A nano-magnetic fluid magnetic induction permeation detection method and system driven by a high gradient strong magnetic field

Abstract

The application belongs to the technical field of nondestructive testing and evaluation, and particularly relates to a kind of nanometer magnetic fluid magnetic induction permeation detection method and system driven by high gradient strong magnetic field. The method comprises: quantitative spraying is carried out on the surface of aluminum alloy test piece in the area to be detected, forming a nanometer magnetic fluid covering layer; an excitation current is applied to the area to be detected, and a spatial non-uniform high gradient strong magnetic field is established in the area to be detected; under the continuous action of the high gradient strong magnetic field, the nanometer magnetic fluid in the nanometer magnetic fluid covering layer penetrates and gathers into the defects on the surface of the aluminum alloy test piece along the magnetic field gradient direction; the area to be detected after magnetic induction permeation development is subjected to image acquisition, and original image data reflecting the aggregation and distribution state of the magnetic fluid are obtained; the original image data are input into the constructed image intelligent recognition algorithm model, feature extraction and defect discrimination are carried out, and defect detection results are obtained. The application can significantly improve the recognition ability of small and narrow crack defects on the surface of aluminum alloy.

Description

Technical Field

[0001] This invention belongs to the field of nondestructive testing and evaluation technology, specifically relating to a high-gradient strong magnetic field driven nanomagnetic fluid magnetic induced permeation detection method and system. Background Technology

[0002] High-speed trains, as a vital component of modern transportation, have become the preferred mode of travel for the public due to their high speed, punctuality, and large load capacity. With the rapid development of rail transit technology, the safety and operational reliability of train structures have become paramount in ensuring operational safety. To balance the requirements of lightweight design and structural strength, high-speed train bodies widely utilize aluminum alloy materials. Aluminum alloys, due to their lightweight, high strength, corrosion resistance, and excellent processing stability, are not only the mainstream material for train bodies but also widely used in key fields such as aerospace and automotive manufacturing. In the manufacturing process of aluminum alloy components, welding is the core process for achieving structural connections. However, due to the influence of welding thermal cycles, material properties, and the process environment, welded joints often have inherent defects such as incomplete penetration, microcracks, undercut, and slag inclusions. These defects can easily become stress concentration sources under service loads, leading to crack propagation and localized damage, ultimately potentially causing structural fracture. Once weld defects cause a malfunction, it will pose a serious threat to the operational safety of the train. Therefore, achieving full-section, high-precision non-destructive testing of aluminum alloy welds in high-speed trains is of great significance for ensuring the safe operation of trains.

[0003] Currently, non-destructive testing methods for microcracks on and near the surface of workpieces mainly include ultrasonic testing, magnetic particle testing, penetrant testing, and magnetic flux leakage testing. However, these methods all have significant limitations in the inspection of aluminum alloy welds in high-speed trains: Traditional ultrasonic testing, when inspecting thin plate structures, has a blind zone due to transducer limitations, making it difficult to effectively capture small, narrow cracks on and near the surface; aluminum alloys are typical non-ferromagnetic materials with a relative permeability close to 1, making it difficult to generate a macroscopic magnetization response under an external magnetic field, leading to the failure of magnetic particle testing and magnetic flux leakage testing; although penetrant testing can detect some surface open defects, its testing process is cumbersome, and its ability to penetrate sub-millimeter-level narrow cracks and closed defects is limited. Furthermore, manual interpretation of results is greatly affected by subjective factors, easily leading to missed or false detections.

[0004] In summary, existing detection methods are insufficient to fully meet the high-precision detection requirements of small and narrow defects in high-speed train welds. Therefore, there is an urgent need to develop a novel detection method that combines high-resolution imaging with intelligent recognition capabilities. This method is applicable to non-ferromagnetic aluminum alloys, enabling high-contrast imaging of small and narrow cracks. Combined with intelligent image analysis, it achieves automatic defect identification and quantitative assessment, thus providing reliable technical support for high-quality full-section inspection of high-speed train welds. Summary of the Invention

[0005] To address the shortcomings of existing technologies in detecting minute and narrow defects in aluminum alloy welds of high-speed trains—namely, insufficient sensitivity, limited material adaptability, and weak intelligent recognition capabilities—this invention provides a high-gradient, strong magnetic field-driven nano-magnetic fluid magnetic infiltration detection method and system. By comprehensively utilizing magnetic field-driven active infiltration of nano-magnetic fluid, high-contrast defect imaging, and intelligent image recognition, this method achieves high-precision full-section detection of minute and narrow defects in aluminum alloy welds. It improves the accuracy and automation level of identifying minute and narrow defects in non-ferromagnetic materials, providing a reliable technical means for the safety assessment of welds in high-speed trains.

[0006] To achieve the above objectives, the present invention provides the following solution: A high-gradient, strong magnetic field-driven method for detecting magnetically induced permeation in nanomagnetic fluids, the method comprising: A quantitative coating is applied to the area to be tested on the surface of the aluminum alloy specimen to form a nano-magnetic fluid coating layer; An excitation current is applied to the region to be detected to establish a spatially non-uniform high-gradient strong magnetic field within the region to be detected. Under the continuous action of the high gradient strong magnetic field, the nano-magnetic fluid in the nano-magnetic fluid coating layer penetrates and accumulates into the defects on the surface of the aluminum alloy specimen along the magnetic field gradient direction; Image acquisition is performed on the area to be detected after magnetically induced permeation development to obtain raw image data reflecting the aggregation and distribution state of the magnetofluid; The original image data is input into the constructed image intelligent recognition algorithm model to perform feature extraction and defect identification, and obtain defect detection results.

[0007] Preferably, the method for quantitatively spraying a nano-magnetic fluid coating onto the surface of an aluminum alloy specimen to form the test area includes: A magnetic fluid spraying device is used to uniformly spray nano-magnetic fluid onto the test area on the surface of an aluminum alloy specimen. The spraying duration is 1-2 seconds, and the spraying pressure is kept stable, so that the nano-magnetic fluid forms a continuous, uniform and stably adhered nano-magnetic fluid coating layer in the test area.

[0008] Preferably, the method for applying an excitation current to the region to be detected to establish a spatially non-uniform high-gradient strong magnetic field within the region to be detected includes: An excitation current is applied to the area to be detected using an excitation magnetization device to establish a spatially non-uniform high-gradient strong magnetic field in the area to be detected, with a magnetic field gradient of not less than 12T / m. The excitation magnetization device adopts an electromagnet structure, including a U-shaped high-permeability magnetic core and a copper coil wound on the U-shaped high-permeability magnetic core. The area to be detected is located in the magnetic field action area between the two magnetic poles, and the duration of a single excitation is set to 1-2 seconds.

[0009] Preferably, the method for the nano-magnetic fluid in the nano-magnetic fluid coating layer to penetrate and accumulate into the defects on the surface of the aluminum alloy specimen along the magnetic field gradient direction under the continuous action of the high gradient strong magnetic field includes: The nano-magnetic fluid in the nano-magnetic fluid coating layer is magnetized under the action of the high gradient strong magnetic field, forming a magnetic domain arrangement structure oriented along the magnetic field direction. During the continuous loading of the high gradient strong magnetic field, when the magnetic volume force is greater than the sum of the surface tension and viscous resistance, the nano-magnetic fluid enters the crack inside the surface of the aluminum alloy specimen and gradually forms an aggregation region, forming identifiable defect magnetic induced penetration imaging characteristics.

[0010] Preferably, the method for acquiring images of the area to be detected after magnetically induced permeation development to obtain raw image data reflecting the aggregation and distribution state of the magnetofluid includes: An industrial camera is used to acquire images of the test area after magnetically induced penetrant development, obtaining raw image data containing defect magnetically induced penetrant development feature information.

[0011] Preferably, the method for inputting the original image data into the constructed image intelligent recognition algorithm model to perform feature extraction and defect identification, and obtaining defect detection results includes: The original image data is subjected to wavelet transform noise reduction and contrast enhancement preprocessing using an adaptive histogram equalization algorithm. Then, the region containing magnetohydrodynamic aggregation is determined based on adaptive threshold segmentation to obtain preprocessed image data. Construct an image intelligent recognition algorithm model and perform adaptive optimization to obtain the optimal image intelligent recognition algorithm model; The preprocessed image data is input into the optimal image intelligent recognition algorithm model for feature extraction and defect identification to obtain defect detection results.

[0012] Preferably, the method for inputting the preprocessed image data into the optimal image intelligent recognition algorithm model to perform feature extraction and defect discrimination, and obtaining defect detection results includes: The preprocessed image data is subjected to multi-scale feature extraction by the backbone feature extraction network in the optimal image intelligent recognition algorithm model to obtain feature maps at different levels. The feature extraction and fusion module and upsampling structure in the neck network of the optimal image intelligent recognition algorithm model are used to perform cross-layer fusion and enhancement processing on the feature maps of different levels, and then the head network is used for localization and classification mapping to obtain defect detection results.

[0013] This invention also provides a high-gradient, strong magnetic field-driven nanomagnetic fluid magnetic induced permeation detection system. The system is used to implement the aforementioned method and includes: a core control unit, a magnetic fluid spraying unit, a magnetization driving unit, an image acquisition unit, an image intelligent recognition unit, and a data storage unit. The core control unit is used to perform time-series logic control and coordinated scheduling of each unit of the system. By preset the trigger time, working duration and process switching sequence of each unit, the system can achieve automated control and overall coordination of the magnetic induction detection process. The magnetic fluid spraying unit is used to quantitatively spray the area to be tested on the surface of the aluminum alloy specimen to form a continuous, uniform and stably adhered nano-magnetic fluid coating layer. The magnetization driving unit is used to generate a high gradient strong magnetic field in the area to be detected, and to magnetize and drive the nano-magnetic fluid in the nano-magnetic fluid coating layer, so that the nano-magnetic fluid penetrates and accumulates into the defect area on the surface of the aluminum alloy specimen under the action of the magnetic field gradient. The image acquisition unit is used to acquire images of the spatial distribution of the magnetic response formed by the nano-magnetic fluid at the defect, and obtain raw image data reflecting the aggregation and distribution state of the magnetic fluid. The image intelligent recognition unit is connected to the image acquisition unit to obtain the original image data, and performs feature extraction and defect discrimination on the original image data through the built-in image intelligent recognition algorithm model to obtain defect detection results; The data storage unit is used to perform structured storage of the raw image data and defect detection results generated during the detection process.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a high-gradient, strong magnetic field-driven magnetic infiltration detection method for nano-magnetic fluids. Starting from the detection mechanism, it analyzes the magnetization response behavior and flow characteristics of nano-magnetic fluids under a strong magnetic field, and studies the dynamic influence of a high-gradient, strong magnetic field on the stress state, infiltration velocity, and infiltration depth of the magnetic fluid. By introducing a high-gradient, strong magnetic field driving mechanism, the nano-magnetic fluids achieve directional infiltration and aggregation in defect regions, overcoming the limitations of traditional magnetic particle or conventional infiltration detection methods in responding insufficiently to microcracks. Combined with intelligent image recognition algorithms, it effectively improves defect detection sensitivity, recognition accuracy, and engineering application efficiency, making the detection results more reliable. This method not only accurately reflects the infiltration behavior and defect morphology characteristics inside microcracks but also adapts to the engineering detection needs of different materials, structural forms, and complex working conditions.

[0015] The magnetic induction penetrant testing method provided by this invention belongs to advanced non-destructive testing technology. Through real-time detection and intelligent analysis of the surface condition of actual engineering specimens, the testing system possesses a higher level of intelligence and engineering applicability. The implementation of this invention can significantly improve the ability to identify minute, narrow crack defects, enhance the accuracy and effectiveness of reliability assessment for key load-bearing structures of high-speed trains, and provide reliable technical support for high-speed train operation safety monitoring. Simultaneously, this invention promotes the development of magnetic induction penetrant non-destructive testing technology towards higher sensitivity, intelligence, and engineering practicality, and has significant potential for widespread application. Attached Figure Description

[0016] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the process of the high-gradient strong magnetic field driven nanomagnetic fluid magnetic induced permeation detection method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the intelligent recognition process according to an embodiment of the present invention; Figure 3 This is a structural diagram of the image intelligent recognition model according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the magnetically induced permeation detection system according to an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] Example 1 This invention provides a high-gradient, high-magnetic-field-driven method for detecting magnetically induced permeation in nanofluids, comprising: A quantitative coating is applied to the area to be tested on the surface of the aluminum alloy specimen to form a nano-magnetic fluid coating layer; An excitation current is applied to the region to be detected to establish a spatially non-uniform high-gradient strong magnetic field within the region to be detected. Under the continuous action of the high gradient strong magnetic field, the nano-magnetic fluid in the nano-magnetic fluid coating layer penetrates and accumulates into the defects on the surface of the aluminum alloy specimen along the magnetic field gradient direction; Image acquisition is performed on the area to be detected after magnetically induced permeation development to obtain raw image data reflecting the aggregation and distribution state of the magnetofluid; The original image data is input into the constructed image intelligent recognition algorithm model to perform feature extraction and defect identification, and obtain defect detection results.

[0021] like Figure 1-Figure 3 As shown, the specific implementation process of the present invention is as follows: S1. The magnetic fluid spraying device is driven by the control system to spray a quantitative amount of the area to be tested, forming a continuous, uniform and stable nano-magnetic fluid coating layer on the surface of the specimen.

[0022] The magnetic fluid spraying device is used to uniformly spray nano-magnetic fluid onto the surface of the aluminum alloy specimen to be tested, so that the nano-magnetic fluid forms a continuous and stable coating layer in the area to be tested.

[0023] The nano-magnetic fluid spraying process is automatically executed by the magnetic fluid spraying device under the control of the control system. A quantitative amount of magnetic fluid is sprayed onto the surface of the specimen in the area to be tested for 1-2 seconds, with the spraying pressure maintained at a stable level. This ensures that the nano-magnetic fluid forms a continuous, uniform, and stably adhered liquid film layer in the testing area, guaranteeing that the magnetic fluid can fully respond to the magnetic field on the surface during subsequent magnetization. To investigate the influence of magnetic response characteristics on permeation behavior and detection results, Fe3O4 nanoparticles with a particle size of 5-15 nm were used, and water-based and oil-based carrier fluids were used to prepare nano-magnetic fluids with a volume fraction of 10%-50%. Comparative permeation detection experiments were conducted under an applied magnetic field.

[0024] S2. The control system starts the excitation magnetization device and applies an excitation current to establish a spatially non-uniform high gradient strong magnetic field in the area to be detected, forming a magnetic field driving environment with a directional gradient distribution, so that the detection area has the conditions for magnetic volume force driving.

[0025] Specifically, the excitation magnetization device adopts an electromagnet structure, including a U-shaped high-permeability magnetic core and a copper coil wound on the U-shaped high-permeability magnetic core, and the area to be measured is located in the effective magnetic field area between the two magnetic poles.

[0026] The control system supplies a current signal to the excitation coil, with the duration of a single excitation set to 1-2 seconds. This establishes a spatially non-uniform high-gradient magnetic field environment in the detection area, ensuring that the magnetic field gradient in the detection area is no less than 12 T / m. Under the influence of the high-gradient magnetic field, the nano-magnetic fluid particles distributed on the surface under test are driven by magnetic volume forces to accelerate their penetration into surface cracks or defects and accumulate, thereby enhancing the magnetic response differences and image contrast in the defect area.

[0027] S3. Under the continuous action of the high gradient strong magnetic field, the control system maintains the excitation state, so that the nano-magnetic fluid generates a magnetization response, and under the magnetic volume force driven mechanism formed by the magnetic field gradient force, it penetrates into the defect opening direction along the magnetic field gradient direction.

[0028] Specifically, during the continuous loading of a high-gradient strong magnetic field, a magnetization response is generated inside the magnetohydrodynamic fluid, and the force it experiences satisfies the expression for the volume force of a magnet: (1) (2) in, F m It is a magnetic volume force. M The magnetization intensity, m 0 is the permeability of free space. x , is the magnetic susceptibility H denoted as , where is the magnetic field strength.

[0029] Nanomagnetic fluids overcome surface tension and flow viscosity resistance to enter the crack interior and accumulate in the defect region, creating a stable magnetic fluid distribution difference between the defect region and the intact region, thus establishing identifiable magnetically induced penetration imaging characteristics.

[0030] Specifically, during the infiltration process, the nanomagnetic fluid needs to overcome the capillary resistance and flow viscous resistance formed by surface tension at the crack inlet. For a width of... w For a crack, the equivalent capillary resistance can be approximated as: (3) The viscous resistance to flow is: (4) in, c The surface tension coefficient, i Contact angle, w The width of the crack; or Let be the viscosity coefficient of the fluid. L This is the length of the penetration path. v This represents the penetration rate.

[0031] During the continuous loading of the magnetic field, when F m> F σ + F μ At this point, the nano-magnetic fluid can overcome resistance and enter the crack interior, gradually forming an accumulation region along the crack interior. In the defect region, the penetration and accumulation of the magnetic fluid makes its distribution density significantly higher than that of the surrounding intact region, thus creating a significant difference in magnetic response at the micro-crack.

[0032] S4. After the magnetization drive is completed, the control system sends a trigger signal to the industrial camera to acquire images of the test area that has completed magnetic induced permeation development and obtain raw image data reflecting the aggregation and distribution state of the magnetofluid. Specifically, after the excitation magnetization device stops excitation, the control system of the detection system sends a trigger signal to the industrial camera, triggering the industrial camera to capture images of the developed test area. The industrial camera automatically adjusts according to various preset parameters to acquire high-definition, high signal-to-noise ratio raw image data reflecting the aggregation and distribution state of the magnetohydrodynamics.

[0033] S5. Input the collected raw image data into the image intelligent recognition algorithm model to perform feature extraction and defect identification.

[0034] Specifically, the following steps are included: S51: The acquired raw image data is subjected to wavelet transform noise reduction and adaptive histogram equalization algorithm for contrast enhancement preprocessing to suppress background noise and enhance the contrast difference between the defect area and the background. Then, the region containing magnetohydrodynamic aggregation is determined based on adaptive threshold segmentation. Finally, the preprocessed image data is obtained. First, wavelet transform denoising filtering is performed on the original image. Wavelet basis functions are selected to perform two-dimensional discrete wavelet decomposition on the image, obtaining low-frequency approximation coefficients and high-frequency detail coefficients in the horizontal, vertical, and diagonal directions. The decomposition process can be expressed as: (5) in, Aj This is a low-frequency approximation component. Hj For high-frequency detail components in the horizontal direction, Vj These are the high-frequency detail components in the vertical direction. DJ These are the high-frequency detail components in the diagonal direction. and ψ These are the scaling function and wavelet function determined by the selected wavelet basis, respectively. j For wavelet decomposition scale levels, ( x , y ) represents the image spatial domain coordinates, ( m , n ) represents the translation parameter of the wavelet transform. m For horizontal translation index, nFor vertical translation index, I This is the function for calculating the grayscale values ​​of the original image.

[0035] A soft thresholding function is applied to the high-frequency coefficients obtained from the decomposition to denoise them. The thresholding formula is as follows: (6) in, s The standard deviation of noise. N The number of coefficients.

[0036] The high-frequency coefficients after thresholding and the retained low-frequency coefficients are reconstructed using inverse wavelet transform to obtain the denoised image. The reconstruction expression is: (7) in, A J For the first J Layer approximation coefficients J The maximum number of wavelet decomposition levels. For the first j The horizontal detail factor of the layer after thresholding. For the first j Vertical detail coefficient of the layer after thresholding. For the first j Diagonal detail coefficients of the layer after thresholding. I ′( x , y () is the output image reconstructed after wavelet denoising.

[0037] Secondly, the denoised image undergoes contrast enhancement processing. A contrast-limited adaptive histogram equalization algorithm is used to divide the image into several local sub-blocks, calculate the gray-level histogram in each sub-block, and correct the histogram according to a preset cropping limit. Finally, adaptive thresholding based on maximum inter-class variance is used to determine magnetohydrodynamic (MHD) aggregation regions. The segmentation process is as follows: The total number of image pixels is M grayscale value i The number of pixels is n i The pixel ratio of the foreground to the background is: (8) (9) in, oh 0 represents the area percentage of the background region. oh 1 represents the area percentage of the foreground region. T The segmentation threshold to be optimized. L This represents the total number of gray levels in the image.

[0038] The average grayscale value of the foreground and background is defined as: (10) (11) in, m 0 represents the average grayscale value of the background area. m 1 represents the average gray value of the foreground region.

[0039] The formula for between-class variance is: (12) Find the threshold that maximizes the inter-class variance by traversing the gray levels. T otsu And perform binarization segmentation on the image: (13) in, I ROI This is the segmented binarized output image. I pre The grayscale values ​​are the preprocessed input image values.

[0040] S52: Construct an image intelligent recognition algorithm model structure for identifying small and narrow defects. The image intelligent recognition algorithm adopts an improved YOLO11 model, mainly including a backbone feature extraction network, a neck network, and a head network, used for crack identification in magnetically induced permeation preprocessed images. The backbone network consists of an input layer, ordinary convolutional layers, and CSP layers. The model receives magnetically induced permeation preprocessed images and gradually expands the number of feature map channels from 64 to 512 through cross-stage partial connections to extract high-level features from the input image. The neck network is composed of a path aggregation network and a feature pyramid network, performing top-down and bottom-up path processing simultaneously to enhance multi-scale feature representation capabilities and embedding a CBAM attention mechanism to improve small target detection capabilities. The CBAM attention mechanism is divided into channel attention and spatial attention, with the following formulas: (14) (15) (16) in, Mc This is a channel attention map. Ms. This is a spatial attention map. For convolution operations with a kernel size of 7×7, This is the intermediate feature map after channel attention weighting. This is the final feature map after being weighted by both channel attention and spatial attention. F For the input feature map, C For the number of channels,H , W For space dimensions, s It is the Sigmoid activation function. MLP It is a multilayer perceptron. AvgPool For global average pooling, MaxPool This is for global max pooling.

[0041] The head network is the output part of the model and is responsible for the final task recognition.

[0042] S53: Construct a loss function based on SIoU Loss, Focal Loss and BCE Loss to address the class imbalance and small target localization accuracy issues in the detection of small and narrow defects. Dynamically adjust the network weight parameters and learning rate according to the deviation between the predicted results and the true values ​​to achieve adaptive optimization of feature weights until the loss function converges, and obtain the optimal image intelligent recognition algorithm model after training.

[0043] The formula for the SIoU Loss function is: (17) in, IoU For intersection-over-union (IoU), the degree of overlap between the predicted bounding box and the ground truth bounding box is calculated. D Ω represents the distance loss value, and Ω represents the shape loss value.

[0044] The formula for the Focal Loss function is: (18) in, p t This represents the model's predicted probability of the true class. α t As a category balance factor, k For focusing parameters.

[0045] The BCE Loss function formula is: (19) in, y For real labels, p This represents the probability predicted by the model.

[0046] Finally, the joint loss function formula is: (20) in, l coord 、l cls , l obj These are the weighting coefficients for location, classification, and reliability loss, respectively.

[0047] S54: Input the preprocessed image data into the optimal image intelligent recognition algorithm model, and through the mapping and parsing of the backbone feature extraction network, neck network, and head network, output a discrimination result containing the micro-narrow defect information at the output end, specifically including: The preprocessed image data is subjected to multi-scale feature extraction by the backbone feature extraction network in the optimal image intelligent recognition algorithm model to obtain feature maps at different levels. The feature extraction and fusion module and upsampling structure in the neck network of the optimal image intelligent recognition algorithm model are used to perform cross-layer fusion and enhancement processing on the feature maps of different levels, and then the head network is used for localization and classification mapping to obtain defect detection results.

[0048] In summary, this invention discloses a high-gradient, strong magnetic field-driven method for detecting magnetically induced penetration of nano-magnetic fluids: A spraying device is used to uniformly spray nano-magnetic fluid onto the surface of the aluminum alloy specimen to be tested, forming a continuous and stable coating layer in the test area. A non-uniform spatial strong magnetic field is generated using an excitation device, driving the nano-magnetic fluid to overcome surface tension and viscous resistance through magnetic volume force, causing the nano-magnetic fluid to penetrate and accumulate into the defect along the magnetic field gradient direction. Subsequently, an industrial camera is used to acquire images of the developed defect area, obtaining image data containing the magnetically induced penetration characteristic information of the defect. Finally, an image intelligent recognition algorithm model based on multi-dimensional feature extraction is established. The acquired image data is input into the image intelligent recognition algorithm model for feature extraction and intelligent recognition, achieving intelligent extraction and high-precision recognition of short weld defects. This invention significantly improves the detection sensitivity of small defects in non-ferromagnetic materials. The method of this invention features high sensitivity for detecting small and narrow defects, applicability to non-ferromagnetic materials, fast detection speed, and high degree of intelligence, and can be widely applied in the field of non-destructive testing of the entire cross-section of welds in high-speed trains.

[0049] Example 2 like Figure 4 As shown, based on the same inventive concept, this invention also provides a high-gradient, strong magnetic field-driven nanomagnetic fluid magnetic induced permeation detection system for implementing the methods described in the foregoing embodiments. The system includes: a core control unit, a magnetic fluid spraying unit, a magnetization driving unit, an image acquisition unit, an image intelligent recognition unit, and a data storage unit. The core control unit is used for the timing logic control and coordinated scheduling of various functional units of the nano-magnetic fluid magnetic induced permeation detection system. By preset the trigger time, working duration, and process switching sequence of each unit, the system achieves automated control and overall coordination of the magnetic induced detection process. The magnetic fluid spraying unit is used to store and spray nano-magnetic fluid, and to quantitatively spray it onto the surface of the test specimen to form a continuous, uniform and stable nano-magnetic fluid coating layer. The magnetization drive unit is used to generate a high gradient strong magnetic field in the test area of ​​the test piece, and to magnetize the nano-magnetic fluid sprayed on the test surface, so that the nano-magnetic fluid penetrates into the defect area and accumulates under the action of the magnetic field gradient. In particular, this module can precisely adjust the amplitude and duration of the pulse current to control the magnetohydrodynamic penetration rate and accumulation depth, thereby obtaining a highly sensitive magnetic response at tiny narrow defects.

[0050] The image acquisition unit is used to capture high-resolution images of the spatial distribution of the magnetic response formed by the nano-magnetic fluid at the defect, and to obtain raw image data reflecting the morphological characteristics of the defect. An image intelligent recognition unit is communicatively connected to the image acquisition unit to obtain the original image data. It has a built-in deep learning image intelligent recognition algorithm model for automatically extracting multi-dimensional features and identifying defect patterns from the original image data. In particular, this unit uses deep learning recognition algorithms to automatically extract features and identify defects in images, and combines the characteristics of magnetohydrodynamic permeation distribution for discrimination, thereby improving the accuracy of identifying tiny and narrow defects.

[0051] The data storage unit is used to store and manage the raw image data, intermediate processing data, and intelligent recognition results generated during the detection process.

[0052] This system has the following advantages: High sensitivity detection capability: By utilizing the directional volume force generated by the high gradient strong magnetic field, the nano-magnetic fluid is actively driven to penetrate and highly accumulate deep into the micro-narrow cracks, which can significantly enhance the magnetic response difference at the micro-narrow cracks and improve the detection capability of micro-narrow crack defects.

[0053] Automated inspection process: Relying on the core control unit, the timing coordination control of spraying, magnetization drive, image acquisition and recognition analysis is realized, so that the inspection process can be completed continuously, stably and automatically.

[0054] Strong intelligent recognition capability: It integrates an intelligent image recognition unit to automatically extract features and identify defects in magnetically induced permeation images, thereby improving the accuracy of defect identification and reducing manual intervention.

[0055] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A high-gradient, strong magnetic field-driven method for detecting magnetically induced permeation in nanomagnetic fluids, characterized in that, The method includes: A quantitative coating is applied to the area to be tested on the surface of the aluminum alloy specimen to form a nano-magnetic fluid coating layer; An excitation current is applied to the region to be detected to establish a spatially non-uniform high-gradient strong magnetic field within the region to be detected. Under the continuous action of the high gradient strong magnetic field, the nano-magnetic fluid in the nano-magnetic fluid coating layer penetrates and accumulates into the defects on the surface of the aluminum alloy specimen along the magnetic field gradient direction; Image acquisition is performed on the area to be detected after magnetically induced permeation development to obtain raw image data reflecting the aggregation and distribution state of the magnetofluid; The original image data is input into the constructed image intelligent recognition algorithm model to perform feature extraction and defect identification, and obtain defect detection results.

2. The method according to claim 1, characterized in that, Methods for quantitatively spraying a nano-magnetic fluid coating onto the surface of an aluminum alloy specimen to form a protective layer include: A magnetic fluid spraying device is used to uniformly spray nano-magnetic fluid onto the test area on the surface of an aluminum alloy specimen. The spraying duration is 1-2 seconds, and the spraying pressure is kept stable, so that the nano-magnetic fluid forms a continuous, uniform and stably adhered nano-magnetic fluid coating layer in the test area.

3. The method according to claim 1, characterized in that, The method for applying an excitation current to the region to be detected and establishing a spatially non-uniform high-gradient strong magnetic field within the region to be detected includes: An excitation current is applied to the area to be detected using an excitation magnetization device to establish a spatially non-uniform high-gradient strong magnetic field in the area to be detected, with a magnetic field gradient of not less than 12T / m. The excitation magnetization device adopts an electromagnet structure, including a U-shaped high-permeability magnetic core and a copper coil wound on the U-shaped high-permeability magnetic core. The area to be detected is located in the magnetic field action area between the two magnetic poles, and the duration of a single excitation is set to 1-2 seconds.

4. The method according to claim 1, characterized in that, The method for infiltrating and accumulating nano-magnetic fluid in the nano-magnetic fluid coating layer into the defects on the surface of the aluminum alloy specimen along the magnetic field gradient direction under the continuous action of the high gradient magnetic field includes: The nano-magnetic fluid in the nano-magnetic fluid coating layer is magnetized under the action of the high gradient strong magnetic field, forming a magnetic domain arrangement structure oriented along the magnetic field direction. During the continuous loading of the high gradient strong magnetic field, when the magnetic volume force is greater than the sum of the surface tension and viscous resistance, the nano-magnetic fluid enters the crack inside the surface of the aluminum alloy specimen and gradually forms an aggregation region, forming identifiable defect magnetic induced penetration imaging characteristics.

5. The method according to claim 1, characterized in that, Methods for acquiring images of the target area after magnetically induced permeation development to obtain raw image data reflecting the aggregation and distribution state of magnetofluid include: An industrial camera is used to acquire images of the test area after magnetically induced penetrant development, obtaining raw image data containing defect magnetically induced penetrant development feature information.

6. The method according to claim 1, characterized in that, The method for inputting the original image data into a constructed image intelligent recognition algorithm model to perform feature extraction and defect identification, and obtaining defect detection results, includes: The original image data is subjected to wavelet transform noise reduction and contrast enhancement preprocessing using an adaptive histogram equalization algorithm. Then, the region containing magnetohydrodynamic aggregation is determined based on adaptive threshold segmentation to obtain preprocessed image data. Construct an image intelligent recognition algorithm model and perform adaptive optimization to obtain the optimal image intelligent recognition algorithm model; The preprocessed image data is input into the optimal image intelligent recognition algorithm model for feature extraction and defect identification to obtain defect detection results.

7. The method according to claim 6, characterized in that, The method for inputting the preprocessed image data into the optimal image intelligent recognition algorithm model to perform feature extraction and defect discrimination, and obtaining defect detection results includes: The preprocessed image data is subjected to multi-scale feature extraction by the backbone feature extraction network in the optimal image intelligent recognition algorithm model to obtain feature maps at different levels. The feature extraction and fusion module and upsampling structure in the neck network of the optimal image intelligent recognition algorithm model are used to perform cross-layer fusion and enhancement processing on the feature maps of different levels, and then the head network is used for localization and classification mapping to obtain defect detection results.

8. A high-gradient, high-magnetic-field-driven nanomagnetic fluid magnetic-induced permeation detection system, said system being used to implement the method described in any one of claims 1-7, characterized in that, The system includes: a core control unit, a magnetic fluid spraying unit, a magnetization drive unit, an image acquisition unit, an image intelligent recognition unit, and a data storage unit. The core control unit is used to perform time-series logic control and coordinated scheduling of each unit of the system. By preset the trigger time, working duration and process switching sequence of each unit, the system can achieve automated control and overall coordination of the magnetic induction detection process. The magnetic fluid spraying unit is used to quantitatively spray the area to be tested on the surface of the aluminum alloy specimen to form a continuous, uniform and stably adhered nano-magnetic fluid coating layer. The magnetization driving unit is used to generate a high gradient strong magnetic field in the area to be detected, and to magnetize and drive the nano-magnetic fluid in the nano-magnetic fluid coating layer, so that the nano-magnetic fluid penetrates and accumulates into the defect area on the surface of the aluminum alloy specimen under the action of the magnetic field gradient. The image acquisition unit is used to acquire images of the spatial distribution of the magnetic response formed by the nano-magnetic fluid at the defect, and obtain raw image data reflecting the aggregation and distribution state of the magnetic fluid. The image intelligent recognition unit is connected to the image acquisition unit to obtain the original image data, and performs feature extraction and defect discrimination on the original image data through the built-in image intelligent recognition algorithm model to obtain defect detection results; The data storage unit is used to perform structured storage of the raw image data and defect detection results generated during the detection process.

Images