Automatic control system and method in the processing of rare earth permanent magnet materials

Abstract

The present application relates to the field of intelligent control, and specifically discloses an automatic control system and method in the processing of rare earth permanent magnet materials, which uses an X-ray image of the rare earth permanent magnet material to be detected collected by an X-ray fluorescence spectrometer, and uses deep learning technology to perform image analysis and processing on the X-ray image of the rare earth permanent magnet material to be detected, thereby determining whether the quality of the rare earth permanent magnet material is qualified, and controlling the early warning device to issue an early warning indication. In this way, tiny defects or abnormalities of rare earth permanent magnet materials can be identified, and high-precision detection of material quality can be achieved, reducing labor costs and time costs, reducing quality risks in the production process, and greatly improving production efficiency and product quality stability.

Description

Technical Field

[0001] The present application relates to the field of intelligent control, and more specifically, to an automatic control system and method for rare earth permanent magnet material processing. Background Art

[0002] Rare earth permanent magnets are a class of permanent magnets made from rare earth elements (such as neodymium, praseodymium, and promethium). These materials typically possess high magnetic energy products and coercive forces, making them widely used in modern electronics, communications, automotive, and medical equipment.

[0003] Quality inspection during the processing of rare earth permanent magnet materials is crucial. Specifically, the quality of rare earth permanent magnet materials directly affects their magnetic properties, such as remanence, coercivity, and maximum magnetic energy product. The presence of defects or impurities can reduce the reliability of rare earth permanent magnet materials. Therefore, quality inspection ensures that rare earth permanent magnet materials meet the performance requirements of their intended applications and identifies and eliminates these defects, thereby improving the reliability of rare earth permanent magnet materials in harsh environments or critical applications.

[0004] However, traditional methods require destructive testing of samples during the processing of rare earth permanent magnet materials, such as metallographic microscopy and mechanical testing. This makes quality inspection of mass-produced materials expensive, time-consuming, and potentially damaging. Furthermore, traditional inspection relies on operator visual inspection, which can fail to detect small defects or impurities in the material. This leads to subjective and inconsistent inspection results, making them prone to errors and inefficient.

[0005] Therefore, an optimized automatic control system for the processing of rare earth permanent magnet materials is expected. Summary of the Invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an automatic control system and method for the processing of rare earth permanent magnet materials, which uses an X-ray image of the rare earth permanent magnet material to be detected collected by an X-ray fluorescence spectrometer, and uses deep learning technology to perform image analysis and processing on the X-ray image of the rare earth permanent magnet material to be detected, thereby determining whether the quality of the rare earth permanent magnet material is qualified, and controlling the early warning device to issue an early warning indication. In this way, tiny defects or abnormalities of rare earth permanent magnet materials can be identified, and high-precision detection of material quality can be achieved, which reduces labor costs and time costs, reduces quality risks in the production process, and greatly improves production efficiency and product quality stability.

[0007] According to one aspect of the present application, an automatic control system for rare earth permanent magnet material processing is provided, which includes:

[0008] An X-ray image acquisition module for rare earth permanent magnet materials, used to acquire an X-ray image of the rare earth permanent magnet material to be tested acquired by an X-ray fluorescence spectrometer;

[0009] a local area feature extraction module, configured to segment the X-ray image of the rare earth permanent magnet material to be detected to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected, and then perform shallow and deep image feature extraction to obtain a sequence of shallow feature maps of the X-ray local area and a sequence of deep feature maps of the X-ray local area;

[0010] a feature enhancement module, configured to perform feature enhancement on the sequence of X-ray local area deep feature maps to obtain a sequence of X-ray local area deep enhanced feature maps;

[0011] a feature dimensionality reduction module, configured to concatenate and pool the sequence of the X-ray local area deep enhancement feature maps and the sequence of the X-ray local area shallow feature maps, respectively, to obtain an X-ray local area deep enhancement cascade feature vector and an X-ray local area shallow cascade feature vector;

[0012] a feature fusion module, configured to perform feature distribution matching based on feature representation conversion on the X-ray local area deep enhancement cascade feature vector and the X-ray local area shallow cascade feature vector to obtain an X-ray fusion feature vector;

[0013] The quality detection module is used to determine whether the quality of the rare earth permanent magnet material is qualified based on the X-ray fusion feature vector, and control the early warning device to issue an early warning indication.

[0014] According to another aspect of the present application, there is provided a method for automatic control during the processing of rare earth permanent magnet materials, comprising:

[0015] Obtaining an X-ray image of the rare earth permanent magnet material to be tested collected by an X-ray fluorescence spectrometer;

[0016] Segmenting the X-ray image of the rare earth permanent magnet material to be detected to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected, and then extracting shallow and deep image features to obtain a sequence of shallow feature maps of the X-ray local area and a sequence of deep feature maps of the X-ray local area;

[0017] Performing feature enhancement on the sequence of X-ray local area deep feature maps to obtain a sequence of X-ray local area deep enhanced feature maps;

[0018] Respectively concatenating and pooling the sequence of the X-ray local area deep enhancement feature maps and the sequence of the X-ray local area shallow feature maps to obtain an X-ray local area deep enhancement cascade feature vector and an X-ray local area shallow cascade feature vector;

[0019] performing feature distribution matching based on feature representation conversion on the X-ray local area deep enhancement cascade feature vector and the X-ray local area shallow cascade feature vector to obtain an X-ray fusion feature vector;

[0020] Based on the X-ray fusion characteristic vector, it is determined whether the quality of the rare earth permanent magnet material is qualified, and the early warning device is controlled to issue an early warning indication.

[0021] Compared with the prior art, the present application provides an automatic control system and method for the processing of rare earth permanent magnet materials. This system uses deep learning technology to analyze and process X-ray images of the rare earth permanent magnet material to be tested, collected by an X-ray fluorescence spectrometer, to determine whether the rare earth permanent magnet material is of qualified quality and control the early warning device to issue an early warning indication. In this way, tiny defects or anomalies in the rare earth permanent magnet material can be identified, achieving high-precision detection of material quality, reducing labor and time costs, reducing quality risks in the production process, and greatly improving production efficiency and product quality stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0023] Figure 1 This is a block diagram of an automatic control system for the rare earth permanent magnet material processing according to an embodiment of the present application.

[0024] Figure 2 4 is a block diagram of a local area feature extraction module in an automatic control system during the processing of rare earth permanent magnet materials according to an embodiment of the present application.

[0025] Figure 3 4 is a block diagram of a feature enhancement module in an automatic control system for rare earth permanent magnet material processing according to an embodiment of the present application.

[0026] Figure 4 This is a flow chart of an automatic control method during the processing of rare earth permanent magnet materials according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] The following describes embodiments of the present disclosure in more detail with reference to the accompanying drawings. While the drawings illustrate certain embodiments of the present disclosure, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of protection of the present disclosure.

[0028] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in a different order and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0029] It should be noted that the terms "first, second, and third" in the embodiments of the present application are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It is understood that the terms "first, second, and third" can be interchanged to represent a specific order or precedence where permitted. It should be understood that the objects distinguished by "first, second, and third" can be interchanged where appropriate, such that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0030] As used in this disclosure and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0031] Various exemplary embodiments, features, and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0032] Rare earth permanent magnets are a class of permanent magnets made from rare earth elements (such as neodymium, praseodymium, and promethium). These materials typically possess high magnetic energy products and coercivity, making them widely used in modern electronics, communications, automobiles, medical devices, and other fields. Specifically, rare earth permanent magnets are generally divided into three categories: neodymium iron boron magnets, cobalt ferrites, and terbium cobalt magnets. Neodymium iron boron magnets are the most common and widely used, exhibiting high magnetic energy products and coercivity, and are widely used in motors, generators, sensors, and other fields. Due to the scarcity and environmental impact of rare earth elements, researchers have recently begun to search for alternatives, such as permanent magnet graphite and magnetic polymers, to reduce dependence on rare earth resources.

[0033] Quality inspection during the processing of rare earth permanent magnet materials is crucial. Specifically, the quality of rare earth permanent magnet materials directly affects their magnetic properties, such as remanence, coercivity, and maximum magnetic energy product. The presence of defects or impurities can reduce the reliability of rare earth permanent magnet materials. Therefore, quality inspection ensures that rare earth permanent magnet materials meet the performance requirements of their intended applications and identifies and eliminates these defects, thereby improving the reliability of rare earth permanent magnet materials in harsh environments or critical applications.

[0034] However, traditional methods require destructive testing of samples during the processing of rare earth permanent magnet materials, such as metallographic microscopy and mechanical testing. This makes quality inspection of mass-produced materials expensive, time-consuming, and potentially damaging. Furthermore, traditional inspection relies on operator visual inspection, which can fail to detect small defects or impurities in the material. This leads to subjective and inconsistent inspection results, making them prone to errors and inefficient.

[0035] Therefore, an optimized automatic control system for the processing of rare earth permanent magnet materials is desired. The system uses X-ray images of the rare earth permanent magnet material to be tested, collected by an X-ray fluorescence spectrometer, and uses deep learning technology to perform image analysis and processing on the X-ray images of the rare earth permanent magnet material to be tested, thereby determining whether the rare earth permanent magnet material is of qualified quality and controlling the early warning device to issue an early warning indication. In this way, tiny defects or anomalies in the rare earth permanent magnet material can be identified, achieving high-precision detection of material quality, reducing labor costs and time costs, reducing quality risks in the production process, and greatly improving production efficiency and product quality stability.

[0036] Figure 1 FIG. 1 is a block diagram of an automatic control system for rare earth permanent magnet material processing according to an embodiment of the present application. Figure 1As shown in FIG, an automatic control system 100 for the processing of rare earth permanent magnet materials according to an embodiment of the present application includes: an X-ray image acquisition module 110 for rare earth permanent magnet materials, for acquiring an X-ray image of the rare earth permanent magnet material to be detected acquired by an X-ray fluorescence spectrometer; a local area feature extraction module 120, for performing image segmentation on the X-ray image of the rare earth permanent magnet material to be detected to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected, and then performing deep and shallow image feature extraction to obtain a sequence of X-ray local area shallow feature maps and a sequence of X-ray local area deep feature maps; a feature enhancement module 130, for performing feature enhancement on the sequence of X-ray local area deep feature maps to obtain an X-ray local area feature map. A sequence of deep-domain enhanced feature maps; a feature dimensionality reduction module 140, for cascading and pooling the sequence of the X-ray local area deep-domain enhanced feature maps and the sequence of the X-ray local area shallow-domain enhanced feature maps, respectively, to obtain an X-ray local area deep-domain enhanced cascade feature vector and an X-ray local area shallow-domain cascade feature vector; a feature fusion module 150, for performing feature distribution matching based on feature representation conversion on the X-ray local area deep-domain enhanced cascade feature vector and the X-ray local area shallow-domain cascade feature vector to obtain an X-ray fusion feature vector; and a quality detection module 160, for determining whether the quality of the rare earth permanent magnet material is qualified based on the X-ray fusion feature vector, and controlling the early warning device to issue an early warning indication.

[0037] In the embodiment of the present application, the X-ray image acquisition module 110 of the rare earth permanent magnet material is used to obtain an X-ray image of the rare earth permanent magnet material to be tested, which is acquired by an X-ray fluorescence spectrometer. It should be understood that, considering that X-ray images can penetrate materials and reveal defects within the material, such as cracks, pores, or non-metallic inclusions, they can show the structure and composition distribution within the material, which can help the detector understand the material's organization, defects, foreign matter, etc., thereby evaluating the quality and performance of the material. Based on this, in the technical solution of the present application, obtaining an X-ray image of the rare earth permanent magnet material to be tested, which is acquired by an X-ray fluorescence spectrometer, can achieve non-destructive testing of the rare earth permanent magnet material, that is, obtaining key information without destroying the material structure, detecting defects in the rare earth permanent magnet material, such as cracks, pores, inclusions, etc., thereby comprehensively evaluating the quality of the rare earth permanent magnet material, helping to formulate quality control strategies, and ensuring that the product meets standard requirements.

[0038] X-ray fluorescence (XRF) is a rapid and non-destructive method for measuring materials. It uses high-energy X-rays or gamma rays to bombard a material, emitting secondary X-rays whose signatures are used for elemental and chemical analysis. XRF is widely used in the study of metals, glass, ceramics, and building materials, as well as in geochemistry, forensics, archaeology, and art analysis.

[0039] In an embodiment of the present application, the local area feature extraction module 120 is used to perform image segmentation on the X-ray image of the rare earth permanent magnet material to be detected to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected, and then perform deep and shallow image feature extraction to obtain a sequence of X-ray local area shallow feature maps and a sequence of X-ray local area deep feature maps. Figure 2 FIG is a block diagram of a local area feature extraction module in an automatic control system for rare earth permanent magnet material processing according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 2 As shown, the local area feature extraction module 120 includes: an image segmentation unit 121, which is used to segment the X-ray image of the rare earth permanent magnet material to be detected to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected; and a deep and shallow fusion convolution unit 122, which is used to input the sequence of X-ray local area images of the rare earth permanent magnet material to be detected into an image feature extractor based on a deep and shallow fusion convolution network model to obtain a sequence of X-ray local area shallow feature maps and a sequence of X-ray local area deep feature maps.

[0040] Specifically, in an embodiment of the present application, the image segmentation unit 121 is used to segment the X-ray image of the rare earth permanent magnet material to be detected to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected. Accordingly, considering that the X-ray image of the rare earth permanent magnet material to be detected contains material information of multiple local areas, and the characteristics of these different local areas are crucial for the quality detection of the entire rare earth permanent magnet material. Therefore, in order to better process and analyze the different layout areas in the X-ray image of the rare earth permanent magnet material to be detected, in the technical solution of the present application, the X-ray image of the rare earth permanent magnet material to be detected is segmented to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected. That is to say, image segmentation of the X-ray image of the rare earth permanent magnet material to be detected can divide an entire large X-ray image into small local area images, and more carefully analyze the characteristics of each X-ray local area, identify local detail information, so as to more accurately locate defects or abnormalities in the rare earth permanent magnet material to be detected, thereby more accurately evaluating the quality of the material, improving the precision and accuracy of material quality detection, helping to accurately evaluate the quality of the material and achieve precise quality control.

[0041] Specifically, in an embodiment of the present application, the deep-shallow fusion convolution unit 122 is used to input the sequence of X-ray local area images of the rare earth permanent magnet material to be detected into an image feature extractor based on a deep-shallow fusion convolution network model to obtain a sequence of shallow feature maps of the X-ray local area and a sequence of deep feature maps of the X-ray local area. It should be understood that, considering that each X-ray local area image of the rare earth permanent magnet material to be detected in the sequence of X-ray local area images of the rare earth permanent magnet material to be detected contains different degrees of deep and shallow feature information, for example, the shallow texture information in the X-ray local area image of the rare earth permanent magnet material to be detected, including particle size, shape, arrangement, structural information, including edges, contours, and geometric shapes, and the deep semantic information in the X-ray local area image of the rare earth permanent magnet material to be detected, including the meaning and relevance of different areas, these deep and shallow information have an important impact on the subsequent quality detection of the rare earth permanent magnet material. Therefore, in the technical solution of the present application, the sequence of X-ray local area images of the rare earth permanent magnet material to be detected is input into an image feature extractor based on a deep-shallow fusion convolutional network model to obtain a sequence of shallow feature maps of the X-ray local area and a sequence of deep feature maps of the X-ray local area. It is worth mentioning that the deep-shallow fusion convolutional network model combines the advantages of deep and shallow feature extraction in the image, and can more comprehensively capture the feature information of the X-ray local area images of each rare earth permanent magnet material to be detected, including local details and semantic information, which helps to improve the richness and characterization ability of features, improve the diversity and expression ability of X-ray local area shallow feature maps and X-ray local area deep feature maps, and help to more accurately describe the content and structure of the X-ray local area images of the rare earth permanent magnet material to be detected.

[0042] In an embodiment of the present application, the feature enhancement module 130 is used to perform feature enhancement on the sequence of X-ray local area deep feature maps to obtain a sequence of X-ray local area deep enhanced feature maps. Specifically, in an embodiment of the present application, the feature enhancement module is used to: input the sequence of X-ray local area deep feature maps into a feature enhancement model based on a spatial attention mechanism to obtain a sequence of X-ray local area deep enhanced feature maps. It should be understood that, considering that each position in each X-ray local area deep feature map in the sequence of X-ray local area deep feature maps has different feature information, therefore, in order to enhance the features of the key areas in each X-ray local area deep feature map and reduce the influence of irrelevant feature information, in the technical solution of the present application, the sequence of X-ray local area deep feature maps is input into a feature enhancement model based on a spatial attention mechanism to obtain a sequence of X-ray local area deep enhanced feature maps. That is, the spatial attention mechanism can help the model focus on specific important areas in the deep feature map of each X-ray local area, improve the perception of deep features of key X-ray local areas, and strengthen the representation of important areas in the deep feature map of X-ray local areas, so as to better capture the key information in the image, thereby providing more effective support and assistance for the quality analysis and detection of rare earth permanent magnet materials.

[0043] Figure 3 FIG is a block diagram of a feature enhancement module in an automatic control system for rare earth permanent magnet material processing according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 3 As shown, the feature enhancement module 130 includes: a local area convolutional coding unit 131, which is used to pass each X-ray local area deep feature map in the sequence of the X-ray local area deep feature map through the convolution coding part of the feature enhancement model based on the spatial attention mechanism to obtain a plurality of high-dimensional X-ray local area deep feature maps; a local area spatial attention unit 132, which is used to input each high-dimensional X-ray local area deep feature map in the high-dimensional X-ray local area deep feature map into the spatial attention part of the feature enhancement model based on the spatial attention mechanism to obtain a plurality of X-ray local area deep spatial attention maps; and a local area attention applying unit 133, which is used to respectively calculate the position point multiplication between each corresponding group of the multiple high-dimensional X-ray local area deep feature maps and the X-ray local area deep spatial attention maps to obtain a sequence of the X-ray local area deep enhanced feature maps.

[0044] In an embodiment of the present application, the feature dimensionality reduction module 140 is used to cascade and pool the sequence of the X-ray local area deep enhancement feature map and the sequence of the X-ray local area shallow feature map to obtain an X-ray local area deep enhancement cascade feature vector and an X-ray local area shallow cascade feature vector. Accordingly, considering that the sequence of the X-ray local area deep enhancement feature map and the sequence of the X-ray local area shallow feature map respectively represent feature information at different levels in the X-ray local area, including structure, texture, color, high-level semantic information and other aspects. Therefore, in order to comprehensively represent the true information in the two sequences, in the technical solution of the present application, it is first necessary to cascade the sequence of the X-ray local area deep enhancement feature map and the sequence of the X-ray local area shallow feature map respectively to obtain a richer and more comprehensive X-ray feature representation. Taking into account that the two feature maps obtained by cascading are of high dimension and inconvenient for subsequent calculation, it is necessary to pool the two feature maps obtained by cascading to obtain an X-ray local area deep enhancement cascade feature vector and an X-ray local area shallow cascade feature vector respectively.

[0045] In an embodiment of the present application, the feature fusion module 150 is used to perform feature distribution matching based on feature representation conversion on the X-ray local area deep enhancement cascade feature vector and the X-ray local area shallow cascade feature vector to obtain an X-ray fusion feature vector. It should be understood that the X-ray local area deep enhancement feature vector and the X-ray local area shallow feature vector respectively represent feature information at different levels in the X-ray local area, including structure, texture, color, high-level semantic information and other aspects. Therefore, in order to comprehensively represent the true information of the X-ray area of ​​the entire rare earth permanent magnet material, in the technical solution of the present application, it is necessary to fuse the X-ray local area deep enhancement cascade feature vector and the X-ray local area shallow cascade feature vector.

[0046] In particular, the deep enhanced cascade feature vector is obtained by applying a deep feature extraction model based on a spatial attention mechanism to a sequence of deep feature maps from local X-ray regions. This process emphasizes capturing complex patterns and subtle structures in the image, such as tiny defects within materials or uneven distribution of elements. Deep features typically contain higher-level semantic information, reflecting more complex visual patterns and associations between objects. In contrast, the shallow cascade feature vector is extracted from a sequence of shallow feature maps from local X-ray regions and primarily captures more basic visual features, such as edges and textures. These features are crucial for understanding the fundamental composition of the image, but they are at a lower level of abstraction, reflecting more local geometric and photometric properties. Due to the fundamental differences in representation between deep and shallow features—the former focuses on global structure and complex patterns, while the latter focuses on local details and simple patterns—even if they are derived from the same original X-ray image, the feature vectors generated after their respective processing pathways will exist in different feature spaces. This means that when trying to directly transfer knowledge from one domain (for example, a model trained only for shallow features) to another domain (i.e., a new environment that integrates shallow and deep features), it is difficult for the model to directly and effectively transfer its learned knowledge. Based on this, in the technical solution of the present application, feature distribution matching based on feature representation conversion is performed on the X-ray local area deep enhanced cascade feature vector and the X-ray local area shallow cascade feature vector to obtain an X-ray fusion feature vector.

[0047] Specifically, in the embodiment of the present application, the feature fusion module includes: a feature vector mapping unit for mapping the X-ray local area deep enhancement cascade feature vector and the X-ray local area shallow cascade feature vector to an intrinsic decomposition space to obtain the X-ray local area deep enhancement cascade eigenfeature vector and the X-ray local area shallow cascade eigenfeature vector; a feature inter-domain calibration unit for performing inter-domain feature calibration on the X-ray local area deep enhancement cascade eigenfeature vector and the X-ray local area shallow cascade eigenfeature vector to obtain the X-ray local area deep enhancement cascade modulated eigenfeature vector. An X-ray fusion fine-grained semantic information field construction unit is used to construct an X-ray fusion fine-grained semantic information field between the X-ray local area deep enhanced cascade modulated eigenvector and the X-ray local area shallow cascade modulated eigenvector; an X-ray fusion feature generation unit is used to fuse the X-ray local area deep enhanced cascade modulated eigenvector and the X-ray local area shallow cascade modulated eigenvector based on the X-ray fusion fine-grained semantic information field to obtain the X-ray fusion feature vector.

[0048] More specifically, in an embodiment of the present application, the X-ray fusion fine-grained semantic information field construction unit is used to: multiply the intrinsic feature vector after deep enhancement cascade modulation of the X-ray local area and the transposed vector of the intrinsic feature vector after shallow cascade modulation of the X-ray local area, and then divide the multiplication by the scale of the intrinsic feature vector after shallow cascade modulation of the X-ray local area to obtain an X-ray fusion fine-grained interaction matrix; and perform convolution encoding on the X-ray fusion fine-grained matrix to obtain the X-ray fusion fine-grained semantic information field.

[0049] More specifically, in an embodiment of the present application, the X-ray fusion feature generation unit is used to: multiply the X-ray fusion fine-grained semantic information field by the X-ray local area deep enhancement cascade modulation intrinsic feature vector to obtain the X-ray local area deep enhancement cascade feature vector after information field mapping; multiply the X-ray fusion fine-grained semantic information field by the X-ray local area shallow cascade modulation intrinsic feature vector to obtain the X-ray local area shallow cascade feature vector after information field mapping; and perform weighted fusion on the X-ray local area deep enhancement cascade feature vector after information field mapping and the X-ray local area shallow cascade feature vector after information field mapping to obtain the X-ray fusion feature vector.

[0050] In an embodiment of the present application, specifically, performing feature distribution matching based on feature representation conversion on the X-ray local area deep enhancement cascade feature vector and the X-ray local area shallow cascade feature vector to obtain an X-ray fusion feature vector includes: processing the X-ray local area deep enhancement cascade feature vector and the X-ray local area shallow cascade feature vector according to the following formula to obtain an X-ray fusion feature vector; wherein the formula is:

[0051]

[0052]

[0053]

[0054]

[0055]

[0056]

[0057]

[0058] v c =αv kt +βv st

[0059] Where V1 represents the X-ray local area deep enhancement cascade feature vector, V2 represents the X-ray local area shallow cascade feature vector, PCA(·) represents principal component analysis, U1 represents the sequence of X-ray local area deep enhancement cascade eigendecomposition vectors, Λ1 represents the X-ray local area deep enhancement cascade diagonal matrix, λ 11 and λ 1m denote the eigenvalues ​​of the first and mth positions of the X-ray local area deep enhancement cascade diagonal matrix, T denotes the transpose of the vector, concat{·;·} denotes the vector cascade, and v 11 、v 12 、v 1m They represent the first, second and mth eigenvectors of the sequence of deep enhancement cascade eigendecomposition vectors of the local X-ray region, V k represents the X-ray local area deep enhancement cascade eigenvector, U2 represents the sequence of X-ray local area shallow cascade eigendecomposition vectors, Λ2 represents the X-ray local area deep enhancement cascade diagonal matrix, λ 21 and λ 2m They represent the eigenvalues ​​of the first and mth positions of the shallow cascade diagonal matrix of the local X-ray region, v 21 、v 22 、v 2m They represent the first, second and mth eigenvectors of the sequence of shallow cascade eigendecomposition vectors of the local X-ray region, V s represents the shallow cascade eigenvector of the local X-ray region, v i represents the i-th eigenvalue of the deep enhancement cascade eigenvector of the local X-ray region, v j represents the jth eigenvalue of the shallow cascade eigenvector of the local X-ray region, represents the square of the vector norm, exp represents the natural exponential function, V′ k represents the intrinsic eigenvector after deep enhancement cascade modulation in the local X-ray region, V′ s represents the intrinsic eigenvector after shallow cascade modulation of the local X-ray region, Represents matrix multiplication, S represents the scale of the vector, M x represents the X-ray fusion fine-grained interaction matrix, Conv 3×3 (M x ) indicates M x Perform convolutional coding based on a 3×3 convolution kernel, Ω represents the X-ray fusion fine-grained semantic information field, v kt represents the deep enhancement cascade feature vector of the local X-ray region after information field mapping, v st represents the shallow cascade feature vector of the local X-ray region after information field mapping, α and β represent weighted hyperparameters, and v crepresents the X-ray fusion feature vector.

[0060] That is, in the technical solution of the present application, the feature distribution matching based on feature representation conversion is performed on the X-ray local area deep enhancement cascade feature vector and the X-ray local area shallow cascade feature vector. The process first maps the X-ray local area deep enhancement cascade feature vector and the X-ray local area shallow cascade feature vector to the eigendecomposition space to obtain the X-ray local area deep enhancement cascade eigenfeature vector and the X-ray local area shallow cascade eigenfeature vector. Those skilled in the art should know that the process of mapping to the eigendecomposition space is a combination of dimensionality reduction and feature extraction. In a specific example, the eigenvalue decomposition technology of principal component analysis (PCA) is used to find the eigendecomposition space, and the eigendecomposition space mapping is performed. The X-ray local area deep enhancement cascade eigenfeature vector and the X-ray local area shallow cascade eigenfeature vector after information field mapping not only represent the projection of the original eigenvector in the new coordinate system, but also reflect the main direction of change of the data, which helps to avoid the "dimensionality disaster", that is, the problem that the sample density drops sharply as the number of features increases.

[0061] Next, after mapping the eigenvector to the eigendecomposition space, it is necessary to perform inter-domain feature calibration on the X-ray local area deep enhanced cascade eigenvector and the X-ray local area shallow cascade eigenvector to adjust the source domain eigenvector to better match the data distribution of the target domain. This operation is intended to capture and compensate for the systematic deviation between the two domains to reduce the hidden disturbance of the class expression of the local feature distribution of the X-ray local area deep enhanced cascade eigenvector and the X-ray local area shallow cascade eigenvector to the overall class expression. After modulation, the X-ray local area deep enhanced cascade modulated eigenvector and the X-ray local area shallow cascade modulated eigenvector can retain the source domain characteristics while being more in line with the actual needs of the target domain, thereby enhancing the model's understanding and application of the target domain data.

[0062] After completing the scale modulation of the feature vector, an X-ray fusion fine-grained semantic information field is constructed between the X-ray local area deep enhanced cascade modulation eigenvector and the X-ray local area shallow cascade modulation eigenvector. The X-ray fusion fine-grained semantic information field is used to describe a more detailed relationship between feature vectors. The introduction of the X-ray fusion fine-grained semantic information field enables the model to go beyond simple point-to-point comparison and consider more complex patterns and contextual information, thereby enhancing the model's sensitivity to subtle differences to improve the quality of feature representation. Finally, the X-ray local area deep enhanced cascade modulation eigenvector and the X-ray local area shallow cascade modulation eigenvector are fused based on the X-ray fusion fine-grained semantic information field to obtain an X-ray fusion feature vector.

[0063] In this embodiment of the present application, the quality detection module 160 is configured to determine whether the rare earth permanent magnet material is of acceptable quality based on the X-ray fusion feature vector and control the early warning device to issue an early warning indication. Specifically, in this embodiment of the present application, the quality detection module includes: a quality acceptance judgment unit configured to input the X-ray fusion feature vector into a classifier to obtain a classification result, wherein the classification result indicates whether the rare earth permanent magnet material is of acceptable quality; and an early warning indication control unit configured to control the early warning device to issue an early warning indication in response to the rare earth permanent magnet material being of unacceptable quality. In other words, the quality of the rare earth permanent magnet material is determined to be acceptable by performing classification processing based on the X-ray fusion feature vector, and a response indicating that the rare earth permanent magnet material is of unacceptable quality is provided. This allows the identification of subtle defects or anomalies in the rare earth permanent magnet material, achieving high-precision material quality detection, reducing labor and time costs, mitigating quality risks in the production process, and significantly improving production efficiency and product quality stability.

[0064] In summary, the automatic control system 100 in the rare earth permanent magnet material processing process based on the embodiment of the present application is explained, which uses an X-ray image of the rare earth permanent magnet material to be detected collected by an X-ray fluorescence spectrometer and uses deep learning technology to perform image analysis and processing on the X-ray image of the rare earth permanent magnet material to be detected, thereby determining whether the quality of the rare earth permanent magnet material is qualified and controlling the early warning device to issue an early warning indication. In this way, tiny defects or abnormalities in rare earth permanent magnet materials can be identified, and high-precision detection of material quality can be achieved, which reduces labor costs and time costs, reduces quality risks in the production process, and greatly improves production efficiency and product quality stability.

[0065] Figure 4 FIG. 1 is a flow chart of an automatic control method for the processing of rare earth permanent magnet materials according to an embodiment of the present application. Figure 4As shown, the automatic control method in the rare earth permanent magnet material processing process according to the embodiment of the present application includes: S110, obtaining an X-ray image of the rare earth permanent magnet material to be detected collected by an X-ray fluorescence spectrometer; S120, performing image segmentation on the X-ray image of the rare earth permanent magnet material to be detected to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected, and then performing deep and shallow image feature extraction to obtain a sequence of X-ray local area shallow feature maps and a sequence of X-ray local area deep feature maps; S130, performing feature enhancement on the sequence of X-ray local area deep feature maps to obtain a sequence of X-ray local area deep enhanced feature maps. sequence; S140, respectively concatenating and pooling the sequence of the X-ray local area deep enhancement feature map and the sequence of the X-ray local area shallow feature map to obtain an X-ray local area deep enhancement cascade feature vector and an X-ray local area shallow cascade feature vector; S150, performing feature distribution matching based on feature representation conversion on the X-ray local area deep enhancement cascade feature vector and the X-ray local area shallow cascade feature vector to obtain an X-ray fusion feature vector; and, S160, determining whether the quality of the rare earth permanent magnet material is qualified based on the X-ray fusion feature vector, and controlling the early warning device to issue an early warning indication.

[0066] Here, those skilled in the art will appreciate that the specific operations of each step in the automatic control method for the processing of rare earth permanent magnet materials have been described in the above reference. Figures 1 to 3 The description of the automatic control system in the rare earth permanent magnet material processing process has been introduced in detail, and therefore, its repeated description will be omitted.

[0067] While various embodiments of the present disclosure have been described above, the above descriptions are illustrative, non-exhaustive, and not intended to be limiting of the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or improvements to existing technologies, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. An automatic control system for rare earth permanent magnet material processing, characterized in that: include: An X-ray image acquisition module for rare earth permanent magnet materials, used to acquire an X-ray image of the rare earth permanent magnet material to be detected acquired by an X-ray fluorescence spectrometer; A local area feature extraction module is used to perform image segmentation on the X-ray image of the rare earth permanent magnet material to be detected to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected, and then perform deep and shallow image feature extraction to obtain a sequence of X-ray local area shallow feature maps and a sequence of X-ray local area deep feature maps; A feature enhancement module, used for performing feature enhancement on the sequence of X-ray local area deep feature maps to obtain a sequence of X-ray local area deep enhanced feature maps; A feature dimension reduction module, used for cascading and pooling the sequence of the X-ray local area deep enhancement feature map and the sequence of the X-ray local area shallow feature map to obtain an X-ray local area deep enhancement cascade feature vector and an X-ray local area shallow cascade feature vector; A feature fusion module, used for performing feature distribution matching based on feature representation conversion on the X-ray local area deep enhanced cascade feature vector and the X-ray local area shallow cascade feature vector to obtain an X-ray fusion feature vector; The quality detection module is used to determine whether the quality of the rare earth permanent magnet material is qualified based on the X-ray fusion feature vector, and control the early warning device to issue an early warning indication.

2. The automatic control system in the rare earth permanent magnet material processing process according to claim 1 is characterized in that: The local area feature extraction module comprises: An image segmentation unit, used for segmenting the X-ray image of the rare earth permanent magnet material to be detected to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected; The deep-shallow fusion convolution unit is used to input the sequence of X-ray local area images of the rare earth permanent magnet material to be detected into an image feature extractor based on the deep-shallow fusion convolution network model to obtain a sequence of shallow feature maps of the X-ray local area and a sequence of deep feature maps of the X-ray local area.

3. The automatic control system in the rare earth permanent magnet material processing process according to claim 2 is characterized in that: The feature enhancement module is used to: input the sequence of deep feature maps of the local X-ray area into a feature enhancement model based on the spatial attention mechanism to obtain the sequence of deep enhanced feature maps of the local X-ray area.

4. The automatic control system in the rare earth permanent magnet material processing process according to claim 3 is characterized in that: The feature fusion module comprises: A feature vector mapping unit, used for mapping the X-ray local area deep enhancement cascade feature vector and the X-ray local area shallow cascade feature vector to an intrinsic decomposition space to obtain an X-ray local area deep enhancement cascade intrinsic feature vector and an X-ray local area shallow cascade intrinsic feature vector; A feature inter-domain calibration unit, used for performing inter-domain feature calibration on the X-ray local area deep enhancement cascade intrinsic feature vector and the X-ray local area shallow cascade intrinsic feature vector to obtain the X-ray local area deep enhancement cascade modulated intrinsic feature vector and the X-ray local area shallow cascade modulated intrinsic feature vector; An X-ray fusion fine-grained semantic information field construction unit is used to construct an X-ray fusion fine-grained semantic information field between the X-ray local area deep enhanced cascade modulation intrinsic feature vector and the X-ray local area shallow cascade modulation intrinsic feature vector; An X-ray fusion feature generation unit is used to fuse the X-ray local area deep enhanced cascade modulated intrinsic feature vector and the X-ray local area shallow cascade modulated intrinsic feature vector based on the X-ray fusion fine-grained semantic information field to obtain the X-ray fusion feature vector.

5. The automatic control system in the rare earth permanent magnet material processing process according to claim 4 is characterized in that: The X-ray fusion fine-grained semantic information field construction unit is used to: The X-ray local area deep enhanced cascade modulation intrinsic feature vector and the X-ray local area shallow cascade modulation intrinsic feature vector after multiplying, and then divided by the scale of the X-ray local area shallow cascade modulation intrinsic feature vector to obtain an X-ray fusion fine-grained interaction matrix; The X-ray fusion fine-grained matrix is ​​convolutionally encoded to obtain the X-ray fusion fine-grained semantic information field.

6. The automatic control system in the rare earth permanent magnet material processing process according to claim 5, characterized in that: The X-ray fusion feature generation unit is used to: Multiplying the X-ray fusion fine-grained semantic information field by the X-ray local area deep enhancement cascade modulated intrinsic feature vector to obtain the X-ray local area deep enhancement cascade feature vector after information field mapping; Multiplying the X-ray fusion fine-grained semantic information field by the X-ray local area shallow cascade modulated intrinsic feature vector to obtain the X-ray local area shallow cascade feature vector after information field mapping; The deep enhanced cascade feature vector of the X-ray local area after the information field mapping and the shallow cascade feature vector of the X-ray local area after the information field mapping are weightedly fused to obtain the X-ray fusion feature vector.

7. The automatic control system in the rare earth permanent magnet material processing process according to claim 6 is characterized in that: The quality detection module comprises: A quality judgment unit, used for inputting the X-ray fusion feature vector into a classifier to obtain a classification result, wherein the classification result is used to indicate whether the quality of the rare earth permanent magnet material is qualified; The early warning indication control unit is used to control the early warning device to issue an early warning indication in response to the rare earth permanent magnet material being of unqualified quality.

8. An automatic control method in the processing of rare earth permanent magnet materials, characterized in that: include: Obtaining an X-ray image of the rare earth permanent magnet material to be tested collected by an X-ray fluorescence spectrometer; Segmenting the X-ray image of the rare earth permanent magnet material to be detected to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected, and then extracting shallow and deep image features to obtain a sequence of shallow feature maps of the X-ray local area and a sequence of deep feature maps of the X-ray local area; Performing feature enhancement on the sequence of X-ray local area deep feature maps to obtain a sequence of X-ray local area deep enhanced feature maps; Respectively cascading and pooling the sequence of the X-ray local area deep enhancement feature map and the sequence of the X-ray local area shallow feature map to obtain an X-ray local area deep enhancement cascade feature vector and an X-ray local area shallow cascade feature vector; Performing feature distribution matching based on feature representation conversion on the X-ray local area deep enhancement cascade feature vector and the X-ray local area shallow cascade feature vector to obtain an X-ray fusion feature vector; Based on the X-ray fusion feature vector, it is determined whether the quality of the rare earth permanent magnet material is qualified, and the early warning device is controlled to issue an early warning indication.

9. The automatic control method in the rare earth permanent magnet material processing process according to claim 8, characterized in that: The method comprises: performing image segmentation on the X-ray image of the rare earth permanent magnet material to be detected to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected, and then performing deep and shallow image feature extraction to obtain a sequence of X-ray local area shallow feature maps and a sequence of X-ray local area deep feature maps, including: Segmenting the X-ray image of the rare earth permanent magnet material to be detected to obtain a sequence of X-ray local area images of the rare earth permanent magnet material to be detected; The sequence of X-ray local area images of the rare earth permanent magnet material to be detected is input into an image feature extractor based on a deep and shallow fusion convolutional network model to obtain a sequence of shallow feature maps of the X-ray local area and a sequence of deep feature maps of the X-ray local area.

10. The automatic control method in the rare earth permanent magnet material processing process according to claim 9, characterized in that: Performing feature enhancement on the sequence of X-ray local area deep feature maps to obtain a sequence of X-ray local area deep enhanced feature maps, comprising: The sequence of the X-ray local area deep feature maps is input into a feature enhancement model based on a spatial attention mechanism to obtain a sequence of the X-ray local area deep enhanced feature maps.

Images