Artificial intelligence-based backscatter analysis method, system, and medium
By employing an AI-based backscattering analysis method, and combining AI prediction models with the SIMNRA program, the inefficiency caused by reliance on human experience in existing technologies is solved, enabling rapid and accurate material analysis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- THE ENG & TECHN COLLEGE OF CHENGDU UNIV OF TECH
- Filing Date
- 2026-02-12
- Publication Date
- 2026-06-05
AI Technical Summary
Existing backscatter analysis methods rely on human experience, resulting in low analysis efficiency and limited analytical sensitivity.
An AI-based backscattering analysis method was adopted. The first backscattering energy spectrum of the material to be analyzed was obtained through experiments. The second backscattering energy spectrum of the material to be analyzed was fitted using a pre-trained AI prediction model and the SIMNRA program. Consistency comparison was then performed to obtain the backscattering analysis results.
It improves the efficiency of backscatter analysis, reduces reliance on human experience, enhances the sensitivity and universality of the analysis, and enables the rapid and accurate acquisition of basic material information under different experimental configurations.
Smart Images
Figure CN122157897A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of materials analysis and artificial intelligence technology, specifically to artificial intelligence-based backscattering analysis methods, systems, and media. Background Technology
[0002] Backscatter analysis is a high-energy ion beam analysis technique used in materials science to analyze and measure the structure and composition of materials. It features high precision and non-destructive analysis. By shining a high-energy ion beam of a specific energy onto the material to be analyzed and detecting the energy of the backscattered ions, the type, concentration, and depth distribution of the target atoms can be determined.
[0003] Commonly used methods include Rutherford backscattering (RBS) and elastic backscattering (EBS). RBS analysis often uses alpha ions as the incident ion. However, this method cannot backscatter nuclides with smaller masses (such as D, T, and He), and its sensitivity is low for lighter elements (such as N and O). Elastic backscattering typically uses protons as the incident ion, and at higher energies, generally 2-5 MeV. Due to the high energy, elastic backscattering not only causes Coulomb scattering but also breaks the Coulomb barrier, resulting in nuclear potential scattering. For lighter elements, it not only causes backscattering but also provides higher sensitivity.
[0004] Existing backscattering analysis techniques typically involve acquiring the backscattering experiment's energy spectrum, inputting it into the SIMNRA program, and then setting initial parameters such as material composition, type, and thickness within the SIMNRA program. An initial fitted simulated backscattering energy spectrum is obtained and compared with the experimental backscattering energy spectrum. The set parameters are continuously optimized until the fitted simulated backscattering energy spectrum matches the experimental backscattering energy spectrum in amplitude and shape. Only then are the current material composition, type, and thickness information considered to be true experimental sample information. This fitting simulation analysis process heavily relies on human experience. If the initial parameters differ significantly from the actual material composition, type, and thickness, extensive parameter optimization is required, which is time-consuming, labor-intensive, and inefficient.
[0005] Therefore, it is necessary to propose an efficient and rapid backscattering analysis method to address the shortcomings of existing backscattering methods, such as reliance on human experience, low analysis efficiency, and high time and labor costs. Summary of the Invention
[0006] To address the shortcomings of existing backscattering analysis techniques, which heavily rely on human experience, resulting in low efficiency and limited sensitivity, this invention aims to provide an artificial intelligence-based backscattering analysis method, system, and medium. The method involves experimentally obtaining the first backscattering energy spectrum of the material to be analyzed, fitting the second backscattering energy spectrum of the material to be analyzed using a trained AI prediction model and the SIMNRA program, and obtaining the backscattering analysis result of the material by comparing the consistency of the first and second backscattering energy spectra. This approach solves the problems of reliance on human experience and low efficiency in existing backscattering methods, thereby improving the efficiency of backscattering analysis.
[0007] The above-mentioned technical objective of the present invention is achieved through the following technical solution:
[0008] This solution provides an artificial intelligence-based backscattering analysis method, which includes:
[0009] The first backscattered energy spectrum of the material to be analyzed was obtained experimentally.
[0010] The first backscattered energy spectrum and experimental process information are input into a trained AI prediction model to predict the basic information of the material to be analyzed; the experimental process information includes the type of incident particle, the energy of the incident particle, and the detection angle.
[0011] The basic information and experimental process information are input into the SIMNRA program to fit the second backscattered energy spectrum of the material to be analyzed.
[0012] The backscattering analysis results of the material to be analyzed are obtained by comparing the consistency of the first backscattering energy spectrum and the second backscattering energy spectrum.
[0013] A further optimized approach is that the method for obtaining the first backscattered energy spectrum includes:
[0014] Experimental analysis of the material under analysis is carried out based on Rutherford backscattering analysis or elastic backscattering analysis, and experimental process information and first backscattering energy spectrum are collected.
[0015] A further optimized solution is that the basic information includes: the thickness of the material to be analyzed, the number of layers of the material to be analyzed, and the composition of each layer.
[0016] A further optimization scheme is that the horizontal and vertical axes of the first and second backscattered energy spectra are consistent, where the horizontal axis represents the channel address or particle energy, and the vertical axis represents the particle count or particle count rate.
[0017] A further optimization is that the training method for the AI prediction model includes:
[0018] A backpropagation neural network was constructed, taking the first backscattered energy spectrum and experimental process information as inputs and the basic information of the material to be analyzed as outputs.
[0019] Construct a training sample database based on the Monte Carlo or SIMNRA procedures;
[0020] Set the loss function and loss threshold for the backpropagation neural network;
[0021] The backpropagation neural network is trained based on the training sample database until it converges and the loss function value is lower than the loss threshold, thus obtaining a trained AI prediction model.
[0022] A further optimization scheme is that the loss function is the mean squared error function:
[0023] ;
[0024] Where n represents the total number of training samples; Indicates the first The true value of each training sample. Indicates the first The predicted value of each training sample.
[0025] A further optimized scheme is that the particle is an alpha particle or a proton.
[0026] A further optimized solution is that the method for obtaining the backscattering analysis results includes:
[0027] The ray count and total ray count with energy k are extracted from the first backscattered energy spectrum and the second backscattered energy spectrum, respectively.
[0028] Substituting the ray count and total ray count at energy k into the following formula, the similarity between the first and second backscattered energy spectra can be calculated. :
[0029] The ray count and total ray count with energy k are extracted from the first backscattered energy spectrum and the second backscattered energy spectrum, respectively.
[0030] Substituting the ray count at energy k and the total ray count into the following formula, the similarity between the first and second backscattered energy spectra can be calculated. :
[0031] ;
[0032] in, This represents the count of rays with energy k at the j-th channel of the first backscattered energy spectrum; This represents the count of rays with energy k at the j-th channel of the second backscattered energy spectrum; This represents the count of rays with energy k at the i-th address of the first backscattered energy spectrum; This represents the count of rays with energy k at the i-th address of the second backscattered energy spectrum;
[0033] Preset similarity threshold e, when similarity >Similarity threshold When e, it is determined that the first backscattered energy spectrum and the second backscattered energy spectrum are inconsistent; when the similarity is... ≤similarity threshold When e, the first backscattered energy spectrum is determined to be consistent with the second backscattered energy spectrum;
[0034] When the first backscattered energy spectrum is consistent with the second backscattered energy spectrum, the current basic information is used as the backscattered analysis result output; when the first backscattered energy spectrum is inconsistent with the second backscattered energy spectrum, the basic information parameters in the SIMNRA program are optimized.
[0035] This solution also provides an AI-based backscattering analysis system for implementing the aforementioned AI-based backscattering analysis method. The system includes:
[0036] The acquisition module is used to experimentally obtain the first backscattered energy spectrum of the material to be analyzed.
[0037] The prediction module is used to input the first backscattering energy spectrum and experimental process information into the trained AI prediction model to predict the basic information of the material to be analyzed; the experimental process information includes the type of incident particles in the backscattering experiment, the energy of the incident particles, and the detection angle.
[0038] The fitting module is used to input the basic information and experimental process information into the SIMNRA program to fit the second backscattered energy spectrum of the material to be analyzed.
[0039] The comparison module is used to obtain the backscattering analysis results of the material to be analyzed by comparing the consistency of the first backscattering energy spectrum and the second backscattering energy spectrum.
[0040] This solution also provides a computer-readable medium storing a computer program, which, when executed by a processor, can implement the artificial intelligence-based backscattering analysis method described above; specifically, it performs the following steps:
[0041] Step 1: Experimentally obtain the first backscattered energy spectrum of the material to be analyzed;
[0042] Step two: Input the first backscattered energy spectrum and experimental process information into the trained AI prediction model to predict the basic information of the material to be analyzed; the experimental process information includes the type of incident particle, the energy of the incident particle, and the detection angle.
[0043] Step 3: Input the basic information and experimental process information into the SIMNRA program to fit the second backscattered energy spectrum of the material to be analyzed;
[0044] Step four: Based on the consistency comparison between the first backscattered energy spectrum and the second backscattered energy spectrum, the backscattered analysis results of the material to be analyzed are obtained.
[0045] Compared with the prior art, the present invention has the following beneficial effects:
[0046] 1. This invention provides an artificial intelligence-based backscattering analysis method, system, and medium. Based on existing backscattering analysis methods, this invention improves upon them by obtaining the first backscattering energy spectrum of the material to be analyzed experimentally, fitting the second backscattering energy spectrum of the material to be analyzed based on a trained AI prediction model and the SIMNRA program, and obtaining the backscattering analysis result of the material to be analyzed by comparing the consistency of the first and second backscattering energy spectra. This solves the problems of existing backscattering methods relying on human experience and having low analysis efficiency, thus improving the efficiency of backscattering analysis methods.
[0047] 2. This invention provides an artificial intelligence-based backscattering analysis method, system, and medium. Addressing the problems of existing backscattering analysis methods relying on human experience, being time-consuming and labor-intensive, and having low analysis efficiency, this invention uses experimental process information such as incident particle type, energy, and detection angle as input to the AI prediction model. This allows the model to dynamically adapt to different experimental configurations, improving the method's versatility under different instruments and parameters. The AI prediction model directly predicts basic material information (such as element type, content, and thickness) from energy dispersive spectroscopy, providing high-quality initial fitting parameters for SIMNRA, significantly reducing the number of manual trial and error attempts, and effectively improving the analysis efficiency of existing backscattering analysis methods. Attached Figure Description
[0048] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:
[0049] Figure 1 This is an intentional framework for an AI-based backscattering analysis method.
[0050] Figure 2 This is a schematic diagram of the backscattering analysis method based on artificial intelligence.
[0051] Figure 3 A schematic diagram of the experimental structure for obtaining the first backscattered energy spectrum of the material to be analyzed;
[0052] Figure 4 This is a schematic diagram of the first backscattered energy spectrum;
[0053] Figure 5 This is a schematic diagram of the AI prediction model training process;
[0054] Figure 6 This is a schematic diagram of the fitting process for the second backscattered energy spectrum;
[0055] Figure 7 This is a schematic diagram of an AI-based backscatter analysis system. Detailed Implementation
[0056] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.
[0057] In existing backscattering analysis techniques, after obtaining the experimental backscattering energy spectrum, initial parameters such as material composition, type, and thickness are set within the SIMNRA program to obtain an initial fitted backscattering energy spectrum. This spectrum is then compared with the experimental backscattering energy spectrum, and the set parameters are continuously optimized until the fitted backscattering energy spectrum matches the experimental backscattering energy spectrum in amplitude and shape. Only then are the current material composition, type, and thickness information considered to be true experimental sample information. This fitting analysis process is highly dependent on human experience. If the initial parameters such as material composition, type, and thickness differ significantly from the actual material composition, type, and thickness, extensive parameter optimization is required, which is time-consuming, labor-intensive, and inefficient.
[0058] In view of this, the present solution provides the following embodiments to solve the above-mentioned technical problems:
[0059] Example 1
[0060] This embodiment provides an artificial intelligence-based backscattering analysis method, such as... Figure 1 and Figure 2 As shown, the method includes:
[0061] Step 1: Experimentally obtain the first backscattered energy spectrum of the material to be analyzed;
[0062] In step one, the method for obtaining the first backscattered energy spectrum includes:
[0063] Experimental analysis of the material under analysis is performed using Rutherford backscattering analysis or elastic backscattering analysis, collecting experimental process information and the first backscattered energy spectrum. Other similar backscattering methods can also be used.
[0064] The specific experimental structure diagram is as follows: Figure 3 As shown, charged particles with a certain energy bombard the sample to be analyzed. The detector detects the energy spectrum information of the scattered particles at an angle of (90+θ) to the incident particle direction. Specifically, the horizontal axis of the backscattered energy spectrum represents the channel address or particle energy, and the vertical axis represents the particle count or count rate. The obtained first backscattered energy spectrum is shown below. Figure 4 As shown, the horizontal axis represents the channel address, which is an integer between 1 and 4500, and the vertical axis represents the count of scattered particles detected during the experiment.
[0065] Step 2: Input the first backscattered energy spectrum and experimental process information into the trained AI prediction model to predict the basic information of the material to be analyzed;
[0066] In step two, the experimental process information includes the type of incident particle, the energy of the incident particle, and the detection angle; the particle is an alpha particle or a proton.
[0067] In step two, particle energy refers to the energy of the particles actually used in the backscattering experiment, and the unit can be any of eV, keV, or MeV.
[0068] In step two, the detection angle refers to the angle between the direction of the detector and the direction of particle incidence when performing a backscattering experiment.
[0069] In step two, as Figure 5 As shown, the training method for the AI prediction model includes:
[0070] S31, construct a backpropagation neural network with the first backscattered energy spectrum and experimental process information as input and the basic information of the material to be analyzed as output; specifically, the AI prediction model can also adopt one or more of any artificial intelligence algorithms such as convolutional neural network, ant algorithm, etc.; in this embodiment, the backpropagation neural network is preferred.
[0071] S32, construct a training sample database based on the Monte Carlo procedure or the SIMNRA procedure;
[0072] In this specific embodiment, a geometric model to be simulated is established in the Geant4 program, including parameters such as material thickness, number of material layers, composition of each material layer, incident particle type, incident energy, and detection angle. The backscattered energy spectrum under the current geometric parameters is simulated, and information such as material thickness, number of material layers, and composition of each material layer is modified to obtain a new backscattered energy spectrum.
[0073] Specifically, the current database is divided into training and testing sets in a 7:3 ratio.
[0074] S34, Set the loss function and loss threshold of the backpropagation neural network; in this embodiment, the mean squared error function is preferred as the loss function:
[0075] ;
[0076] Where n represents the total number of training samples; Indicates the first The true value of each training sample. Indicates the first The predicted value of each training sample.
[0077] S35, train the backpropagation neural network based on the training sample database until the backpropagation neural network converges and the loss function value is lower than the loss threshold, and obtain the trained AI prediction model.
[0078] If the values of the loss functions on the training and test sets are below the threshold and convergence occurs, the AI prediction model outputs the current number of hidden layers, number of nodes, activation function, and the final values of the loss functions on the training and test sets. If the values of the loss functions on the training and test sets are not below the threshold and convergence occurs, the number of hidden layers, number of nodes, and activation function of the neural network are further optimized.
[0079] The basic information includes: the thickness of the material to be analyzed, the number of layers of the material to be analyzed, and the composition of each layer.
[0080] Step 3: Input the basic information and experimental process information into the SIMNRA program to fit the second backscattered energy spectrum of the material to be analyzed;
[0081] In this step, SIMNRA software is a commonly used Microsoft Windows program for fitting backscattering spectra, such as... Figure 6 As shown, when the material parameters of the response are input (such as the material thickness, the number of material layers, and the composition of each layer), the SIMNRA software can automatically provide the fitted backscattered energy spectrum based on these parameters.
[0082] Step four: Based on the consistency comparison between the first backscattered energy spectrum and the second backscattered energy spectrum, the backscattered analysis results of the material to be analyzed are obtained.
[0083] The horizontal axis of the first backscattered energy spectrum and the second backscattered energy spectrum represents the channel address or particle energy, and the vertical axis represents the particle count or particle count rate.
[0084] The method for obtaining the backscattering analysis results includes:
[0085] S51, extract the ray count and total ray count with energy k from the first backscattered energy spectrum and the second backscattered energy spectrum respectively;
[0086] S52, Substitute the ray count and total ray count of energy k into the following formula to calculate the similarity between the first backscattered energy spectrum and the second backscattered energy spectrum. :
[0087] The ray count and total ray count with energy k are extracted from the first backscattered energy spectrum and the second backscattered energy spectrum, respectively.
[0088] Substituting the ray count at energy k and the total ray count into the following formula, the similarity between the first and second backscattered energy spectra can be calculated. :
[0089] ;
[0090] in, This represents the count of rays with energy k at the j-th channel of the first backscattered energy spectrum; This represents the count of rays with energy k at the j-th channel of the second backscattered energy spectrum; This represents the count of rays with energy k at the i-th address of the first backscattered energy spectrum; This represents the count of rays with energy k at the i-th address of the second backscattered energy spectrum;
[0091] S53, Preset similarity threshold e, when similarity >Similarity threshold When e, it is determined that the first backscattered energy spectrum and the second backscattered energy spectrum are inconsistent; when the similarity is... ≤similarity threshold When e, the first backscattered energy spectrum is determined to be consistent with the second backscattered energy spectrum;
[0092] Specifically, in this embodiment, the similarity threshold... e is set to 10 -3 ;
[0093] S54: When the first backscattered energy spectrum is consistent with the second backscattered energy spectrum, the current basic information is used as the backscattered analysis result output; when the first backscattered energy spectrum is inconsistent with the second backscattered energy spectrum, the basic information parameters in the SIMNRA program are optimized.
[0094] This embodiment provides an AI-based backscattering analysis method, system, and medium. It improves upon existing backscattering analysis methods by obtaining the first backscattering energy spectrum of the material to be analyzed experimentally, fitting the second backscattering energy spectrum of the material to be analyzed based on a trained AI prediction model and the SIMNRA program, and obtaining the backscattering analysis result of the material to be analyzed by comparing the consistency of the first and second backscattering energy spectra. This solves the problems of existing backscattering methods relying on human experience and having low analysis efficiency, thus improving the efficiency of backscattering analysis methods.
[0095] This invention addresses the problems of existing backscatter analysis methods, which rely on human experience, are time-consuming and labor-intensive, and have low analytical efficiency. By using experimental process information such as incident particle type, energy, and detection angle as input to the AI prediction model, the model can dynamically adapt to different experimental configurations, improving the universality of the method under different instruments and parameters. The AI prediction model directly predicts basic material information (such as element type, content, and thickness) from the energy spectrum, providing high-quality initial fitting parameters for SIMNRA, significantly reducing the number of manual trial and error attempts, and effectively improving the analytical efficiency of existing backscatter analysis.
[0096] Example 2
[0097] This embodiment provides an artificial intelligence-based backscattering analysis system, such as... Figure 7 As shown, the system for implementing the AI-based backscattering analysis method described in Example 1 includes:
[0098] The acquisition module is used to experimentally obtain the first backscattered energy spectrum of the material to be analyzed.
[0099] The prediction module is used to input the first backscattering energy spectrum and experimental process information into the trained AI prediction model to predict the basic information of the material to be analyzed; the experimental process information includes the type of incident particles in the backscattering experiment, the energy of the incident particles, and the detection angle.
[0100] The fitting module is used to input the basic information and experimental process information into the SIMNRA program to fit the second backscattered energy spectrum of the material to be analyzed.
[0101] The comparison module is used to obtain the backscattering analysis results of the material to be analyzed by comparing the consistency of the first backscattering energy spectrum and the second backscattering energy spectrum.
[0102] The acquisition module includes an experimental unit, a first acquisition unit, and a second acquisition unit;
[0103] The experimental unit is used to perform experimental analysis on the materials to be analyzed based on Rutherford backscattering analysis or elastic backscattering analysis.
[0104] The first data acquisition unit is used to collect experimental process information;
[0105] The second acquisition unit is used to acquire the first backscattered energy spectrum.
[0106] The basic information includes: the thickness of the material to be analyzed, the number of layers of the material to be analyzed, and the composition of each layer.
[0107] The horizontal and vertical axes of the first and second backscattered energy spectra are consistent, with the horizontal axis representing channel address or particle energy and the vertical axis representing particle count or particle count rate.
[0108] The prediction module includes a model training sub-module, which includes: a network construction unit, a dataset training unit, a function setting unit, and a training unit.
[0109] The network building unit is used to construct a backpropagation neural network that takes the first backscattered energy spectrum and experimental process information as input and the basic information of the material to be analyzed as output.
[0110] The dataset training unit is used to build a training sample database based on the Monte Carlo or SIMNRA procedure;
[0111] The function setting unit is used to set the loss function and loss threshold of the backpropagation neural network;
[0112] The training unit is used to train the backpropagation neural network based on the training sample database until the backpropagation neural network converges and the loss function value is lower than the loss threshold, thus obtaining a trained AI prediction model.
[0113] The loss function is the mean squared error function:
[0114] ;
[0115] Where n represents the total number of training samples; Indicates the first The true value of each training sample. Indicates the first The predicted value of each training sample.
[0116] The particle is either an alpha particle or a proton.
[0117] The comparison module includes: an extraction unit, a calculation unit, a judgment unit, and an output optimization unit;
[0118] The extraction unit is used to extract the k-ray count and the total ray count from the first backscattered energy spectrum and the second backscattered energy spectrum, respectively.
[0119] The calculation unit is used to substitute the ray count and total ray count of energy k into the following formula to calculate the similarity between the first backscattered energy spectrum and the second backscattered energy spectrum. :
[0120] The ray count and total ray count with energy k are extracted from the first backscattered energy spectrum and the second backscattered energy spectrum, respectively.
[0121] Substituting the ray count at energy k and the total ray count into the following formula, the similarity between the first and second backscattered energy spectra can be calculated. :
[0122] ;
[0123] in, This represents the count of rays with energy k at the j-th channel of the first backscattered energy spectrum; This represents the count of rays with energy k at the j-th channel of the second backscattered energy spectrum; This represents the count of rays with energy k at the i-th address of the first backscattered energy spectrum; This represents the count of rays with energy k at the i-th address of the second backscattered energy spectrum;
[0124] The determination unit is used to preset the similarity threshold. e, when similarity >Similarity threshold When e, it is determined that the first backscattered energy spectrum and the second backscattered energy spectrum are inconsistent; when the similarity is... ≤similarity threshold When e, the first backscattered energy spectrum is determined to be consistent with the second backscattered energy spectrum;
[0125] The output optimization unit is used to output the backscattering analysis result with the current basic information when the first backscattering energy spectrum and the second backscattering energy spectrum are consistent; and to optimize the basic information parameters in the SIMNRA program when the first backscattering energy spectrum and the second backscattering energy spectrum are inconsistent.
[0126] This embodiment provides an AI-based backscattering analysis method, system, and medium. It improves upon existing backscattering analysis methods by obtaining the first backscattering energy spectrum of the material to be analyzed experimentally. A second backscattering energy spectrum is then fitted using a pre-trained AI prediction model and the SIMNRA program. The backscattering analysis results are obtained by comparing the consistency of the first and second backscattering energy spectra. This approach solves the problems of existing backscattering methods relying on human experience and having low analysis efficiency, thus improving the efficiency of backscattering analysis methods. Users do not need to master the complex parameter adjustment techniques of the SIMNRA program; the AI automatically provides optimized input parameters, making this traditional professional tool easier to use widely.
[0127] Example 3
[0128] This embodiment provides a computer-readable medium storing a computer program thereon. The computer program, when executed by a processor, can implement the artificial intelligence-based backscattering analysis method as described in Embodiment 1; specifically, it performs the following steps:
[0129] Step 1: Experimentally obtain the first backscattered energy spectrum of the material to be analyzed;
[0130] Step two: Input the first backscattered energy spectrum and experimental process information into the trained AI prediction model to predict the basic information of the material to be analyzed; the experimental process information includes the type of incident particle, the energy of the incident particle, and the detection angle.
[0131] Step 3: Input the basic information and experimental process information into the SIMNRA program to fit the second backscattered energy spectrum of the material to be analyzed;
[0132] Step four: Based on the consistency comparison between the first backscattered energy spectrum and the second backscattered energy spectrum, the backscattered analysis results of the material to be analyzed are obtained.
[0133] Traditional backscattered energy dispersive spectroscopy (EDS) analysis heavily relies on the operator's expertise and experience, requiring repeated manual parameter adjustments for fitting, a time-consuming process susceptible to subjective bias. This solution utilizes a pre-trained AI prediction model to directly predict basic material information (such as element type, content, and thickness) from the EDS spectrum, providing high-quality initial fitting parameters for the SIMNRA program. This significantly reduces manual trial and error, enabling one-click rapid analysis. The AI prediction model can complete the process in seconds, and combined with the precise fitting of the SIMNRA program, it reduces the analysis process, which might have taken hours or even days, to minutes, making it particularly suitable for high-throughput, batch material testing scenarios.
[0134] In this approach, the AI prediction model, trained on a large amount of data, can capture the complex features hidden in the energy spectrum, avoiding misjudgments caused by insufficient operator experience and improving the scientific rigor and consistency of initial parameter settings. A collaborative workflow of AI prediction and SIMNRA physical model fitting is employed to achieve cross-validation: the AI prediction model provides rapid initial screening, the SIMNRA program performs fine-grained fitting based on physical principles, and finally, consistency comparison ensures the reliability of the results, reducing the limitations of a single method.
[0135] This approach explicitly uses experimental process information such as incident particle type, energy, and detection angle as input to the AI prediction model, enabling the model to dynamically adapt to different experimental configurations and improving the method's versatility across different instruments and parameters. With the accumulation of training data, the AI prediction model can be continuously iterated and upgraded, further enhancing its ability to analyze new materials and complex energy spectra, thus giving the overall method the potential for continuous evolution.
[0136] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a set-top box (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.
[0137] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A backscattering analysis method based on artificial intelligence, characterized in that, The method includes: The first backscattered energy spectrum of the material to be analyzed was obtained experimentally. The first backscattered energy spectrum and experimental process information are input into a trained AI prediction model to predict the basic information of the material to be analyzed; the experimental process information includes the type of incident particle, the energy of the incident particle, and the detection angle. The basic information and experimental process information are input into the SIMNRA program to fit the second backscattered energy spectrum of the material to be analyzed. The backscattering analysis results of the material to be analyzed are obtained by comparing the consistency of the first backscattering energy spectrum and the second backscattering energy spectrum.
2. The backscattering analysis method based on artificial intelligence according to claim 1, characterized in that, The method for obtaining the first backscattered energy spectrum includes: Experimental analysis of the material under analysis is carried out based on Rutherford backscattering analysis or elastic backscattering analysis, and experimental process information and first backscattering energy spectrum are collected.
3. The backscattering analysis method based on artificial intelligence according to claim 1, characterized in that, The basic information includes: the thickness of the material to be analyzed, the number of layers of the material to be analyzed, and the composition of each layer.
4. The backscattering analysis method based on artificial intelligence according to claim 1, characterized in that, The horizontal and vertical axes of the first and second backscattered energy spectra are consistent, with the horizontal axis representing channel address or particle energy and the vertical axis representing particle count or particle count rate.
5. The backscattering analysis method based on artificial intelligence according to claim 3, characterized in that, The training methods for the AI prediction model include: A backpropagation neural network was constructed, taking the first backscattered energy spectrum and experimental process information as inputs and the basic information of the material to be analyzed as outputs. Construct a training sample database based on the Monte Carlo or SIMNRA procedures; Set the loss function and loss threshold for the backpropagation neural network; The backpropagation neural network is trained based on the training sample database until it converges and the loss function value is lower than the loss threshold, thus obtaining a trained AI prediction model.
6. The backscattering analysis method based on artificial intelligence according to claim 5, characterized in that, The loss function is the mean squared error function: ； Where n represents the total number of training samples; Indicates the first The true value of each training sample. Indicates the first The predicted value of each training sample.
7. The backscattering analysis method based on artificial intelligence according to claim 1, characterized in that, The particle is either an alpha particle or a proton.
8. The backscattering analysis method based on artificial intelligence according to claim 1, characterized in that, The method for obtaining the backscattering analysis results includes: The ray count and total ray count with energy k are extracted from the first backscattered energy spectrum and the second backscattered energy spectrum, respectively. Substituting the ray count at energy k and the total ray count into the following formula, the similarity between the first and second backscattered energy spectra can be calculated. : ； in, This represents the count of rays with energy k at the j-th channel of the first backscattered energy spectrum; This represents the count of rays with energy k at the j-th channel of the second backscattered energy spectrum; This represents the count of rays with energy k at the i-th address of the first backscattered energy spectrum; This represents the count of rays with energy k at the i-th address of the second backscattered energy spectrum; Preset similarity threshold e, when similarity Similarity threshold When e, it is determined that the first backscattered energy spectrum and the second backscattered energy spectrum are inconsistent; when the similarity is... ≤similarity threshold When e, the first backscattered energy spectrum is determined to be consistent with the second backscattered energy spectrum; When the first backscattered energy spectrum is consistent with the second backscattered energy spectrum, the current basic information is used as the backscattered analysis result output; when the first backscattered energy spectrum is inconsistent with the second backscattered energy spectrum, the basic information parameters in the SIMNRA program are optimized.
9. A backscattering analysis system based on artificial intelligence, characterized in that, The system for implementing the AI-based backscattering analysis method according to any one of claims 1-8, the system comprising: The acquisition module is used to experimentally obtain the first backscattered energy spectrum of the material to be analyzed. The prediction module is used to input the first backscattering energy spectrum and experimental process information into the trained AI prediction model to predict the basic information of the material to be analyzed; the experimental process information includes the type of incident particles in the backscattering experiment, the energy of the incident particles, and the detection angle. The fitting module is used to input the basic information and experimental process information into the SIMNRA program to fit the second backscattered energy spectrum of the material to be analyzed. The comparison module is used to obtain the backscattering analysis results of the material to be analyzed by comparing the consistency of the first backscattering energy spectrum and the second backscattering energy spectrum.
10. A computer-readable medium having a computer program stored thereon, characterized in that, The computer program, executed by a processor, can implement the artificial intelligence-based backscattering analysis method as described in any one of claims 1-8; specifically, it performs the following steps: Step 1: Experimentally obtain the first backscattered energy spectrum of the material to be analyzed; Step two: Input the first backscattered energy spectrum and experimental process information into the trained AI prediction model to predict the basic information of the material to be analyzed; the experimental process information includes the type of incident particle, the energy of the incident particle, and the detection angle. Step 3: Input the basic information and experimental process information into the SIMNRA program to fit the second backscattered energy spectrum of the material to be analyzed; Step four: Based on the consistency comparison between the first backscattered energy spectrum and the second backscattered energy spectrum, the backscattered analysis results of the material to be analyzed are obtained.