Material detection method, electronic equipment and storage medium

Abstract

The embodiment of the invention discloses a material detection method, electronic equipment and a storage medium, and relates to the technical field of spectrum detection.The method comprises the steps that spectrum sequence data corresponding to a to-be-detected composite material is obtained; performing feature analysis on the spectrum sequence data by using a pre-trained neural network model to obtain an analysis result, the pre-trained neural network model being obtained by training a training sample, and the training sample comprising historical spectrum sequence data and a corresponding historical analysis result; and determining the components of the to-be-detected composite material according to the analysis result. According to the embodiment of the invention, feature analysis is carried out on the spectral sequence data through the pre-trained neural network model, so that the accuracy of determining the composite material components can be improved.

Description

Technical Field

[0001] This application relates to the field of spectral detection technology, and in particular to a material detection method, electronic device and storage medium. Background Technology

[0002] With the continuous development of new materials technology, composite materials are increasingly favored by various industries due to their excellent comprehensive properties. Composite materials are usually composed of at least two components with significantly different properties. The final performance of the composite material depends not only on the characteristics of each component itself, but also on the ratio of the components, their spatial distribution, and the quality of the interfacial bonding.

[0003] Currently, composite materials are widely used in many key fields such as aerospace, automobile manufacturing, and communication equipment, making the accurate identification of their internal components increasingly important. Among related testing technologies, spectroscopic detection technology is a commonly used method for identifying the components of composite materials. However, in practical applications, this technology still suffers from low accuracy and requires further improvement and refinement.

[0004] Application content

[0005] In view of this, one of the objectives of this application is to provide a material testing method, electronic device and storage medium that can improve the accuracy of determining the composition of composite materials.

[0006] To achieve the above objectives, the technical solution of this application is implemented as follows:

[0007] In a first aspect, embodiments of this application provide a material testing method, which includes:

[0008] Obtain the spectral sequence data corresponding to the composite material to be tested;

[0009] The pre-trained neural network model is used to perform feature analysis on spectral sequence data to obtain analysis results. The pre-trained neural network model is trained from training samples, which include historical spectral sequence data and their corresponding historical analysis results.

[0010] Based on the analysis results, the components of the composite material to be tested were determined.

[0011] In one possible implementation, the pre-trained neural network model includes convolutional layers, pooling layers, and fully connected layers, with the output of the convolutional layer connected to the input of the pooling layer, and the output of the pooling layer connected to the input of the fully connected layer.

[0012] Feature analysis of spectral sequence data is performed using a pre-trained neural network model to obtain the analysis results, including:

[0013] The first feature data is obtained by using a convolutional layer to extract features from the spectral sequence data.

[0014] The first feature data is downsampled using a pooling layer to obtain the second feature data;

[0015] The second feature data is transformed using a fully connected layer to obtain the analysis results.

[0016] In one possible implementation, the convolutional layer includes a first convolutional layer, a second convolutional layer, and a third convolutional layer, wherein the output of the first convolutional layer is connected to the input of the second convolutional layer, and the output of the second convolutional layer is connected to the input of the third convolutional layer.

[0017] Feature extraction is performed on the spectral sequence data using convolutional layers to obtain the first feature data, including:

[0018] The first convolutional layer is used to extract features from the spectral sequence data to obtain the first sub-feature data;

[0019] The second convolutional layer is used to extract features from the first sub-feature data to obtain the second sub-feature data;

[0020] The third sub-feature data is obtained by extracting features from the second sub-feature data using the third convolutional layer. The first feature data includes the third sub-feature data.

[0021] In one possible implementation, the number of features in the first sub-feature data, the second sub-feature data, and the third sub-feature data increases sequentially, and the sequence lengths of the first sub-feature data, the second sub-feature data, and the third sub-feature data decrease sequentially.

[0022] In one possible implementation, the first convolutional layer, the second convolutional layer, and the third convolutional layer each include a subconvolutional layer, a modified linear unit function layer, and a max pooling layer. The output of the subconvolutional layer is connected to the input of the modified linear unit function layer, and the output of the modified linear unit function layer is connected to the input of the max pooling layer.

[0023] In one possible implementation, the fully connected layer includes a first fully connected layer and a second fully connected layer, with the output of the first fully connected layer connected to the input of the second fully connected layer.

[0024] The second feature data is transformed using a fully connected layer to obtain the analysis results, including:

[0025] The second feature data is nonlinearly combined using the first fully connected layer to obtain combined feature data;

[0026] The combined feature data is classified and mapped using the second fully connected layer to obtain the analysis results.

[0027] In one possible implementation, acquiring the spectral sequence data corresponding to the composite material to be detected includes:

[0028] The composite material to be tested is irradiated with a first parallel ray to obtain transmitted light, wherein the first parallel ray is formed by the convergence of broadband light rays, and the broadband light rays are generated by a broadband light source;

[0029] The transmitted light is calibrated to obtain the second parallel light ray;

[0030] The second parallel light rays are collected to form spectral sequence data.

[0031] In one possible implementation, the composite material to be tested is irradiated with a first parallel light beam, including:

[0032] The angle of the first parallel light ray is adjusted so that the adjusted first parallel light ray irradiates the composite material to be tested perpendicularly.

[0033] Secondly, embodiments of this application provide a material testing system, which includes:

[0034] The acquisition module is used to acquire the spectral sequence data corresponding to the composite material to be tested;

[0035] The analysis module is used to perform feature analysis on spectral sequence data using a pre-trained neural network model to obtain analysis results. The pre-trained neural network model is trained from training samples, which include historical spectral sequence data and their corresponding historical analysis results.

[0036] The determination module is used to determine the composition of the composite material to be tested based on the analysis results.

[0037] Thirdly, embodiments of this application provide an electronic device, which includes a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, it implements the material detection method provided in the first aspect.

[0038] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by one or more processors, implements the material detection method provided in the first aspect.

[0039] The material testing method provided in this application acquires the spectral sequence data corresponding to the composite material to be tested. Then, a pre-trained neural network model is used to perform feature analysis on the spectral sequence data to obtain analysis results. The pre-trained neural network model is trained using training samples, including historical spectral sequence data and their corresponding historical analysis results. Finally, the composition of the composite material to be tested can be determined based on the analysis results. This application's method, by using a pre-trained neural network model to perform feature analysis on the spectral sequence data, can improve the accuracy of determining the composition of composite materials. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. It should be understood that the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 A flowchart illustrating a material testing method provided in this application embodiment;

[0042] Figure 2 A schematic diagram of the neural network architecture included in a material detection method provided in an embodiment of this application;

[0043] Figure 3 A schematic diagram of the acquisition device involved in a material testing method provided in an embodiment of this application;

[0044] Figure 4 This is a schematic diagram of the functional modules of a material testing system provided in an embodiment of this application;

[0045] Figure 5 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application.

[0046] Explanation of reference numerals in the attached figures:

[0047] 300. Data acquisition device;

[0048] 310. Broad-spectrum light source;

[0049] 320. Spectrometer;

[0050] 330. First collimator;

[0051] 340. Second collimator;

[0052] 350. Mirror module;

[0053] 400. Material testing system;

[0054] 410. Acquisition Module;

[0055] 420. Analysis Module;

[0056] 430. Determine the module;

[0057] 501. Processor;

[0058] 502. Memory;

[0059] 503. Communication interface;

[0060] 510. Bus. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0062] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0063] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0064] In various embodiments of this application, the expression "or" or "at least one of A and / or B" includes any combination or all combinations of the words listed simultaneously. For example, the expression "A or B" or "at least one of A and / or B" may include A, may include B, or may include both A and B.

[0065] In the description of this application, it should be noted that if terms such as "upper," "lower," "inner," or "outer" are used to indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of the invention is usually placed during use, they are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0066] Furthermore, the terms "first" and "second" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0067] It should be noted that, where there is no conflict, the features in the embodiments of this application can be combined with each other.

[0068] Furthermore, in the embodiments of this application, the term "connection" can refer to "electrical connection" or "direct connection." "Electrical connection" can refer to a direct electrical connection between two components, or it can refer to an electrical connection between two components via one or more normally open tubes or other components.

[0069] To facilitate a better understanding of the solutions in the embodiments of this application, the relevant technologies will be introduced first below.

[0070] Transmittance spectroscopy is a technique that measures the transmission behavior of a material to light within a specific wavelength range, reflecting its internal chemical bonding structure, component concentration, and microscopic physical characteristics. When multiple materials are mixed to form a composite structure, light propagating within it is affected by the different absorption and scattering effects of each component, resulting in a series of complex characteristics in its transmittance curve. These characteristic curves exhibit absorption peaks, transmission valleys, or continuous variations within a specific wavelength range, containing rich information about the composition of the composite material. Therefore, non-destructive identification technology based on transmittance spectroscopy demonstrates unique advantages in meeting the dual requirements of efficiency and accuracy in industrial-grade testing. By irradiating a sample of the material to be tested (such as the composite material to be tested in the embodiments of this application) with light and acquiring its transmittance spectrum at a specific wavelength, characteristic curves containing information about the material composition and structure can be obtained. This method is not only applicable to transparent or translucent materials, but can also be used under certain conditions for component identification of optically complex structures such as thin film coatings and multilayer composite films. Due to its non-destructive nature, high sensitivity, and ability to characterize the internal structure of samples, transmittance spectroscopy has become one of the important methods in composite material testing.

[0071] Spectroscopic detection technology, as a non-contact and non-destructive material identification method, has received widespread attention in the field of composite material testing. Transmittance spectroscopy, in particular, can reflect key information such as the internal structure and molecular composition of materials by detecting their transmission characteristics to light in specific wavelengths. When different materials are combined to form composite materials, the transmittance spectrum curve of the composite material will reflect the optical response characteristics of each component.

[0072] The Rectified Linear Unit (ReLU) is one of the most commonly used activation functions in deep neural networks.

[0073] Pooling layers are algorithms used in convolutional neural networks to downsample (reduce dimensionality) feature maps, thereby reducing computational cost and the number of parameters, while enhancing the model's invariance to feature translation and local deformation.

[0074] A fully connected layer is the most basic layer structure in a neural network, in which each neuron is connected to all neurons in the layer above it.

[0075] Robustness refers to the ability of a system or algorithm to maintain stability and output accuracy when faced with input disturbances, noise interference, changes in model parameters, or non-ideal conditions.

[0076] The Sigmoid activation function is a commonly used non-linear function in convolutional neural networks, typically used in the output layer.

[0077] The binary cross-entropy loss function is a commonly used loss function in binary or multi-label classification tasks, used to measure the difference between the predicted probability and the true label.

[0078] To address the technical problems in the background art, embodiments of this application provide a material testing method, a material testing system, an electronic device, a computer-readable storage medium, and a computer program product. The material testing method provided in the embodiments of this application will be described first.

[0079] Please see Figure 1 , Figure 1 This is a flowchart of a material testing method provided in an embodiment of this application. The material testing method can be applied to the material testing system or electronic device in the following embodiments. The electronic device includes personal computers, servers, mobile devices, cloud computing platforms, and supercomputers, etc. The following will describe the material testing method from the perspective of its application in electronic devices.

[0080] like Figure 1 The material testing method shown specifically includes the following steps 110 to 130.

[0081] Step 110: Obtain the spectral sequence data corresponding to the composite material to be tested.

[0082] Step 120: Use a pre-trained neural network model to perform feature analysis on the spectral sequence data and obtain the analysis results. The pre-trained neural network model is trained from training samples, which include historical spectral sequence data and their corresponding historical analysis results.

[0083] Step 130: Based on the analysis results, determine the components of the composite material to be tested.

[0084] The material testing method provided in this application acquires the spectral sequence data corresponding to the composite material to be tested. Then, a pre-trained neural network model is used to perform feature analysis on the spectral sequence data to obtain analysis results. The pre-trained neural network model is trained using training samples, including historical spectral sequence data and their corresponding historical analysis results. Finally, the composition of the composite material to be tested can be determined based on the analysis results. This application's method, by using a pre-trained neural network model to perform feature analysis on the spectral sequence data, can improve the accuracy of determining the composition of composite materials.

[0085] The following will discuss how Figure 1 The steps of the Chinese method are explained in detail.

[0086] In step 110, the electronic device can acquire the spectral sequence data corresponding to the composite material to be tested, providing a data basis for subsequent analysis to determine the composition of the composite material to be tested.

[0087] The composite materials to be tested are composed of at least two different materials and require component analysis. For example, in the aerospace and automotive manufacturing fields, composite materials include, but are not limited to, glass fiber reinforced plastics, carbon fiber reinforced composites, aramid fiber composites, metal matrix composites, ceramic matrix composites, and aerogel composites.

[0088] The aforementioned spectral sequence data can be used to reflect the transmittance of light irradiated by different wavelengths onto the composite material under test. For example, the spectral sequence data can be expressed as T(λ)=[T(λ1),T(λ2),...,T(λ)]. n )], T(λ1) represents the transmittance of the light corresponding to the light with wavelength λ1 illuminating the composite material under test.

[0089] In some embodiments, the composite material to be tested can be placed on a corresponding system sample stage, and then the system sample stage can be controlled by an electronic device to generate corresponding light to irradiate the composite material to be tested. The light irradiating the composite material to be tested and passing through the composite material can be collected by a corresponding acquisition device, and corresponding spectral sequence data can be generated.

[0090] In step 120, the electronic device can use a pre-trained neural network model to perform feature analysis on the spectral sequence data to obtain accurate analysis results, and the composition of the composite material to be tested can be determined based on the analysis results.

[0091] In some embodiments, the pre-trained neural network model described above includes a pre-trained one-dimensional convolutional neural network (1D CNN).

[0092] In some embodiments, before acquiring the spectral sequence data corresponding to the composite material to be detected, the electronic device may train a pre-trained neural network model based on training samples.

[0093] The training samples mentioned above include historical spectral sequence data and the corresponding historical analysis results.

[0094] In some embodiments, the electronic device may train a pre-built initial neural network model based on training samples, and stop training when a preset stopping condition is met, thereby obtaining the pre-trained neural network model.

[0095] In some embodiments, the preset notification conditions described above include any one of the following:

[0096] The number of training attempts exceeds the preset training attempt threshold;

[0097] The training duration exceeds the preset training duration threshold;

[0098] In the pre-built initial network model, the loss function decreases less than the preset value over 10 consecutive training iterations.

[0099] The historical analysis results mentioned above may include the historical component labels corresponding to the historical spectral sequence data, as well as the probability values ​​corresponding to the historical component labels.

[0100] Accordingly, the above analysis results may include the component labels corresponding to the above spectral sequence data and the probability values ​​corresponding to the component labels. The component labels corresponding to the above spectral sequence data can also be understood as the component labels corresponding to the above composite material to be tested.

[0101] Different component labels can be used to characterize different materials.

[0102] For example, the component labels include 001, 002, and 003, corresponding to three different materials.

[0103] In some embodiments, the component labels described above are unique.

[0104] It should be noted that the above spectral sequence data is one-dimensional vector data. Using a pre-trained one-dimensional neural network model, such as 1D CNN, to process the spectral sequence data can greatly reduce the number of parameters that need to be learned and stored compared to processing data using a two-dimensional neural network model, thereby reducing the computational load and memory requirements of the model.

[0105] In step 130, the electronic device can determine the composition of the composite material to be tested based on the analysis results.

[0106] In some embodiments, the analysis results include component labels and corresponding probability values ​​for the component labels. The determination of the components of the composite material to be tested based on the analysis results includes:

[0107] The components of the composite material to be tested are determined based on the preset probability threshold, component labels, and the probability values ​​corresponding to the component labels.

[0108] In some embodiments, the components of the composite material to be tested are determined based on a preset probability threshold, component labels, and the probability values ​​corresponding to the component labels, including:

[0109] The material corresponding to the target component label is determined as the component of the composite material to be tested. The target component label is the component label corresponding to the probability value greater than the preset probability value threshold.

[0110] For example, if the component labels include 001, 002, and 003, with corresponding probability values ​​of 0.6, 0.3, and 0.05 respectively, and the preset probability threshold is 0.1, then the electronic device can identify component labels 001 and 002 as target component labels, and the materials indicated by component labels 001 and 002 can be identified as components of the composite material to be tested.

[0111] Although the preset probability threshold is shown as 0.1 for illustrative purposes, more values ​​can be selected as the preset probability threshold according to actual needs, all of which are within the protection scope of the embodiments of this application.

[0112] Please see Figure 2 , Figure 2 This is a schematic diagram of the neural network architecture included in a material detection method provided in an embodiment of this application.

[0113] exist Figure 2 In this process, the input data to the convolutional layer is the transmittance curve, which is the spectral sequence data corresponding to the composite material to be tested. The "prediction" output by the fully connected layer represents the above analysis result. "S" represents the composite material to be tested.

[0114] In one possible implementation, the pre-trained neural network model includes convolutional layers, pooling layers, and fully connected layers, with the output of the convolutional layer connected to the input of the pooling layer, and the output of the pooling layer connected to the input of the fully connected layer.

[0115] Feature analysis of spectral sequence data is performed using a pre-trained neural network model to obtain the analysis results, including:

[0116] The first feature data is obtained by using a convolutional layer to extract features from the spectral sequence data.

[0117] The first feature data is downsampled using a pooling layer to obtain the second feature data;

[0118] The second feature data is transformed using a fully connected layer to obtain the analysis results.

[0119] In this embodiment, the pre-trained neural network model is the aforementioned 1D CNN model.

[0120] The pre-trained neural network model in this embodiment, namely the 1D CNN model, includes convolutional layers, pooling layers, and fully connected layers. It can be understood that the three work together to construct a highly compressed and interference-resistant data processing chain, which can improve the accuracy of determining the components of the composite material to be detected, while also improving the efficiency of determining the components of the composite material to be detected.

[0121] The above convolutional layer is a one-dimensional convolutional layer.

[0122] The first feature data mentioned above is the local feature data obtained after feature extraction.

[0123] The aforementioned second feature data is the key feature data obtained by downsampling and dimensionality reduction of the first feature data. This not only reduces the amount of feature data and alleviates the computational pressure on the fully connected layer, but also ensures the accuracy and reliability of the analysis results.

[0124] The above analysis results can be found in the description of the foregoing embodiments, and will not be repeated here.

[0125] The output of the fully connected layer can be the output of the 1D CNN model.

[0126] In some embodiments, the pooling layer described above is a global adaptive average pooling layer.

[0127] In one possible implementation, the first convolutional layer, the second convolutional layer, and the third convolutional layer each include a subconvolutional layer, a modified linear unit function layer, and a max pooling layer. The output of the subconvolutional layer is connected to the input of the modified linear unit function layer, and the output of the modified linear unit function layer is connected to the input of the max pooling layer.

[0128] The first, second, and third convolutional layers mentioned above all include sub-convolutional layers, modified linear unit function layers, and max pooling layers, which can be understood as follows:

[0129] The first convolutional layer includes a first sub-convolutional layer, a first rectified linear unit (ReLU) function layer, and a first max pooling layer.

[0130] The second convolutional layer consists of a second sub-convolutional layer, a second modified linear unit function layer, and a second max pooling layer.

[0131] The third convolutional layer consists of a third sub-convolutional layer, a third modified linear unit function layer, and a third max pooling layer.

[0132] In some embodiments:

[0133] The first sub-convolutional layer has 1 input channel, 16 output channels, and a kernel length of 3.

[0134] The second sub-convolutional layer has 1 input channel, 32 output channels, and a kernel length of 3.

[0135] The third sub-convolutional layer has 1 input channel, 64 output channels, and a kernel length of 3.

[0136] Taking the first sub-convolutional layer as an example, the electronic device can use the first sub-convolutional layer to extract features from spectral sequence data. There are 16 output channels, and each channel can correspond to the feature data extracted by a convolutional kernel.

[0137] In some embodiments:

[0138] The first ReLU function layer, the second ReLU function layer, and the third ReLU function layer mentioned above are the same.

[0139] Taking the first ReLU function layer as an example, the first ReLU function layer can perform nonlinear transformation on the feature data extracted by the first sub-convolutional layer.

[0140] In some embodiments:

[0141] The first max pooling layer, the second max pooling layer, and the third max pooling layer are the same.

[0142] Taking the first pooling layer as an example, the first pooling layer can compress data that has undergone nonlinear transformation by the first ReLU function layer.

[0143] In some embodiments, after the data from the nonlinear transformation performed by the first ReLU function layer is compressed, the first pooling layer can output sequence data with the corresponding sequence length halved.

[0144] In composite material scenarios, the optical responses of each component often overlap and interfere with each other on the spectral curves, blurring the characteristic boundaries between different materials and posing a challenge to subsequent classification tasks. Especially in multi-component systems, even small amounts of dopants can significantly affect the overall transmittance characteristics, further increasing the complexity of identification.

[0145] In one possible implementation, the convolutional layer includes a first convolutional layer, a second convolutional layer, and a third convolutional layer, wherein the output of the first convolutional layer is connected to the input of the second convolutional layer, and the output of the second convolutional layer is connected to the input of the third convolutional layer.

[0146] Feature extraction is performed on the spectral sequence data using convolutional layers to obtain the first feature data, including:

[0147] The first convolutional layer is used to extract features from the spectral sequence data to obtain the first sub-feature data;

[0148] The second convolutional layer is used to extract features from the first sub-feature data to obtain the second sub-feature data;

[0149] The third sub-feature data is obtained by extracting features from the second sub-feature data using the third convolutional layer. The first feature data includes the third sub-feature data.

[0150] This application embodiment extracts local features by setting three one-dimensional convolutional layers, which can realize advanced semantic feature extraction to identify deep spectral features that are strongly correlated with specific material categories. This can avoid the situation where the optical responses of each component in composite material scenarios overlap and interfere with each other on the spectral curves, resulting in blurred feature boundaries of different materials, thereby improving the overall recognition accuracy and generalization ability.

[0151] The data processing procedure for each convolutional layer, including its sub-convolutional layers, ReLU function layers, and max pooling layers, as described in the foregoing embodiments, is as follows:

[0152] The electronic device inputs spectral sequence data into the first sub-convolutional layer. This first sub-convolutional layer uses 16 convolutional kernels of length 3 to maintain the sequence length of the input spectral sequence data, extracting local fluctuations and edge features. This highlights the absorption or transmission changes of the material at specific wavelengths, aiding in subsequent structure identification. The first sub-convolutional layer then enhances its nonlinear expression capability through a first ReLU function layer and halves the sequence length through a first max-pooling layer, obtaining the aforementioned first sub-feature data. This achieves feature compression and noise reduction, thereby improving the network's robustness to local patterns.

[0153] The electronic device inputs the first sub-feature data into the second sub-convolutional layer, which further deepens the feature abstraction level based on the first sub-convolutional layer. The second sub-convolutional layer uses 32 output channels, with the same kernel length of 3, maintaining the same receptive field setting, and continues to perform nonlinear transformation and size compression through a second ReLU function layer and a second max-pooling layer. The second sub-convolutional layer is mainly used to capture mid-level features at a wider scale, such as the correlation between multiple feature points. The second sub-convolutional layer enhances the learning ability of the pre-trained neural network model, i.e., the 1D CNN model, for cross-band feature coupling, which is beneficial for identifying hidden material information in complex spectra, thereby improving the accuracy of determining the components in the composite material to be detected. The output of the second convolutional layer is the aforementioned second sub-feature data.

[0154] The electronic device inputs the second sub-feature data into the third sub-convolutional layer. This third sub-convolutional layer further expands the feature data mapping dimension to 64 channels, modeling potential complex variation patterns in the spectrum in a high-dimensional space. Combined with a third max-pooling layer, the sequence length corresponding to the second sub-feature data is further halved, achieving the final high-level semantic feature extraction. At this stage, the pre-trained neural network model, i.e., the 1DCNN model, can identify deep-level spectral features strongly correlated with specific material categories, thereby improving the overall accuracy and generalization ability of the identification.

[0155] Through three layers of convolution and pooling operations, the spectral sequence data corresponding to the composite material to be detected is mapped into a high-dimensional, low-temporal-resolution feature tensor, providing a rich and compressed representation basis for the classification network.

[0156] In one possible implementation, the number of features in the first sub-feature data, the second sub-feature data, and the third sub-feature data increases sequentially, and the sequence lengths of the first sub-feature data, the second sub-feature data, and the third sub-feature data decrease sequentially.

[0157] The description process of this embodiment is similar to that of the foregoing embodiments, and will not be repeated here.

[0158] In one possible implementation, the fully connected layer includes a first fully connected layer and a second fully connected layer, with the output of the first fully connected layer connected to the input of the second fully connected layer.

[0159] The second feature data is transformed using a fully connected layer to obtain the analysis results, including:

[0160] The second feature data is nonlinearly combined using the first fully connected layer to obtain combined feature data;

[0161] The combined feature data is classified and mapped using the second fully connected layer to obtain the analysis results.

[0162] The aforementioned second fully connected layer can be regarded as the data output layer of a pre-trained neural network model, i.e., a 1D CNN model.

[0163] In some embodiments, the second fully connected layer can use the Sigmoid activation function, combined with a binary cross-entropy loss function, to independently model and predict each class. This allows for the simultaneous identification of multiple material features, adapting to complex and non-mutually exclusive material combinations in real-world scenarios.

[0164] In one possible implementation, acquiring the spectral sequence data corresponding to the composite material to be detected includes:

[0165] The composite material to be tested is irradiated with a first parallel ray to obtain transmitted light. The first parallel ray is formed by the convergence of broadband light rays, which are generated by a broadband light source 310.

[0166] The transmitted light is calibrated to obtain the second parallel light ray;

[0167] The second parallel light rays are collected to form spectral sequence data.

[0168] The aforementioned broadband light source 310 can be found in [reference needed]. Figure 3 .

[0169] This application embodiment improves the efficiency of transmitting light collection by using a first parallel light formed by the convergence of broadband light to irradiate the composite material to be tested, and by calibrating the corresponding transmitted light and collecting the calibrated second parallel light. This improves the efficiency of acquiring spectral sequence data.

[0170] To further improve the accuracy of component detection in the composite material under test, the spectral sequence data corresponding to the composite material needs to cover a wide wavelength range, thereby enabling the extraction of rich spectral feature information. In some embodiments, the broadband light source 310 can be selected as near-infrared light as the excitation source, and its operating wavelength range can be set to 1100 nm to 1700 nm to have a wide spectral coverage capability. The light in the wavelength range exemplified in this embodiment can penetrate the surface layer of the composite material under test, and can obtain feature information of its internal structure, especially showing a strong response to vibrational modes of covalent bonds in the material.

[0171] It is important to note that, compared to visible light, near-infrared light is safer for the human eye in spectral measurements and is suitable for long-term stable measurements. Furthermore, near-infrared light has lower sensitivity to the surface color of the composite material being tested, and can more accurately reflect the actual optical properties of the material, making it suitable for composite materials with significant color differences.

[0172] In one possible implementation, the composite material to be tested is irradiated with a first parallel light beam, including:

[0173] The angle of the first parallel light ray is adjusted so that the adjusted first parallel light ray irradiates the composite material to be tested perpendicularly.

[0174] This embodiment of the application avoids dispersion interference caused by non-perpendicular illumination by illuminating the composite material under test with the adjusted first parallel light perpendicularly, thereby improving the signal-to-noise ratio of the collected spectrum and increasing the overall measurement accuracy and robustness.

[0175] To achieve the above-mentioned calibration of transmitted light and convergence of broadband light to form the first parallel ray, please refer to [link to relevant documentation]. Figure 3 , Figure 3 This is a schematic diagram of a collection device involved in a material testing method provided in an embodiment of this application.

[0176] In some embodiments, the first parallel light ray may be formed by converging broadband light rays through a first collimator 330.

[0177] In some embodiments, the above-described calibration of the transmitted light to obtain a second parallel light ray includes:

[0178] The transmitted light is calibrated by the second collimator 340 to obtain the second parallel light.

[0179] In some embodiments, the above-described acquisition of the second parallel light rays to form spectral sequence data includes:

[0180] The second parallel light rays are collected by the spectrometer 320, and spectral sequence data are formed in the spectrometer 320.

[0181] In some embodiments, the above-described angle adjustment of the first parallel light beam to ensure that the adjusted first parallel light beam perpendicularly illuminates the composite material to be tested includes:

[0182] The angle of the first parallel light is adjusted by the reflector module 350 so that the adjusted first parallel light shines perpendicularly on the composite material to be tested.

[0183] exist Figure 3 In the acquisition device 300 shown, a broadband light source 310 is connected to a first collimator 330, and a spectrometer 320 is connected to a second collimator 340. Furthermore, electronic devices can be connected to the spectrometer 320 and the broadband light source 310 respectively to control the broadband light source 310 to generate light of corresponding wavelengths, or to acquire the spectral sequence data corresponding to the composite material to be tested from the spectrometer 320. It should be noted that... Figure 3 The connection between the electronic equipment and the spectrometer 320 and the broadband light source 310 is not shown in the figure.

[0184] Specifically, the broadband light source 310 generates broadband light, which is initially collimated by the first collimator 330 to form first parallel light, thereby ensuring good spatial consistency of the broadband light. Then, the illumination angle of the first parallel light is adjusted by the reflector module 350, so that the first parallel light passing through the reflector module 350 can perpendicularly illuminate the composite material to be tested.

[0185] Still based on Figure 3 To further eliminate system-level interference, in some embodiments of the illustrated device, electronic equipment can control the device to perform initialization measurement operations. Specifically, in the absence of the composite material to be detected, the broadband light source 310 directly passes through, as shown... Figure 3 The spectral sequence data following the optical path shown are used to establish a system baseline response curve. This eliminates systematic errors, including light source stability and specular reflection errors, resulting in more accurate spectral sequence data. Furthermore, based on the operations described in the foregoing embodiments, it improves the accuracy of determining the composition of the composite material to be tested.

[0186] Material transmittance spectral curves enable high-precision subsequent classification processing.

[0187] Corresponding to the above method embodiments, this application also provides a material testing system. Please refer to [link to relevant documentation]. Figure 4 , Figure 4 This application provides a schematic diagram of the functional modules of a material testing system 400, which includes:

[0188] The acquisition module 410 is used to acquire the spectral sequence data corresponding to the composite material to be detected;

[0189] Analysis module 420 is used to perform feature analysis on spectral sequence data using a pre-trained neural network model to obtain analysis results. The pre-trained neural network model is trained from training samples, which include historical spectral sequence data and their corresponding historical analysis results.

[0190] The determination module 430 is used to determine the composition of the composite material to be tested based on the analysis results.

[0191] The material testing system provided in this application embodiment can achieve the following: Figure 1 The various processes implemented in the Chinese method embodiments can achieve similar or the same technical effects, and will not be described again here to avoid repetition.

[0192] In one possible implementation, the pre-trained neural network model includes convolutional layers, pooling layers, and fully connected layers, with the output of the convolutional layer connected to the input of the pooling layer, and the output of the pooling layer connected to the input of the fully connected layer.

[0193] Analysis module 420 is also specifically used for:

[0194] The first feature data is obtained by using a convolutional layer to extract features from the spectral sequence data.

[0195] The first feature data is downsampled using a pooling layer to obtain the second feature data;

[0196] The second feature data is transformed using a fully connected layer to obtain the analysis results.

[0197] In one possible implementation, the convolutional layer includes a first convolutional layer, a second convolutional layer, and a third convolutional layer, wherein the output of the first convolutional layer is connected to the input of the second convolutional layer, and the output of the second convolutional layer is connected to the input of the third convolutional layer.

[0198] Analysis module 420 is also specifically used for:

[0199] The first convolutional layer is used to extract features from the spectral sequence data to obtain the first sub-feature data;

[0200] The second convolutional layer is used to extract features from the first sub-feature data to obtain the second sub-feature data;

[0201] The third sub-feature data is obtained by extracting features from the second sub-feature data using the third convolutional layer. The first feature data includes the third sub-feature data.

[0202] In one possible implementation, in some embodiments, the number of features of the first sub-feature data, the second sub-feature data, and the third sub-feature data increases sequentially, and the sequence length of the first sub-feature data, the second sub-feature data, and the third sub-feature data decreases sequentially.

[0203] In one possible implementation, the first convolutional layer, the second convolutional layer, and the third convolutional layer each include a subconvolutional layer, a modified linear unit function layer, and a max pooling layer. The output of the subconvolutional layer is connected to the input of the modified linear unit function layer, and the output of the modified linear unit function layer is connected to the input of the max pooling layer.

[0204] In one possible implementation, the fully connected layer includes a first fully connected layer and a second fully connected layer, with the output of the first fully connected layer connected to the input of the second fully connected layer.

[0205] Analysis module 420 is also specifically used for:

[0206] The second feature data is nonlinearly combined using the first fully connected layer to obtain combined feature data;

[0207] The combined feature data is classified and mapped using the second fully connected layer to obtain the analysis results.

[0208] In one possible implementation, the acquisition module 410 is further specifically used for:

[0209] The composite material to be tested is irradiated with a first parallel ray to obtain transmitted light, wherein the first parallel ray is formed by the convergence of broadband light rays, and the broadband light rays are generated by a broadband light source;

[0210] The transmitted light is calibrated to obtain the second parallel light ray;

[0211] The second parallel light rays are collected to form spectral sequence data.

[0212] In one possible implementation, the acquisition module 410 is further specifically used for:

[0213] The angle of the first parallel light ray is adjusted so that the adjusted first parallel light ray irradiates the composite material to be tested perpendicularly.

[0214] Figure 5 A schematic diagram of the hardware structure of the electronic device provided in an embodiment of this application is shown.

[0215] The electronic device may include a processor 501 and a memory 502 storing computer program instructions.

[0216] Specifically, the processor 501 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0217] Memory 502 may include mass storage for data or instructions. For example, and not limitingly, memory 502 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 502 may include removable or non-removable (or fixed) media. Where appropriate, memory 502 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 502 is non-volatile solid-state memory.

[0218] In some embodiments, memory 502 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described in the methods provided according to embodiments of this application.

[0219] The processor 501 implements the method provided in the above embodiments by reading and executing computer program instructions stored in the memory 502.

[0220] In one example, the electronic device may also include a communication interface 503 and a bus 510. The processor 501, memory 502, and communication interface 503 are connected via the bus 510 and communicate with each other.

[0221] The communication interface 503 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0222] Bus 510 includes hardware, software, or both, that couples components of an electronic device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 510 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0223] Furthermore, in conjunction with the methods provided in the above embodiments, this application embodiment can be implemented using a computer-readable storage medium. This computer-readable storage medium stores computer program instructions; when executed by a processor, these computer program instructions implement any of the methods in the above embodiments.

[0224] Furthermore, in conjunction with the methods provided in the above embodiments, this application embodiment can provide a computer program product to implement the methods. This program product is stored in a storage medium and executed by at least one processor to implement the various processes of the embodiments of the methods provided in the above embodiments, achieving similar or identical technical effects. To avoid repetition, further details are omitted here.

[0225] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0226] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0227] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0228] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0229] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A material testing method, characterized in that, The method includes: Obtain the spectral sequence data corresponding to the composite material to be tested; The spectral sequence data are subjected to feature analysis using a pre-trained neural network model to obtain analysis results. The pre-trained neural network model is trained from training samples, which include historical spectral sequence data and their corresponding historical analysis results. Based on the analysis results, the composition of the composite material to be tested is determined.

2. The method according to claim 1, characterized in that, The pre-trained neural network model includes convolutional layers, pooling layers, and fully connected layers. The output of the convolutional layer is connected to the input of the pooling layer, and the output of the pooling layer is connected to the input of the fully connected layer. The step of using a pre-trained neural network model to perform feature analysis on the spectral sequence data to obtain analysis results includes: The convolutional layer is used to extract features from the spectral sequence data to obtain first feature data; The pooling layer is used to downsample the first feature data to obtain the second feature data; The second feature data is transformed using the fully connected layer to obtain the analysis result.

3. The method according to claim 2, characterized in that, The convolutional layer includes a first convolutional layer, a second convolutional layer, and a third convolutional layer. The output of the first convolutional layer is connected to the input of the second convolutional layer, and the output of the second convolutional layer is connected to the input of the third convolutional layer. The step of extracting features from the spectral sequence data using the convolutional layer to obtain first feature data includes: The first convolutional layer is used to extract features from the spectral sequence data to obtain the first sub-feature data; The second convolutional layer is used to extract features from the first sub-feature data to obtain the second sub-feature data; The third convolutional layer is used to extract features from the second sub-feature data to obtain the third sub-feature data, and the first feature data includes the third sub-feature data.

4. The method according to claim 3, characterized in that, The number of features in the first sub-feature data, the second sub-feature data, and the third sub-feature data increases sequentially, and the sequence lengths of the first sub-feature data, the second sub-feature data, and the third sub-feature data decrease sequentially.

5. The method according to claim 3, characterized in that, The first convolutional layer, the second convolutional layer, and the third convolutional layer each include a sub-convolutional layer, a modified linear unit function layer, and a max pooling layer. The output of the sub-convolutional layer is connected to the input of the modified linear unit function layer, and the output of the modified linear unit function layer is connected to the input of the max pooling layer.

6. The method according to claim 2, characterized in that, The fully connected layer includes a first fully connected layer and a second fully connected layer, wherein the output of the first fully connected layer is connected to the input of the second fully connected layer. The step of using the fully connected layer to perform feature transformation on the second feature data to obtain the analysis result includes: The second feature data is nonlinearly combined using the first fully connected layer to obtain combined feature data; The combined feature data is classified and mapped using the second fully connected layer to obtain the analysis results.

7. The method according to claim 1, characterized in that, The acquisition of the spectral sequence data corresponding to the composite material to be tested includes: The composite material to be tested is irradiated with a first parallel ray to obtain transmitted light, wherein the first parallel ray is formed by the convergence of broadband light rays, and the broadband light rays are generated by a broadband light source; The transmitted light is calibrated to obtain a second parallel light ray; The second parallel light ray is collected to form the spectral sequence data.

8. The method according to claim 7, characterized in that, The step of irradiating the composite material to be tested with a first parallel light beam includes: The angle of the first parallel light is adjusted so that the adjusted first parallel light shines perpendicularly onto the composite material to be tested.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, wherein the memory stores a computer program, which, when executed by the processor, implements the material testing method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by one or more processors, implements the material testing method according to any one of claims 1 to 8.

Images