阅读说明：本技术 一种复合磁场磁光成像无损检测系统及方法 (A kind of resultant field magneto-optic imaging non-destructive detection system and method ) 是由 高向东 季玉坤 马女杰 张艳喜 游德勇 于 2019-08-12 设计创作，主要内容包括：本发明涉及一种复合磁场磁光成像无损检测系统及方法,所述系统包括：二维运动平台、恒定磁场源、交变磁场源、磁光传感器、交直流电源、处理器；恒定磁场源设于二维运动平台上方,交变磁场源设于恒定磁场源的外侧,交变磁场源于交直流电源连接；磁光传感器设于恒定磁场源的磁场范围内；磁光传感器和处理器连接。所述方法包括以下步骤：S1：对待检测工件进行励磁；S2：对待检测工件的磁光图像进行采集；S3：利用卷积神经网络(CNN),对磁光图像进行缺陷检测；S4：将检查结果可视化。本发明复合磁场磁光成像无损检测技术是将焊接缺陷直接转化为图像进行处理,无损检测的识别精度高,可视化程度高。(The present invention relates to a kind of resultant field magneto-optic imaging non-destructive detection system and method, the system comprises: two-dimension moving platform, stationary magnetic field source, alternating magnetic field source, magneto-optical sensor, power of alterating and direct current, processor；Stationary magnetic field source is set to above two-dimension moving platform, and alternating magnetic field source is set to the outside in stationary magnetic field source, and alternating magnetic field is connected derived from power of alterating and direct current；Magneto-optical sensor is set in the magnetic field range in stationary magnetic field source；Magneto-optical sensor is connected with processor.It the described method comprises the following steps: excitation S1: being carried out to workpiece to be detected；S2: the magneto optic images of workpiece to be detected are acquired；S3: utilizing convolutional neural networks (CNN), carries out defects detection to the magneto optic images；S4: inspection result is visualized.Resultant field magneto-optic imaging non-destructive detection technique of the present invention is that welding defect is converted into image to handle, and the accuracy of identification of non-destructive testing is high, and visualization is high.)

1. a kind of resultant field magneto-optic imaging non-destructive detection system, which is characterized in that the system comprises: two-dimension moving platform (1), stationary magnetic field source (2), alternating magnetic field source (3), magneto-optical sensor (4), power of alterating and direct current (5), processor (6)；

Stationary magnetic field source (2) is set to the top of two-dimension moving platform (1), between stationary magnetic field source (2) and two-dimension moving platform (1) Workpiece to be detected (7) can be placed；

Two-dimension moving platform (1) is used to place workpiece to be detected (7), and workpiece to be detected (7) can be driven to carry out two along the horizontal plane Maintenance and operation is dynamic；

Alternating magnetic field source is set to the outside of stationary magnetic field source (2), and alternating magnetic field source (3) are connect with power of alterating and direct current (5)；

Magneto-optical sensor (4) is set in the magnetic field range of stationary magnetic field source (2)；Magneto-optical sensor (4) and processor (6) connection；

Magnetic Field is become the magneto optic images and passes to processor (6) by magneto-optical sensor (4)；Processor (6) to the magneto optic images into Row processing, realizes non-destructive testing.

2. resultant field magneto-optic imaging non-destructive detection system according to claim 1, which is characterized in that the constant magnetic The structure of field source (2) is "u"-shaped；The U-shaped mechanism in stationary magnetic field source (2) it is opening down, magneto-optical sensor (4) be set to perseverance In the opening at the U-shaped opening end in fixed-field source (2).

3. resultant field magneto-optic imaging non-destructive detection system according to claim 2, which is characterized in that the alternation magnetic Field source (3) is "u"-shaped, is set on the outside of stationary magnetic field source (2) U-shaped structure, so that the U-shaped structure of stationary magnetic field source (2) It is Chong Die with the U-shaped structure in alternating magnetic field source, and the two is in same perpendicular.

4. resultant field magneto-optic imaging non-destructive detection system according to claim 3, which is characterized in that alternating magnetic field source (3) it is formed by silicon steel sheet and copper wire winding；The both ends of copper wire are connect with the positive and negative anodes of power of alterating and direct current (5) respectively.

5. a kind of detection side applied to the described in any item resultant field magneto-optic imaging non-destructive detection systems of claim 1-4 Method, which is characterized in that the described method comprises the following steps:

S1: excitation is carried out to workpiece to be detected using Composite Field, resultant field is stationary magnetic field and alternating magnetic field common activation shape At；

S2: by magneto-optical sensor, the magneto optic images of workpiece to be detected are acquired, and the magneto optic images information is transmitted to place Manage device；

S3: processor carries out defects detection to the magneto optic images with convolutional neural networks；

S4: inspection result is visualized, and passes to monitoring personnel.

6. detection method according to claim 5, which is characterized in that S3 the following steps are included:

S3.1: sample calibration；

S3.2: establish prediction model: prediction model uses 6 convolutional layers, and the magneto optic images after pretreatment will turn to 19 × 19 × 50 matrix；

If input image resolution is n*m, 601*601*3 is converted input images into as CNN input layer, is had in CCN model Following parameter:

f^[l]: the size of convolution kernel；

s^[l]: each step is across several lattice；

The quantity of convolution kernel；

Wherein [l] represents l layers；

L layers of dimension are as follows:

Wherein, in this modelFor this layer height,For this slice width degree,N can be used^[l]It indicates；

L layers of convolution kernel can indicate are as follows:

It wherein 19 × 19 indicates to divide the image into 19 × 19 target frames, 50=5 × (5+5), wherein first 5 indicates each mesh Mark frame can detecte five objects, and second 5 refers to and can detecte 5 kinds of defects, and each defect is indicated by 5 kinds of information, and five kinds Information is respectively as follows: probability of occurrence, defect center abscissa, defect center ordinate, defect width, defect length；

S3.3: initialization weight initializes every layer of convolution kernel；

S3.4: being trained model using the sample demarcated, and obtains the model of high-accuracy detection test specimen defect；

S3.5: real-time defects detection is realized to the magneto optic images that transmission is come in using trained model.

Technical field

The present invention relates to lossless defects detection fields, detect more particularly, to a kind of resultant field magneto-optic imaging non-destructive System and method.

Background technique

All can largely use metal casting in fields such as automobile manufacture, space flight and aviation, machine toolings, metal product and we Life it is closely bound up.Due to processing technology be difficult to control, operating condition is severe, quality requirement is high and various random disturbances because The defects of influence of element, casting can inevitably result from sand holes, pin hole, slag inclusion.In welding procedure, due to welding surroundings It is scarce to be easy to appear crackle, stomata, solid inclusion, incomplete fusion, lack of penetration etc. in welding process for the factors such as severe and machine fatigue It falls into.In order to guarantee its product quality, it is necessary to detect defect in timely and efficient manner.In the actual production process, in addition to estimating table Outside planar defect and forming defect, usually also need to detect defect using non-destructive testing technology, thus a kind of effective defect without Damage detection method has important practical significance.

Following several method: (1) magnetic powder detection method is concentrated mainly on for the non-destructive testing of defect both at home and abroad at present, it should Method is only limitted to magnetic conductive material, the work that is cleaned multiple times and demagnetized to measured workpiece surface is needed before detection, it is ensured that surface Smoothly, magnetic line of force distribution is not influenced.Then uniformly it is covered with magnetic powder on the surface, measured workpiece defect can generate irregularly after magnetization The magnetic line of force, these defects will be shown by the distribution of magnetic powder.Magnetic testing is generally used for the surface of ferromagnetic workpiece And the detection of near surface flaw, cost is relatively low, and sensitivity is higher and to the requirement of measured workpiece amorphism.Magnetic testing method work Sequence is more complicated, is not suitable for on-line checking or thicker workpiece.(2) penetrant detecting method, principle are based on liquid Capillarity is the lossless detection method for detecting workpiece surface flaws, specifically includes two methods of fluorescence and coloring.It is glimmering The principle of light detection is to immerse measured workpiece in fluorescent liquid, and because of capillarity, fluorescent liquid has been filled inside defect.It removes Surface liquid issues visible light and shows defect since photoelectric effect fluorescent liquid is under ultraviolet irradiation.The original of coloration detection Reason is similar with fluorescence detection, it does not need special equipment, and workpiece only is sucked out in the coloring liquid being adsorbed in defect with toner Surface and show defect.This method sensitivity when detecting workpiece surface open-delta winding is high, to surface moisture or there is painting The sample of layer, meeting extreme influence detection effect, and also the judgement of this method is heavily dependent on the experience of inspector.(3) Ray detection method is that have certain attenuation law during passing through testee using ray (x-ray, gamma-rays etc.), According to a kind of method for detecting inside workpiece defect by the transmitted intensity after each position decaying of workpiece.Its decaying of different objects Degree is different, the degree of decaying by the thickness of object, the material category of object and ray type and determine.Ray detection master It is used to detect inside workpiece volume flaw, and the thickness of workpiece is not easy more than 80mm, can be done according to the attenuation coefficient of material It is corresponding to thicken or be thinned.This method testing cost is high, and detection device is larger, and the ray radiation of generation is to human injury pole Greatly, lower to the detection sensitivity of micro-crack defect.(4) ultrasonic detection method, principle are using ultrasonic wave in tested work When being propagated in part, it will receive the influence of measured workpiece material acoustics characteristic and the variation of its interior tissue, pass through the influence of ultrasonic wave Degree and status analysis, to detect the variation of material property and structure.The detection efficiency of the detection method is higher, and at This is lower, but relatively other detection methods, to the more demanding of operator.This method has for distinguishing different types of defect Certain difficulty, maximum disadvantage is exactly to need couplant when detecting.(5) eddy current detection method, testing principle are to be based on Electromagnetic induction phenomenon, changing magnetic field generate vortex in conductor workpiece, if in workpiece existing defects, be mingled with, conductivity When variation or structure change, the flowing of vortex will affect, so that superposition magnetic field changes, may determine that according to the variation in magnetic field The defect of workpiece.This method has many advantages, such as that detection efficiency is high, is suitable for on-line checking, is not necessarily to couplant and non-contact detecting, And it is higher to the sensitivity of near surface or surface defect.But it is only suitable for the detection of conductive material surface and near surface, it is difficult To judge type, the shapes and sizes of defect.(6) stationary magnetic field and alternating magnetic field magneto-optic imaging non-destructive detection method.Both sides Method is to realize stationary magnetic field and low-frequency alternating magnetic with permanent magnet and low frequency ac respectively using Faraday magnetooptical effect Field, workpiece is in induced magnetic field, if workpiece existing defects, deformity, which can occur, for the magnetic line of force shows, and the magnetic line of force of this distortion will The magnetic field of distortion is generated, and the perpendicular field profile at this is caused to change, magneto-optical sensor can convert the vertical magnetic field For the magneto optic images, analyzes workpiece, defect and be converted into analysis the magneto optic images.(7) other detection methods.Such as laser hologram non-destructive testing, It is by the method for external world's load, it to be made to cause local change in corresponding body surface body surface and internal defect Shape, is observed and more this deformation with holographic, and records the deformation of the body surface under different extraneous loads effects Situation is observed and is analyzed, and then whether there is defect inside judgment object.Acoustic emission testing technology is object in the external world Under conditioning, defect or object abnormal position stress are concentrated and are deformed or are broken, and are released in the form of elastic wave Strain energy, with instrument detection and analysis acoustic emission signal and the technology of determining acoustic emission source.Infrared detection technology, when detecting One constant hot-fluid can be injected into workpiece, if existing defects in workpiece, due to the thermal diffusion system of defect area and nondefective zone Number is different, then the Temperature Distribution in workpiece surface will be variant, surface temperature corresponding to internal defective and nondefective zone Degree is just different, and the infrared waves (heat radiation) thus issued are also just different, can be with response infrared wave using infrared detector And it is converted into the function of correspondingly sized electric signal, point by point scanning workpiece surface is obtained with the distribution shape of workpiece surface temperature Condition determines the position of inside workpiece defect to find workpiece surface temperature abnormal area.

The shortcomings that above-mentioned detection technique: Magnetic testing is limited to ferrimagnet, and has strict demand to workpiece surface；Infiltration Detection is limited to Surface-breaking defect；The testing cost of ray detection is high, and detection device is larger, and the ray radiation of generation hurts human body Evil is very big；To the more demanding of operator, distinguishing different types of defect has certain difficulty and needs coupling for ultrasound examination Mixture；EDDY CURRENT is only suitable for the detection of conductive material surface and near surface, and be difficult to judge the type of workpiece, defect, shape and Size；Can the defect that laser hologram non-destructive testing depends on interior of articles cause the mutually strain of body surface under external force Shape；Acoustic emission testing technology needs to detect by sensitive electronic instrument since the intensity of acoustic emission signal is generally very weak； The magneto optic images are easily saturated under existing fixed-field, can not detect deeper defect, and existing alternating magnetic field magneto-optic imaging non-destructive detection method It is difficult to accurately detect inside workpiece tiny flaw；Infrared detection technology mainly measures workpiece surface Warm status, cannot be true The Warm status for determining inside workpiece, it is expensive compared with other detecting instruments or routine monitoring equipment.

Summary of the invention

The present invention is the defect for overcoming the accuracy of identification of non-destructive testing described in the above-mentioned prior art inadequate, is provided a kind of multiple Close magnetic field magneto-optic imaging non-destructive detection system and method.

The present invention is that new lossless detection method is proposed based on the prior art, be easily saturated according to the magneto optic images under fixed-field with And tiny flaw difficulty test problems under alternating magnetic field, propose magneto-optic imaging non-destructive detection technique under resultant field, it is existing to solve Technical deficiency improves non-destructive testing range and accuracy of identification.Resultant field magneto-optic imaging non-destructive detection technique is by welding defect It is converted into image to be handled, the accuracy of identification of non-destructive testing is high, and visualization is high, and image processing method is relatively opened It puts, application prospect is wide.

The system comprises: two-dimension moving platform, stationary magnetic field source, alternating magnetic field source, magneto-optical sensor, alterating and direct current Source, processor；

Stationary magnetic field source is set to above two-dimension moving platform, can be placed between stationary magnetic field source and two-dimension moving platform to be checked The workpiece of survey；

Two-dimension moving platform is used to place workpiece to be detected, and workpiece to be detected can be driven to carry out two maintenance and operations along the horizontal plane It is dynamic；The two-dimension moving platform is the platform for placing workpiece, and can be realized platform and carry out two dimensional motion in the horizontal plane, Such as: cross slid platform etc..

Alternating magnetic field source is set to the outside in stationary magnetic field source, and alternating magnetic field is connected derived from power of alterating and direct current；

Magneto-optical sensor is set in the magnetic field range in stationary magnetic field source；Magneto-optical sensor is connected with processor.

Magnetic Field is become the magneto optic images and passes to processor by magneto-optical sensor；Processor to the magneto optic images at Reason realizes non-destructive testing.

Preferably, the structure in the stationary magnetic field source is "u"-shaped；The opening court of the U-shaped mechanism in stationary magnetic field source Under, magneto-optical sensor is set in the opening at the U-shaped opening end in stationary magnetic field source.

Preferably, the alternating magnetic field source is "u"-shaped, is set on the outside of the U-shaped structure of stationary magnetic field source, so that permanent The U-shaped structure in fixed-field source is Chong Die with the U-shaped structure in alternating magnetic field source, and the two is in same perpendicular；Stationary magnetic field The outside of the U-shaped structure in source close to alternating magnetic field source U-shaped structure inside, U-shaped structure is using U-shaped mouth inside as inside, U-shaped knot Structure back is outside.

Preferably, alternating magnetic field source is formed by silicon steel sheet and copper wire winding；The both ends of copper wire respectively with power of alterating and direct current Positive and negative anodes connection.

The method of the invention the following steps are included:

S1: excitation is carried out to workpiece to be detected using Composite Field, resultant field is that stationary magnetic field and alternating magnetic field swash jointly It encourages and to be formed；

S2: by magneto-optical sensor, the magneto optic images of workpiece to be detected are acquired, and the magneto optic images information is transmitted To processor；

S3: processor carries out defects detection to the magneto optic images with convolutional neural networks (CNN)；

S4: inspection result is visualized, and passes to monitoring personnel.

Preferably, S3 the following steps are included:

S3.1: sample calibration；

S3.2: establish prediction model: prediction model uses 6 convolutional layers, and the magneto optic images after pretreatment will turn to 19 × 19 × 50 matrix；

Since magneto-optical sensor acquisition image resolution ratio is not fixed, if input image resolution is n*m, by input figure As being converted into 601*601*3 as CNN input layer, there is following parameter in CCN model:

f^[l]: the size of convolution kernel；

s^[l]: each step is across several lattice；

The quantity of convolution kernel；

Wherein [l] represents l layers；

L layers of dimension are as follows:

Wherein, in this modelFor this layer height,For this slice width degree,N can be used^[l]It indicates；

L layers of convolution kernel can indicate are as follows:

It wherein 19 × 19 indicates to divide the image into 19 × 19 target frames, 50=5 × (5+5), wherein indicating for first 5 every A target frame can detecte five objects, and second 5 refers to and can detecte 5 kinds of defects, and each defect is by 5 (thirds 5) kinds Information indicates that five kinds of information are respectively as follows: probability of occurrence, defect center abscissa, defect center ordinate, defect width, defect Length；

S3.3: initialization weight initializes every layer of convolution kernel；

S3.4: being trained model using the sample demarcated, and obtains the model of high-accuracy detection test specimen defect；

S3.5: real-time defects detection is realized to the magneto optic images that transmission is come in using trained model.

Compared with prior art, the beneficial effect of technical solution of the present invention is:

(1) present invention does not need large volume using stationary magnetic field source and alternating magnetic field source compound rectifier excitation technology, excitation unit Coil, detection device may be implemented integrated, possess small in size, light weight advantage.

(2) resultant field magneto-optic imaging non-destructive detection technique solves the magneto optic images under stationary magnetic field and is easily saturated and is difficult to The problem of detecting deeper defect.

(3) resultant field magneto-optic imaging non-destructive detection technique solves the detection of tiny flaw difficulty and the skin that becomes under alternating magnetic field The problem of effect.

(4) device security is high, and easy to operate, small volume, low energy consumption, adaptable, and detection accuracy is high.

(5) resultant field magneto-optic imaging non-destructive detection technique is that welding defect is converted into image to handle, can High depending on change degree, image processing method is relatively more open, and application prospect is wide.

Detailed description of the invention

Fig. 1 is resultant field magneto-optic imaging non-destructive detecting system schematic diagram described in embodiment 1.

Fig. 2 is resultant field magneto-optic imaging nondestructive inspection method flow chart described in embodiment 2.

Fig. 3 is convolutional neural networks illustraton of model, carries out weld defects detection to the magneto optic images.,

In figure, 1- two-dimension moving platform, the stationary magnetic field 2- source, 3- alternating magnetic field source, 4- magneto-optical sensor, 5- alterating and direct current Source, 6- processor, 7- workpiece to be detected.

Specific embodiment

The attached figures are only used for illustrative purposes and cannot be understood as limitating the patent；

In order to better illustrate this embodiment, the certain components of attached drawing have omission, zoom in or out, and do not represent actual product Size；

To those skilled in the art, it is to be understood that certain known features and its explanation, which may be omitted, in attached drawing 's.

The following further describes the technical solution of the present invention with reference to the accompanying drawings and examples.