has a difference frequency component or not to obtain a second judgment result;
s530: if the first judgment result and/or the second judgment result are/is yes, the preset signal exists in the ultrasonic output signal.
Specifically, when the mixed ultrasonic signal is affected by a crack on the surface of the metal material, a sum frequency component or a difference frequency component is generated. The preset signals of the preset type are the sum frequency component and the difference frequency component.
In this way, the ultrasonic output signal is analyzed and judged based on the ultrasonic signal analysis, and a first judgment result can be obtained by judging whether the sum frequency component exists in the ultrasonic output signal. And judging whether the difference frequency component exists in the ultrasonic output signal to obtain a second judgment result.
If the first judgment result and the second judgment result are both negative, a preset signal of a preset type does not exist in the ultrasonic output signal, and no crack exists on the surface of the metal material to be detected. The first surface defect detection result can be output as a metal material surface defect detection result of the metal material to be detected, and the detection is completed.
If the first judgment result and/or the second judgment result are/is yes, it is indicated that the crack exists on the surface of the metal material to be detected, and the crack on the surface of the metal material to be detected needs to be further detected.
According to the embodiment of the application, the surface of the metal material to be detected is detected by adopting frequency mixing ultrasonic detection, and whether cracks exist on the surface of the metal material is judged by judging whether sum frequency components or difference frequency components exist in ultrasonic output signals, so that the method is accurate and efficient, whether crack defects which are difficult to find by naked eyes exist on the surface of the metal material can be effectively judged, and the detection accuracy and efficiency are improved.
S600: if so, extracting to obtain the preset signal, inputting the preset signal into a surface crack analysis model, and obtaining a second surface defect detection result;
if the ultrasonic output signal has a preset signal, i.e., a sum frequency component or a difference frequency component, the sum frequency component and/or the difference frequency component in the preset signal needs to be further used as a data basis for detecting and analyzing the surface crack of the metal material to be detected.
Specifically, a preset signal in the ultrasonic output signal is extracted and input into a surface crack analysis model for analyzing and detecting the surface cracks of the metal material, so that a defect detection result of the surface cracks of the metal material to be detected is obtained.
As shown in fig. 2, step S600 in the method provided in the embodiment of the present application includes:
s610: setting and obtaining a preset time range;
s620: acquiring preset signals and surface crack detection results of other metal materials in the preset time range within historical time to obtain a constructed data set;
s630: constructing the surface crack analysis model based on a BP neural network;
s640: adopting the constructed data set to perform supervision training, verification and testing on the surface crack analysis model to obtain the surface crack analysis model;
s650: and inputting the preset signal into the surface crack analysis model to obtain a second surface defect detection result.
Specifically, a preset time range for acquiring other metal materials to perform frequency mixing ultrasonic detection data is set, and the preset time range can be preferably a range with a longer time span so as to improve the data volume of the acquired data.
Step S620 in the method provided in the embodiment of the present application includes:
s621: acquiring historical ultrasonic output signals for performing ultrasonic frequency mixing detection on other metal materials with the same size characteristics and type characteristics in the preset time range within the historical time to obtain a historical ultrasonic output signal set;
s622: obtaining a historical preset signal set according to the historical ultrasonic output signal set;
s623: acquiring the results of other metal materials with the same size characteristics and type characteristics to obtain a historical surface microcrack detection result set;
s624: dividing and identifying the historical ultrasonic output signal set and the historical surface microcrack detection result set to obtain training data, verification data and test data;
s625: and taking the training data, the verification data and the test data as the construction data set.
Based on the preset time range, acquiring and acquiring ultrasonic output signals of other metal materials in the preset time range in the historical time for frequency mixing ultrasonic detection, and acquiring a historical ultrasonic output signal set. And further acquiring preset signals in the historical ultrasonic output signals, namely sum frequency components and difference frequency components generated by surface cracks in the mixed frequency ultrasonic detection of other metal materials, so as to obtain a historical preset signal set.
And acquiring surface crack detection results of other metal materials based on the preset time range to obtain a historical surface crack detection result set. And the data in the historical surface microcrack detection result set and the historical preset signal set correspond to each other one by one according to the surface of the metal material.
The other metal materials are preferably metal materials with the same size and type as those of the metal materials to be detected, detection parameters of the other metal materials for the frequency mixing ultrasonic detection are the same as those of the frequency mixing ultrasonic detection, and due to the fact that crack positions, lengths, trends, angles and the like on the surfaces of the metal materials are different, preset signals acquired are different. The surface crack detection result can be obtained by detecting the surface of the metal material based on a microscopic technology and used as construction data for constructing a surface crack analysis model.
And dividing and identifying a historical preset signal set and a historical surface microcrack detection result set of other collected metal materials to obtain training data, verification data and test data for constructing a surface crack analysis model, wherein the training data, the verification data and the test data are used as a construction data set.
According to the embodiment of the application, the preset signals of the frequency mixing ultrasonic detection of other metal materials with the same size characteristics and the same type characteristics and the corresponding surface microcrack detection results are acquired and used as construction data for constructing the surface crack analysis model, the model performance of the surface crack analysis model is improved, and the efficiency and the accuracy of the metal material surface crack detection are improved.
After the constructed data set is obtained, based on a BP neural network, the network structure of the surface crack analysis model is designed by combining the waveform, frequency, amplitude and peak characteristic data of the sum frequency component and the difference frequency component in the preset signal and the characteristic data of the trend, length, position, angle and the like of the crack in the surface crack detection result, and the surface crack analysis model is constructed.
And then carrying out supervision training on the surface crack analysis model by adopting training data, verification data and test data in the constructed data set, carrying out supervision training and updating on model parameters based on a gradient descent method until the output result of the surface crack analysis model is converged or reaches a preset accuracy rate, finishing the training of the model, verifying and testing the model, and obtaining the surface crack analysis model if the accuracy rate of the model meets the preset requirement.
And inputting a preset signal of the metal material to be detected into the surface crack analysis model based on the constructed surface crack analysis model, carrying out nonlinear logic analysis budget on the surface crack analysis model according to the waveform, frequency, amplitude and peak characteristic data of the preset signal to obtain an output result, wherein the output result comprises a surface microcrack detection result, and specifically comprises data such as the trend, length, position and angle of the crack on the surface of the metal material to be detected as a second surface defect detection result.
According to the method, the surface crack analysis model is constructed and trained based on the collection of a large amount of frequency mixing ultrasonic detection data and metal material surface crack detection data, the method for intelligently analyzing the metal material surface cracks according to the frequency mixing ultrasonic detection data is constructed, the accuracy is high, and the accuracy and the comprehensiveness of the metal material surface defect detection are effectively improved by combining the detection of other defects in the content.
S700: and inputting the first surface defect detection result and the second surface defect detection result into a surface defect evaluation space to obtain a metal material surface defect detection result.
And based on the first surface defect detection result and the second surface defect detection result obtained by the preliminary detection, acquiring which defects are distributed on the surface of the metal material to be detected, and further evaluating the defect level of the surface of the metal material to be detected based on the detection results to be used as reference data for detecting the defects on the surface of the metal material.
As shown in fig. 3, step S700 in the method provided in the embodiment of the present application includes:
s710: acquiring a historical first surface defect detection result set and a historical second surface defect detection result set for performing surface defect detection on different metal materials within historical time;
s720: constructing a two-dimensional coordinate system according to the historical first surface defect detection result set and the historical second surface defect detection result set;
s730: inputting a historical first surface defect detection result and a historical second surface defect detection result of each metal material into the two-dimensional coordinate system to obtain a plurality of coordinate points;
s740: clustering the coordinate points to obtain a plurality of clustering results;
s750: setting different surface defect evaluation results for each clustering result to obtain the surface defect evaluation space;
s760: inputting the first surface defect detection result and the second surface defect detection result into the surface defect evaluation space to obtain corresponding current clustering results;
s770: and taking the surface defect evaluation result of the current clustering result as the detection result of the surface defect of the metal material.
Specifically, based on the detection method, a plurality of metal materials with the same size and type and different surface defect conditions are detected, a first surface defect detection result and a second surface defect detection result obtained by detection are acquired, and a historical first surface defect detection result set and a historical second surface defect detection result set are obtained.
And constructing a two-dimensional coordinate system according to the historical first surface defect detection result set and the historical second surface defect detection result set, wherein the abscissa of the two-dimensional coordinate system is the first surface defect detection result, and the ordinate is the second surface defect detection result.
And respectively fitting the historical first surface defect detection result and the historical second surface defect detection result of each metal material. Illustratively, fitting is carried out according to the number, the size and the like of various defects of the historical first surface defect detection result to obtain parameters capable of reflecting the historical first surface defect detection result, and fitting is carried out according to the number, the position, the size, the direction, the trend and the like of cracks in the historical second surface defect detection result to obtain parameters capable of reflecting the historical second surface defect detection result. In this way, the historical first surface defect detection result and the historical second surface defect detection result of each metal material are input into the two-dimensional coordinate system, and the coordinate points corresponding to the plurality of metal materials are obtained.
And clustering the plurality of coordinate points, optionally calculating Euclidean distances among the plurality of coordinate points, and clustering every two coordinate points with the Euclidean distances smaller than a threshold value to obtain a plurality of clustering results. The threshold may be set according to the distribution of the plurality of coordinate points and the levels of the first surface defect detection result and the second surface defect detection result.
And (3) supervising and setting different surface defect evaluation results for each clustering result, setting different evaluation results as the evaluation results of the metal material surface defect detection according to the levels of the first surface defect detection result and the second surface defect detection result in each clustering result, for example, the evaluation results of serious defects, few cracks and the like, and obtaining the constructed surface defect evaluation space.
And inputting a first surface defect detection result and a second surface defect detection result of the current metal material to be detected into the surface defect evaluation space to obtain a corresponding coordinate point to be detected and obtain a clustering result corresponding to the coordinate point to be detected.
Optionally, if the to-be-detected coordinate point falls into a certain clustering result, a corresponding clustering result is obtained, and if the to-be-detected coordinate point does not fall into a certain clustering result, the closest clustering result of the to-be-detected coordinate point is taken as the corresponding current clustering result.
And taking the surface defect evaluation result corresponding to the current clustering result as a metal material surface defect detection result, and completing the detection and analysis evaluation of the metal material to be detected.
In summary, the embodiment of the present application has at least the following technical effects:
according to the method, the metal material surface quality defect detection is divided into two parts, namely cracks and other defects, by constructing a specific metal material surface quality defect detection and analysis method, the detection and analysis of the metal material surface defects are carried out by adopting a method in machine learning and combining a method for analyzing the surface cracks by frequency mixing ultrasonic waves, the detection intelligence degree is high, the efficiency is high, the accuracy is high, a method for analyzing and evaluating the metal material surface defect detection results is established, the metal material surface defect detection results can be analyzed and evaluated uniformly, intelligently and accurately according to the detection results, and the technical effects of improving the efficiency, the accuracy and the intelligence degree of the metal material surface detection and analysis are achieved.
Example two
Based on the same inventive concept as the intelligent detection method for the surface defects of the metal material in the previous embodiment, as shown in fig. 4, the present application provides an intelligent detection system for the surface defects of the metal material, wherein the system includes:
the surface image acquisition module 11 is used for acquiring a plurality of surface images of the metal material to be detected to obtain a surface image set;
the defect detection model building module 12 is used for building a surface defect detection model according to the size characteristics and the type characteristics of the metal material to be detected;
a first surface defect detection module 13, configured to input the surface image set into the surface defect detection model, so as to obtain a first surface defect detection result;
the ultrasonic detection module 14 is configured to transmit a mixed frequency ultrasonic signal to the metal material to be detected, and detect and acquire an ultrasonic output signal of the mixed frequency ultrasonic signal after passing through the metal material to be detected;
a preset signal judgment module 15, configured to judge whether a preset signal of a preset type exists in the ultrasonic output signal, and if not, output the first surface defect detection result to obtain a metal material surface defect detection result;
the second surface defect detection module 16 is configured to extract and obtain the preset signal if the first surface defect detection result is positive, and input the preset signal into the surface crack analysis model to obtain a second surface defect detection result;
and the detection result output module 17 is configured to input the first surface defect detection result and the second surface defect detection result into a surface defect evaluation space to obtain a metal material surface defect detection result.
Further, the surface image capturing module 11 is further configured to implement the following functions:
setting and obtaining a plurality of acquisition angles for detecting the surface image;
and acquiring a plurality of surface images of the metal material to be detected according to the plurality of acquisition angles to obtain the surface image set.
Further, the defect detection model building module 12 is further configured to implement the following functions:
setting and obtaining a preset time range;
acquiring a surface image of the metal material for surface defect detection within the preset time range in the historical time to obtain a historical surface image set;
traversing the historical surface image set according to the size characteristic and the type characteristic to obtain a historical surface image of the metal material which is the same as the metal material to be detected as a sample surface image set;
and constructing the surface defect detection model according to the sample surface image set.
Wherein constructing the surface defect detection model according to the sample surface image set comprises:
marking surface defects of the sample surface images in the sample surface image set to obtain a defect image set;
setting a defect detection result information set according to the defect image set;
dividing and identifying the defect image set and the defect detection result information set to obtain training data, verification data and test data;
constructing the surface defect detection model based on a convolutional neural network;
and carrying out supervision training, verification and testing on the surface defect detection model by adopting the training data, the verification data and the testing data until the accuracy of the surface defect detection model reaches a preset requirement, and obtaining the surface defect detection model.
Further, the preset signal determining module 15 is further configured to implement the following functions:
judging whether a sum frequency component exists in the ultrasonic output signal or not to obtain a first judgment result;
judging whether a difference frequency component exists in the ultrasonic output signal or not to obtain a second judgment result;
if the first judgment result and/or the second judgment result are/is yes, the preset signal exists in the ultrasonic output signal.
Further, the second surface defect detecting module 16 is further configured to implement the following functions:
setting and obtaining a preset time range;
acquiring preset signals and surface crack detection results of other metal materials in the preset time range within historical time to obtain a constructed data set;
constructing the surface crack analysis model based on a BP neural network;
adopting the constructed data set to perform supervision training, verification and testing on the surface crack analysis model to obtain the surface crack analysis model;
and inputting the preset signal into the surface crack analysis model to obtain a second surface defect detection result.
Acquiring preset signals and surface crack detection results of other metal materials in the preset time range in historical time, wherein the acquiring comprises the following steps:
acquiring historical ultrasonic output signals for performing ultrasonic frequency mixing detection on other metal materials with the same size characteristics and type characteristics in the preset time range within the historical time to obtain a historical ultrasonic output signal set;
obtaining a historical preset signal set according to the historical ultrasonic output signal set;
acquiring the results of other metal materials with the same size characteristics and type characteristics to obtain a historical surface microcrack detection result set;
dividing and identifying the historical ultrasonic output signal set and the historical surface microcrack detection result set to obtain training data, verification data and test data;
and taking the training data, the verification data and the test data as the construction data set.
Further, the detection result output module 17 is further configured to implement the following functions:
acquiring a historical first surface defect detection result set and a historical second surface defect detection result set for performing surface defect detection on different metal materials within historical time;
constructing a two-dimensional coordinate system according to the historical first surface defect detection result set and the historical second surface defect detection result set;
inputting a historical first surface defect detection result and a historical second surface defect detection result of each metal material into the two-dimensional coordinate system to obtain a plurality of coordinate points;
clustering the coordinate points to obtain a plurality of clustering results;
setting different surface defect evaluation results for each clustering result to obtain the surface defect evaluation space;
inputting the first surface defect detection result and the second surface defect detection result into the surface defect evaluation space to obtain corresponding current clustering results;
and taking the surface defect evaluation result of the current clustering result as the detection result of the surface defect of the metal material.
The specification and figures are merely exemplary of the application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the application. It will be apparent to those skilled in the art that various changes and modifications may be made in the present application without departing from the scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the present application and its equivalent technology, it is intended that the present application include such modifications and variations.

Claims (8)

1. An intelligent detection method for surface defects of a metal material, which is characterized by comprising the following steps:
acquiring a plurality of surface images of a metal material to be detected to obtain a surface image set;
constructing a surface defect detection model according to the size characteristics and the type characteristics of the metal material to be detected;
inputting the surface image set into the surface defect detection model to obtain a first surface defect detection result;
emitting a mixing ultrasonic signal to the metal material to be detected, and detecting and acquiring an ultrasonic output signal of the mixing ultrasonic signal after passing through the metal material to be detected;
judging whether preset signals of preset types exist in the ultrasonic output signals or not; if not, outputting the first surface defect detection result to obtain a metal material surface defect detection result;
if so, extracting to obtain the preset signal, inputting the preset signal into a surface crack analysis model, and obtaining a second surface defect detection result;
inputting the first surface defect detection result and the second surface defect detection result into a surface defect evaluation space to obtain a metal material surface defect detection result, wherein the method comprises the following steps: acquiring a historical first surface defect detection result set and a historical second surface defect detection result set for performing surface defect detection on different metal materials within historical time; constructing a two-dimensional coordinate system according to the historical first surface defect detection result set and the historical second surface defect detection result set; inputting a historical first surface defect detection result and a historical second surface defect detection result of each metal material into the two-dimensional coordinate system to obtain a plurality of coordinate points; clustering the coordinate points to obtain a plurality of clustering results; setting different surface defect evaluation results for each clustering result to obtain the surface defect evaluation space; inputting the first surface defect detection result and the second surface defect detection result into the surface defect evaluation space to obtain corresponding current clustering results; and taking the surface defect evaluation result of the current clustering result as the detection result of the surface defects of the metal material.
2. The method of claim 1, wherein said acquiring a plurality of surface images of the metallic material to be detected comprises:
setting and obtaining a plurality of acquisition angles for detecting the surface image;
and acquiring a plurality of surface images of the metal material to be detected according to the plurality of acquisition angles to obtain the surface image set.
3. The method according to claim 1, wherein said building a surface defect detection model based on the dimensional characteristics and the type characteristics of said metallic material to be detected comprises:
setting and obtaining a preset time range;
acquiring a surface image of the metal material for surface defect detection within the preset time range in the historical time to obtain a historical surface image set;
traversing the historical surface image set according to the size characteristic and the type characteristic to obtain a historical surface image of the metal material which is the same as the metal material to be detected as a sample surface image set;
and constructing the surface defect detection model according to the sample surface image set.
4. The method of claim 3, wherein constructing the surface defect detection model from the set of sample surface images comprises:
marking surface defects of the sample surface images in the sample surface image set to obtain a defect image set;
setting a defect detection result information set according to the defect image set;
dividing and identifying the defect image set and the defect detection result information set to obtain training data, verification data and test data;
constructing the surface defect detection model based on a convolutional neural network;
and performing supervision training, verification and testing on the surface defect detection model by adopting the training data, the verification data and the testing data until the accuracy of the surface defect detection model reaches a preset requirement, and obtaining the surface defect detection model.
5. The method of claim 1, wherein determining whether a preset signal of a preset type is present in the ultrasound output signal comprises:
judging whether a sum frequency component exists in the ultrasonic output signal or not to obtain a first judgment result;
judging whether a difference frequency component exists in the ultrasonic output signal or not to obtain a second judgment result;
if the first judgment result and/or the second judgment result are/is yes, the preset signal exists in the ultrasonic output signal.
6. The method of claim 1, wherein inputting the predetermined signal into a surface crack analysis model to obtain a second surface defect detection result comprises:
setting and obtaining a preset time range;
acquiring preset signals and surface crack detection results of other metal materials in the preset time range within historical time to obtain a constructed data set;
constructing the surface crack analysis model based on a BP neural network;
adopting the constructed data set to perform supervision training, verification and testing on the surface crack analysis model to obtain the surface crack analysis model;
and inputting the preset signal into the surface crack analysis model to obtain a second surface defect detection result.
7. The method according to claim 6, wherein the acquiring of the preset signals and the surface crack detection results of other metal materials in the preset time range in the historical time comprises:
acquiring historical ultrasonic output signals for performing ultrasonic frequency mixing detection on other metal materials with the same size characteristics and type characteristics in the preset time range within the historical time to obtain a historical ultrasonic output signal set;
obtaining a historical preset signal set according to the historical ultrasonic output signal set;
acquiring surface crack detection results of other metal materials with the same size characteristics and type characteristics to obtain a historical surface microcrack detection result set;
dividing and identifying the historical ultrasonic output signal set and the historical surface microcrack detection result set to obtain training data, verification data and test data;
and taking the training data, the verification data and the test data as the construction data set.
8. An intelligent detection system for surface defects of a metal material, the system comprising:
the surface image acquisition module is used for acquiring a plurality of surface images of the metal material to be detected to obtain a surface image set;
the defect detection model building module is used for building a surface defect detection model according to the size characteristics and the type characteristics of the metal material to be detected;
the first surface defect detection module is used for inputting the surface image set into the surface defect detection model to obtain a first surface defect detection result;
the ultrasonic detection module is used for transmitting a mixing ultrasonic signal to the metal material to be detected, and detecting and acquiring an ultrasonic output signal of the mixing ultrasonic signal after passing through the metal material to be detected;
the preset signal judgment module is used for judging whether preset signals of preset types exist in the ultrasonic output signals or not, and if not, outputting the first surface defect detection result to obtain a metal material surface defect detection result;
the second surface defect detection module is used for extracting and obtaining the preset signal if the first surface defect detection module is used for detecting the second surface defect, and inputting the preset signal into the surface crack analysis model to obtain a second surface defect detection result;
the detection result output module is used for inputting the first surface defect detection result and the second surface defect detection result into a surface defect evaluation space to obtain a metal material surface defect detection result and acquiring a historical first surface defect detection result set and a historical second surface defect detection result set for performing surface defect detection on different metal materials within historical time; constructing a two-dimensional coordinate system according to the historical first surface defect detection result set and the historical second surface defect detection result set; inputting a historical first surface defect detection result and a historical second surface defect detection result of each metal material into the two-dimensional coordinate system to obtain a plurality of coordinate points; clustering the coordinate points to obtain a plurality of clustering results; setting different surface defect evaluation results for each clustering result to obtain the surface defect evaluation space; inputting the first surface defect detection result and the second surface defect detection result into the surface defect evaluation space to obtain corresponding current clustering results; and taking the surface defect evaluation result of the current clustering result as the detection result of the surface defect of the metal material.
CN202210848144.7A 2022-07-19 2022-07-19 Intelligent detection method and system for metal material surface defects Active CN114926462B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210848144.7A CN114926462B (en) 2022-07-19 2022-07-19 Intelligent detection method and system for metal material surface defects

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210848144.7A CN114926462B (en) 2022-07-19 2022-07-19 Intelligent detection method and system for metal material surface defects

Publications (2)

Publication Number Publication Date
CN114926462A CN114926462A (en) 2022-08-19
CN114926462B true CN114926462B (en) 2022-11-08

Family

ID=82815944

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210848144.7A Active CN114926462B (en) 2022-07-19 2022-07-19 Intelligent detection method and system for metal material surface defects

Country Status (1)

Country Link
CN (1) CN114926462B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115424095B (en) * 2022-11-03 2023-04-07 湖北信通通信有限公司 Quality analysis method and device based on waste materials
CN115933574B (en) * 2023-01-09 2024-04-26 浙江聚优非织造材料科技有限公司 Intelligent control method and system for producing non-woven product
CN115809501B (en) * 2023-02-08 2023-05-23 万得福实业集团有限公司 Intelligent generation method and system for metal plate curtain wall processing and installation scheme
CN115830030B (en) * 2023-02-22 2023-05-05 日照皓诚电子科技有限公司 Appearance quality assessment method and system for quartz wafer
CN116674300B (en) * 2023-05-24 2023-11-14 常州润来科技有限公司 Automatic ink-jet marking system and method based on copper pipe flaw detection result

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105424798A (en) * 2015-11-07 2016-03-23 哈尔滨理工大学 Method for actively detecting defects in metal thin-walled structure part
CN111325738B (en) * 2020-02-28 2023-06-23 湖北工业大学 Intelligent detection method and system for transverse hole peripheral cracks
CN113887667A (en) * 2021-10-28 2022-01-04 中国人民解放军空军工程大学 Composite material defect detection method and system based on infrared and ultrasonic signal fusion

Also Published As

Publication number Publication date
CN114926462A (en) 2022-08-19

Similar Documents

Publication Publication Date Title
CN114926462B (en) Intelligent detection method and system for metal material surface defects
CN109782274B (en) Water damage identification method based on time-frequency statistical characteristics of ground penetrating radar signals
CN101975818B (en) Detection system and method of characteristic substance
CN110082429B (en) Tunnel lining nondestructive testing auxiliary judgment method combining machine learning
CN113758932B (en) Deep learning-based visual detection method for defects of lithium battery diaphragm
CN102539532B (en) Ultrasonic C scanning imaging method based on two-dimensional neighborhood synthetic aperture focusing
CN112816556B (en) Defect detection method, device, equipment and storage medium
CN111325738B (en) Intelligent detection method and system for transverse hole peripheral cracks
CN102288606A (en) Pollen viability measuring method based on machine vision
CN110568082A (en) cable wire breakage distinguishing method based on acoustic emission signals
CN113237951A (en) Metal plate fatigue damage ultrasonic guided wave detection method based on shape context dynamic time warping
CN109596639A (en) Defect detecting system and defect inspection method
CN102621221A (en) Defect classification method based on phased ultrasonic wave
CN117074430B (en) Method and device for detecting surface defects of stainless steel pipe
CN117109481A (en) Method for detecting and evaluating overall flatness and overcurrent capacity of concrete with overcurrent surface
CN117871679A (en) High-sensitivity ultrasonic defect detection method and device
CN117934404A (en) Stone surface defect detection method and system
CN117169477A (en) Building indoor ground hollowing degree assessment method
CN113899785A (en) Detection method for lithium battery based on ultrasonic and infrared flaw detection
CN114881938A (en) Grain size detection method and system based on wavelet analysis and neural network
CN113624759A (en) Apple nondestructive testing method based on machine learning
CN118362635B (en) Slurry quality real-time detection and analysis method based on ultrasonic image
CN110806443A (en) Steel plate crack evaluation method based on self-organizing statistical model
CN118277842B (en) Substrate classification method and system based on deep sea multi-beam water body bottom echo information
CN113312400B (en) Float glass grade judging method and device

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