CN113536679B - Point source dose rate correction method based on artificial neural network - Google Patents

Point source dose rate correction method based on artificial neural network Download PDF

Info

Publication number
CN113536679B
CN113536679B CN202110816233.9A CN202110816233A CN113536679B CN 113536679 B CN113536679 B CN 113536679B CN 202110816233 A CN202110816233 A CN 202110816233A CN 113536679 B CN113536679 B CN 113536679B
Authority
CN
China
Prior art keywords
data
neural network
artificial neural
dose rate
energy spectrum
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202110816233.9A
Other languages
Chinese (zh)
Other versions
CN113536679A (en
Inventor
王泽宇
莫洪
汤晓斌
龚频
王鹏
梁大戬
周程
朱晓翔
蒋若澄
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Aeronautics and Astronautics
Original Assignee
Nanjing University of Aeronautics and Astronautics
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing University of Aeronautics and Astronautics filed Critical Nanjing University of Aeronautics and Astronautics
Priority to CN202110816233.9A priority Critical patent/CN113536679B/en
Publication of CN113536679A publication Critical patent/CN113536679A/en
Application granted granted Critical
Publication of CN113536679B publication Critical patent/CN113536679B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Analysis (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Evolutionary Computation (AREA)
  • Computational Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Computational Linguistics (AREA)
  • Algebra (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Databases & Information Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Operations Research (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Measurement Of Radiation (AREA)

Abstract

The application discloses a point source dose rate correction method based on an artificial neural network, which comprises the following steps: s1, energy spectrum data and corresponding dose rate values of the unmanned aerial vehicle radioactivity monitoring system at different heights are obtained and are respectively used as input parameters and output parameters; s2, dividing part of energy spectrum data at different heights into training data, dividing the other part of energy spectrum data into test data, constructing an artificial neural network model by utilizing input parameters and output parameters in the training data, and importing the training data into the artificial neural network model for training to obtain a trained artificial neural network model; s3, respectively importing the test data into the trained artificial neural network model to obtain an ideal output result, and comparing errors between the ideal output result and the corresponding test data; and if the error is greater than or equal to the set precision expected value, repeating the steps S2 and S3, and if the error is less than the set precision expected value, training the debugged artificial neural network model to be a point source dose rate correction algorithm.

Description

Point source dose rate correction method based on artificial neural network
Technical Field
The application relates to the technical field of unmanned aerial vehicle radioactivity monitoring, in particular to a point source dose rate correction method based on an artificial neural network.
Background
The unmanned aerial vehicle radioactivity monitoring is to take a fixed wing or rotor unmanned aerial vehicle as a platform, rely on airborne radioactivity monitoring load equipment such as a radiation detector and the like to measure the radioactivity in the air, and is applicable to the fields such as nuclear accidents, environmental radioactivity monitoring, mineral geological investigation and the like. The conventional detector carried by the unmanned aerial vehicle radioactive monitoring system acquires energy spectrum information of nuclides, and when radiation evaluation is carried out, the type of different radionuclides can be distinguished, the air absorption dosage rate of the nuclides can not be measured at the same time, and a multi-parameter evaluation means is not provided. If the dosage rate is to be intuitively measured, a dosage rate meter needs to be carried, so that the operation difficulty of the unmanned aerial vehicle is greatly improved, the working efficiency is reduced, and the unmanned aerial vehicle platform is not beneficial to carrying out radiation monitoring work. Meanwhile, when the unmanned aerial vehicle-mounted system works, in order to intuitively observe the radiation level of each environment, the position information of each point source and the surrounding radiation level when a nuclear accident occurs, the detected and acquired dose rate needs to be subjected to high correction, the air absorption dose rate of the radiation source is converted to the same level, and the dose rate thermodynamic diagram is drawn, so that the radiation level of each position can be intuitively monitored. Fig. 1 shows a schematic diagram of a radiation monitoring system using a drone in practical use. As shown in fig. 1, a sodium iodide detector (thallium excitation) is mounted on the unmanned aerial vehicle to form a unmanned aerial vehicle radiation monitoring system. And when the unmanned aerial vehicle flies at different heights, detecting the dosage rate of the radioactive point source in the air on the ground. However, in this case, the detected dose rate at the same altitude is often inaccurate due to the variation in the altitude of the unmanned aerial vehicle, and thus, correction is required.
In order to solve the problems, researchers at home and abroad have made related research work. For aviation radiation monitoring in the point source mode, researchers have now equipped large NaI (Tl) detectors (the detection system consists of 18 4.2L NaI (Tl) crystals with a total detection volume of 75.6L) on a large aircraft to highly correct the dose rate generated by the ground point source in order to find the high attenuation coefficient in the point source mode, and the correction method is calibrated according to the empirical formula provided by IAEA. The method has the defects that firstly, the volume of a loaded NaI (Tl) detector is too large to be suitable for a small unmanned aerial vehicle, and the operation environment is limited; secondly, the conditions of the used empirical formula are more ideal, the actual application scene is not considered, and the applicability is poor;
Secondly, researchers also adopt energy spectrum dose rate conversion functions (GE functions) to realize conversion between the energy spectrum counts obtained by detection and the corresponding dose rates. The method is complex in work and multiple in implementation steps, and causes great difficulty in real-time online data processing during the unmanned aerial vehicle radiation monitoring system; therefore, according to the unmanned aerial vehicle radioactivity monitoring principle and the practical application requirement, it is necessary to design a radionuclide dose rate height correction scheme aiming at a point source, which can be monitored on line in real time.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a dose rate correction method based on an artificial neural network, which is used for training key data in an energy spectrum acquired by a detector by adopting a prediction model based on the artificial neural network so as to further realize prediction.
The invention provides a point source dose rate correction method based on an artificial neural network, which comprises the following steps: s1, acquiring corresponding dose rate values of energy spectrum data of an unmanned aerial vehicle radioactivity monitoring system at different heights, wherein the dose rate values are respectively used as input parameters and output parameters; s2, dividing part of the energy spectrum data at different heights into training data, dividing the other part into test data, constructing an artificial neural network model by using the input parameters and the output parameters in the training data, and importing the training data into the artificial neural network model for training to obtain a trained artificial neural network model; s3, respectively importing the test data into the trained artificial neural network model to obtain an ideal output result, and comparing errors between the ideal output result and the corresponding test data; and if the error is larger than or equal to the set precision expected value, repeating the steps S2 and S3, and if the error is smaller than the set precision expected value, training the debugged artificial neural network model to be a point source dose rate correction algorithm.
Further, the method for acquiring the energy spectrum of the unmanned aerial vehicle radioactivity monitoring system at different heights comprises the following steps: monte Carlo software was used to simulate the energy spectrum of the unmanned aerial vehicle radioactivity monitoring system at different heights.
Further, the method for obtaining the input parameters and the output parameters in the energy spectrum data with different heights and constructing the artificial neural network model comprises the following steps: carrying out normalization processing on the energy spectrum data at different heights; the number of main components which mainly contribute to the dose rate deposition after normalization processing is extracted as an input parameter; the air absorption dose rate of the radioactive source deposited in the air is taken as an output parameter; calculating an implicit layer by using the input parameter and the output parameter; the structure of the artificial neural network model is an input layer, an hidden layer and an output layer, and the structure of the artificial neural network model is utilized to construct the model of the artificial neural network.
Further, the method for extracting the number of main components which mainly contribute to the dose rate deposition after normalization treatment comprises the following steps: extracting data related to the air deposition dose rate from the energy spectrum data at different heights; carrying out standardization processing on the related data to obtain standardized data; calculating correlation coefficients among the standardized data, and forming a correlation coefficient matrix; calculating each characteristic value of the correlation coefficient matrix; and calculating the contribution rate and the accumulated contribution rate of each related data by using the characteristic values, wherein when the accumulated contribution rate is larger than a threshold value, the corresponding minimum characteristic value quantity is used as the main component quantity contributing mainly, namely the input parameter.
Further, the data relating to the air deposition dose rate includes: the total energy peak count, energy, single escape peak, double escape peak, annihilation peak and corresponding measurement time and height of each nuclide.
Further, the method for obtaining the standardized data by performing the standardized processing on the related data is to use the following formula:
Wherein x ij represents the j index of the i-th energy spectrum; data is shown after x ij has undergone a normalization operation; wherein the method comprises the steps of Is representative of the average value of all of the above-mentioned energy spectrum data; as the variance of the data, the data variance is represented, and s j is the standard deviation between the two data represented.
Further, the method for calculating the correlation coefficient between the standardized data uses the following formula:
Wherein r ij is the correlation coefficient between the energy spectrum data x i and x j; x i is the average value of the ith energy spectrum data; x j is the average value of the j-th energy spectrum data; n represents the number of spectrums used for calculation, and x kj represents the j index of the k energy spectrum.
Further, the expected precision value is 10 -3.
Further, the method further comprises the steps of: if the error is smaller than the expected precision value, measuring the dose rate of the same radioactive source at different heights by using an unmanned aerial vehicle radioactivity monitoring system in an actual scene, and comparing the dose rate with the actual output result to verify the quality of the artificial neural network model.
The method disclosed by the application solves the problems of long time and high cost of manpower and material resources in the dose verification work in the dose rate height correction process in the existing point source mode, can improve the efficiency and quality of dose correction, and the result is beneficial to analyzing the verification result and reduces the verification cost. By using the BPNN artificial neural network, the efficiency of the whole dose rate height correction can be effectively improved, and the integrated correction of the nuclide dose rate height correction is realized.
Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
Fig. 1 is a schematic diagram of a radiation monitoring system using an unmanned aerial vehicle in practical application.
FIG. 2 is a flow chart of a point source dose rate correction method based on a neural artificial neural network.
FIG. 3 is a schematic diagram of a Nal (TI) detector model.
Detailed Description
The following detailed description of specific embodiments of the invention is provided in connection with the accompanying drawings and examples in order to provide a better understanding of the aspects of the invention and advantages thereof. However, the following description of specific embodiments and examples is for illustrative purposes only and is not intended to be limiting of the invention.
The following disclosure provides many different embodiments, or examples, for implementing different features of the invention. In order to simplify the present disclosure, components and arrangements of specific examples are described below. They are, of course, merely examples and are not intended to limit the invention. Furthermore, the present invention may repeat reference numerals and/or letters in the various examples, which are for the purpose of brevity and clarity, and which do not themselves indicate the relationship between the various embodiments and/or arrangements discussed. In addition, the present invention provides examples of various specific processes and materials, but one of ordinary skill in the art will recognize the application of other processes and/or the use of other materials.
First, the origin of BPNN (back propagation neural network algorithm) under the explanation is needed. Artificial neural networks (ARTIFICIAL NEURAL NETWORK, ANN) are a growing research hotspot in the field of artificial intelligence since the 80 s of the 20 th century. The research of artificial neural networks has been in progress in the last decade, and the Back-Propagation algorithm (Back-Propagation) has been proposed as the earliest and most commonly used multi-layer perceptron network training method, and has been used until 70 and 80 years. The back propagation artificial neural network (BPNN) using the error back propagation algorithm as a core is also widely used.
The artificial neural network (BP) model topological structure comprises an input layer, an hidden layer and an output layer, wherein neurons of the input layer are determined by the dimension of the sample attribute, the number of neurons of the output layer is determined by the sample classification number, and the hidden layer can be determined according to the actual situation. Essentially, the BP algorithm uses a gradient descent method to calculate the minimum value of an objective function by using a network error square objective function. The basic BP algorithm includes two processes, forward propagation of the signal and backward propagation of the error. BPNN (back propagation neural network) has many advantages such as strong nonlinear mapping capability, high self-learning and adaptive capability, strong fault tolerance capability, etc. By using the BPNN, the efficiency of the whole dose rate height correction can be effectively improved, and the integrated correction of the nuclide dose rate height correction is realized.
In addition, the MCNP simulation software utilized in the present application needs to be briefly described below. The MCNP software is the software using the Monte Carlo simulation method. The subject nature of nuclear technology has led to the inability to perform actual measurements or experimental work in many cases, where computer simulation calculations have shown unique advantages. Meanwhile, for the nuclear reaction problem with complex structure and reaction mechanism, the general numerical method is difficult to solve, while the MC (Monte Carlo) method can accurately simulate the physical process in practice, and solve the problem that the traditional numerical method is difficult to solve, so that the method is widely applied to the research of the nuclear-related field. The MC method grows gradually along with the development of atomic energy industry in the middle 40 th century, the basic idea is a statistical sampling method based on random number selection, the traditional empirical method cannot approach to a real physical process, satisfactory results are difficult to obtain, the Monte Carlo method has remarkable advantages in solving the particle transport problem, then a plurality of MC simulation programs are gradually developed, EGS, MCNP, GEANT4 and the like are mainly used at present, wherein MCNP is developed by the American Larmose national laboratory of los, photon, electron, neutron and the like transport problem in substances in a three-dimensional geometry, the applicable photon energy range is 1E-3 MeV-1E 5MeV, the electron energy range is 1E-3 MeV-1E 3MeV, and the neutron energy range is 1E-11 MeV-20 MeV. The MCNP has complete program function, various material reaction section data are rich, the variance reduction method is various, the universality is strong and the use is simple.
Fig. 2 shows a flow chart of the point source dose rate correction method based on the artificial neural network.
As shown in fig. 2, the energy spectrum, i.e. simulation data, of the unmanned aerial vehicle radioactivity monitoring system at different heights is firstly obtained, and the simulation is performed by using MCNP software. And extracting energy spectrum data and dose rate values at different heights as input parameters and output parameters. And dividing part of the energy spectrum data at different heights into training data, and dividing the other part into test data. It should be noted that the number of test data is not less than 50, and the number of training data is far greater than the number of test data, for example, 50, and 450 in the embodiment of the present application.
And constructing an artificial neural network model, namely a BPNN algorithm model, by using the input parameters and the output parameters in the training data, respectively importing the training data into the artificial neural network model for training, and obtaining the trained artificial neural network model.
And respectively importing the test data into the trained artificial neural network model to obtain an ideal output result, comparing the ideal output under the same height with the corresponding test data, and if the error is smaller than the precision expected value, obtaining the artificial neural network training model as a required point source air dose rate correction algorithm. If the error is greater than or equal to the expected precision value, the artificial neural network training model does not meet the requirement, and the artificial neural network model needs to be continuously trained by repeating the steps. In addition, if the error is smaller than the precision expected value, the correction algorithm is verified by the energy spectrum data actually measured by the unmanned aerial vehicle airborne radiation monitoring system on site.
Each step will be described in detail below. The application adopts MCNP to simulate the emissivity of ground point sources in the air with different heights. Firstly, selecting a proper unmanned aerial vehicle flight altitude interval and a nuclide energy interval, and selecting a proper airborne detection system.
And secondly, using MCNP to simulate and acquire energy spectrum data of the unmanned aerial vehicle radioactivity monitoring system at different heights. And divides the energy spectrum data into test data and training data. And extracting data related to the air deposition dosage rate from the training data, carrying out normalization processing on all information data related to the air deposition dosage rate, and selecting the number of principal components which mainly contribute to the dosage rate deposition as an input parameter. The air absorption dose rate of the radiation source deposited in air is taken as an output parameter.
And thirdly, calculating the hidden layer node number according to the input parameters and the output parameters. Constructing an artificial neural network model. And respectively importing the training data into the constructed artificial neural network model to perform model training. And obtaining the trained artificial neural network model.
The method for constructing the artificial neural network model comprises the following steps: and carrying out normalization processing on the energy spectrum data at different heights. And extracting the number of principal components which mainly contribute to the dose rate deposition after normalization processing as an input parameter. The air absorption dose rate of the radioactive source deposited in the air is taken as an output parameter; the number of hidden layer nodes is calculated using the input parameters and the output parameters. The structure of the artificial neural network model is that the input parameters-hidden layer node number-output parameters are used for constructing the model of the artificial neural network.
The method for extracting the number of main components which mainly contribute to the deposition of the dose rate after normalization processing is to extract data related to the dose rate of air deposition from energy spectrum data at different heights. And carrying out standardization processing on the related data to obtain standardized data. And calculating correlation coefficients among the standardized data, and forming a correlation coefficient matrix. Calculating each characteristic value of the correlation coefficient matrix; and calculating the contribution rate and the accumulated contribution rate of each related data by utilizing each characteristic value, and determining the corresponding minimum characteristic value number as the main component number contributing mainly when the accumulated contribution rate exceeds a threshold value.
Data relating to the air deposition dose rate includes the total energy peak count, energy, single escape peak, double escape peak, annihilation peak and corresponding measurement times and heights for each species.
The method for obtaining the standardized data by carrying out standardized processing on the related data comprises the following steps of:
Wherein x ij represents the j index of the i-th energy spectrum; data is shown after x ij has undergone a normalization operation; wherein the method comprises the steps of Is representative of the average of all of the energy spectrum data; As the variance of the data, the data variance is represented, and s j is the standard deviation between the two data represented. n represents the number of energy spectrums used for calculation.
Wherein the mean and standard deviation can be calculated according to the following formula:
The meaning of each term in the formula is the same as above.
The method for calculating the correlation coefficient between the standardized data is to use the following formula:
Wherein r ij is a correlation coefficient between the energy spectrum data x i and x j; x i is the average value of the ith energy spectrum data; x j is the average value of the j-th energy spectrum data; n represents the number of spectrums used for calculation, and x kj represents the j index of the k energy spectrum.
Combining the correlation coefficients into a correlation coefficient matrix R
Calculating a correlation coefficient matrix, and sequencing the obtained eigenvalues and eigenvectors, namely solving an eigenvalue equation:
|λI-R|＝0
and I is an identity matrix of a corresponding dimension, the eigenvalue lambda 1,λ2,λ3,...,λ17 and the eigenvector corresponding to the eigenvalue lambda 1,λ2,λ3,...,λ17 are obtained by calculating the above formula, and the eigenvectors are ordered according to the size.
Then, the contribution rate of each component and the accumulated contribution value are calculated to screen the main component, and the calculation formula of the contribution rate alpha l of the first main component is as follows:
The calculation formula of the accumulated contribution rate G k of the k principal components is as follows:
when the cumulative contribution rate is greater than 80%, the corresponding minimum k value, namely the number of main components, is also input information.
And fourthly, testing the trained artificial neural network model, respectively importing test data into the trained artificial neural network model to obtain an ideal output result, comparing errors between the test data and the corresponding test data, and if the errors are smaller than a set precision expected value, obtaining the trained network model as a required correction method.
The method for comparing the error between the ideal output result and the corresponding test data is according to the following formula:
Wherein m i represents an ideal output result obtained by the trained artificial neural network model, o i represents an actual output result obtained by the trained artificial neural network model, and n represents the number of samples.
Empirically, the expected model accuracy is typically set to 10 -3.
In addition, part of the input layer, output layer and hidden layer information extracted from the MCNP software can be used for training an artificial neural network model, and the other part can be used for testing the identification capability of the BN network model before training the network model.
In addition, as shown in fig. 2, the trained artificial neural network is utilized to verify the data sample of the radioactive source, which uses the unmanned plane radioactivity monitoring system to measure the quality, at a certain distance, so as to verify the training quality of the artificial neural network model.
Examples
For ease of understanding, the present application discloses in detail a first embodiment modified by applying the method.
Step one: an appropriate on-board detection system is selected. The invention selects a NaI (T1) detector system, which is composed of a 2X 2 inch NaI (T1) detector, a multichannel amplitude analyzer and matched MAESTRO software, wherein the energy range of the detector is 30keV to 3MeV, the energy resolution is 7.7 percent (Cs-137, 662 KeV), and the multichannel amplitude analyzer is 1024 channels. According to the detection principle of a NaI (T1) detector, gamma particles and the detector act on a crystal part mainly, and in the NaI (T1) detector model diagram shown in figure 3, the crystal geometric dimension and surrounding cladding material structure are shown as the figure, wherein the thickness of an aluminum shell is 2.5mm, the thickness of a glass material is SiO 2, the thickness of the glass material is 2mm, the material of a reflecting layer is MgO, and the thickness of the reflecting layer is 0.5mm. The model used in the MCNP simulation process was also modeled in accordance with the data described above.
Step two: according to the method provided by the invention, the flight height interval of the unmanned aerial vehicle selected in a simulation way is 0-100 m, the value is 0.2 m apart from 0-100 m to obtain a height value, 500 groups of height data are selected in total, and 500 groups of simulation data can be correspondingly obtained; the corresponding energy information is expressed as the energy of the incident photon, set to energy values representing different nuclides, the content comprising: am-241 (59.5 KeV), cs-137 (662 KeV), co-60 (1173 KeV and 1332 KeV), K-40 (1460 KeV).
Step three: according to the method provided by the invention, the MCNP simulates the selected detector model: the point source is at ground level 0 meters and the detector of the unmanned aerial vehicle radioactivity monitoring system is directly above the point source, with the distance taking the distance interval described above (as shown in fig. 1). The source term of the simulation is isotropic, and the emission weight is 1. And (3) establishing a model by using a corresponding curved surface card and a cell card, and acquiring an energy spectrogram of the sodium iodide crystal detector by using an F8 card. According to the method provided by the invention, in the MCNP simulation process, the authenticity of the simulated energy spectrum is related to the number of emission particles, and the simulation can be carried out by adopting the emission particle number of 1×10 11.
Step four: and training the BP artificial neural network by taking 450 groups of data in 500 groups of data obtained through simulation as training data, and testing the identification capacity of the BP artificial neural network by taking the rest 50 groups of data as test data.
Step five: according to the method provided by the invention, firstly, the energy spectrum obtained by MCNP is normalized, and then the energy spectrum characteristic extraction is carried out, so that the total energy peak count and the energy corresponding to the Am-241 nuclide in the energy spectrum are obtained; the total energy peak count and energy corresponding to Cs-137 nuclides; the count and energy of the totipotent peak corresponding to the 1173KeV peak of Co-60; the total energy peak count, energy and single escape peak and double escape peak corresponding to 1332KeV peak of Co-60; the count, energy and single escape peak and double escape peak of the all-round peak corresponding to 1460KeV peak of K-40; together with the annihilation peaks in the energy spectrum, the corresponding measurement time and height are 17 data relating to the airborne deposition dose rate of the nuclear species. Because the energy spectrum information is numerous, the energy spectrum data obtained by extraction is multidimensional data, the extracted data is required to be subjected to dimension reduction treatment, and the data with the largest contribution to the corresponding dose rate is selected.
Step six: according to the method provided by the invention, 450 groups of data of the training set are selected, and then a 450X 17 data matrix X is constructed according to 17 data selected in the fifth step as variables.
Where i=1, 2,3,..450, j=1, 2,3,..17. i represents 450 data of the training set, j represents 17 energy spectrum characteristic value data obtained by extraction.
Step seven: and D, processing the data in the step six to obtain a standardized matrix. The purpose is to eliminate the difference in dimensionality and magnitude order between the individual data and obtain normalized dataWherein the data within the matrixCan be calculated by the following formula:
Wherein x ij represents the j index of the i-th energy spectrum; data is shown after x ij has undergone a normalization operation; wherein the method comprises the steps of Is the average of all data; as the variance of the data, the data variance is represented, and s j is the standard deviation between the two data represented.
In the above formula, n represents the number of energy spectrums for calculation, and n=450.
To obtain the contribution value of each component, a matrix R (correlation coefficient matrix) is calculated:
wherein the calculation formula of r ij is as follows:
In the above formula, r ij is the correlation coefficient of the original variables x i and x j; x i is the ith energy spectrum; x j is the j-th energy spectrum characteristic value; n represents the number of spectrums used for calculation.
And (3) calculating a characteristic value and a characteristic vector obtained by the correlation coefficient matrix, and sequencing, namely solving a characteristic equation:
|λI-R|＝0
and I is an identity matrix of a corresponding dimension, the eigenvalue lambda 1,λ2,λ3,...,λ17 and the eigenvector corresponding to the eigenvalue lambda 1,λ2,λ3,...,λ17 are obtained by calculating the above formula, and the eigenvectors are ordered according to the size.
Then, the contribution rate of each component and the accumulated contribution value are calculated to screen the main component, and the calculation formula of the contribution rate alpha l of the first main component is as follows:
The calculation formula of the accumulated contribution rate G k of the k principal components is as follows:
As calculated by the above equation and the embodiment provided by the present invention, when the number of principal components is equal to 12, the cumulative contribution rate is greater than 80%, and it is considered that these twelve data are principal components of the present embodiment.
The 12 main components are the energy of the rays emitted by the four radioactive sources Am-241 (59.5 KeV), cs-137 (662 KeV), co-60 (1173 KeV and 1332 KeV) and K-40 (1460 KeV) in the simulation, the count of the totipotent peaks corresponding to 5 energies in the energy spectrum, and the time and the height of the measurement.
Step eight: the 12 characteristic parameters obtained in the step seven are applied to the input parameters of the artificial neural network, meanwhile, the output parameters are the deposition air absorption dose rate of each nuclide in the air, and the total output parameters are 4, so that the number of used input nodes N=12 and the number of used output nodes M=4. Determining the number of hidden layer nodes Q may be obtained by an empirical formula:
the optimal node number selected by the method is 10. Therefore, the BPNN of the present invention has a structure of 12-10-4.
Step nine: the 450 groups of training data after the random grouping is conducted are imported into an artificial neural network model, and the model is trained;
Step ten: the test data are imported into a trained artificial neural network model, and error MSE between an output result obtained by the model and the corresponding test data is compared: calculated according to equation 1, where n=450;
Step eleven: comparing the MSE with the expected model precision value, and if the MSE is smaller than 10 -3, considering that model training is completed; if the MSE is larger than the precision expected value, the network parameters need to be continuously modified until the MSE is smaller than the precision expected value, and training is finished;
Step twelve: according to the method provided by the invention, finally, the trained artificial neural network model is utilized to verify the data samples of the three artificial radioactive sources of Am-241, cs-137 and Co-60 measured by using the unmanned plane radioactivity monitoring system at the position of 1 meter, so as to verify the network training quality.
It is apparent that the above examples are only illustrative of the present invention and are not limiting of the embodiments. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. It is not necessary here nor is it exhaustive of all embodiments. And obvious variations or modifications thereof are contemplated as falling within the scope of the present invention.

Claims (7)

1. The point source dose rate correction method based on the artificial neural network is characterized by comprising the following steps of:
s1, energy spectrum data and corresponding dose rate values of the unmanned aerial vehicle radioactivity monitoring system at different heights are obtained and are respectively used as input parameters and output parameters;
S2, dividing part of energy spectrum data at different heights into training data, dividing the other part of energy spectrum data into test data, constructing an artificial neural network model by utilizing the input parameters and the output parameters in the training data, and importing the training data into the artificial neural network model for training to obtain a trained artificial neural network model;
wherein, the method for obtaining the input parameters and the output parameters in the energy spectrum data with different heights and constructing the artificial neural network model comprises the following steps of,
A, carrying out normalization processing on energy spectrum data at different heights;
b, extracting the number of main components which mainly contribute to the deposition of the dose rate after normalization processing as an input parameter, wherein the method is that data related to the deposition dose rate of air in energy spectrum data at different heights are extracted; carrying out standardization processing on the related data to obtain standardized data; calculating correlation coefficients among the standardized data, and forming a correlation coefficient matrix; calculating each characteristic value of the correlation coefficient matrix; calculating the contribution rate and the accumulated contribution rate of each related data by using the characteristic values, and when the accumulated contribution rate is greater than a threshold value, taking the corresponding minimum quantity of characteristic values as the quantity of main components contributing mainly, namely the input parameters;
c, taking the air absorption dose rate of the radioactive source deposited in the air as an output parameter;
d, calculating an implicit layer by using the input parameter and the output parameter;
e, the structure of the artificial neural network model is that input parameters, hidden layers and output parameters are input, and the structure of the artificial neural network model is utilized to construct the model of the artificial neural network;
s3, respectively importing the test data into the trained artificial neural network model to obtain an ideal output result, and comparing errors between the ideal output result and the corresponding test data;
And if the error is larger than or equal to the set precision expected value, repeating the steps S2 and S3, and if the error is smaller than the set precision expected value, training the debugged artificial neural network model to be a point source dose rate correction algorithm.
2. The method for correcting the point source dose rate based on the artificial neural network according to claim 1, wherein the method for acquiring the energy spectrum of the unmanned aerial vehicle radioactivity monitoring system at different heights is as follows: monte Carlo software was used to simulate the energy spectrum of the unmanned aerial vehicle radioactivity monitoring system at different heights.
3. The artificial neural network-based point source dose rate modification method of claim 2, wherein the data related to the air deposition dose rate comprises:
The total energy peak count, energy, single escape peak, double escape peak, annihilation peak and corresponding measurement time and height of each nuclide.
4. The method for correcting the point source dose rate based on the artificial neural network as claimed in claim 3, wherein the method for normalizing the related data to obtain the normalized data uses the following formula:
Wherein x ij represents the j index of the i-th energy spectrum; data is shown after x ij has undergone a normalization operation; wherein the method comprises the steps of Is representative of the average of all of the energy spectrum data; s j is the standard deviation between the two data represented.
5. The artificial neural network-based point source dose rate correction method of claim 4, wherein the method of calculating the correlation coefficient between the normalized data is by using the following formula:
Wherein r ij is a correlation coefficient between the energy spectrum data x i and x j; is the average value of the ith energy spectrum data; Is the average value of the j-th energy spectrum data; n represents the number of spectrums used for calculation, x kj represents the j index of the k energy spectrum, and x ki represents the i index of the k energy spectrum.
6. The artificial neural network-based point source dose rate correction method of claim 5, wherein the precision expectation is 10 -3.
7. The artificial neural network-based point source dose rate correction method of claim 1, further comprising:
if the error is smaller than the precision expected value, measuring the dose rate of the same radioactive source at different heights by using an unmanned aerial vehicle radioactivity monitoring system in an actual scene, and comparing the dose rate with an actual output result to verify the quality of the artificial neural network model.
CN202110816233.9A 2021-07-20 2021-07-20 Point source dose rate correction method based on artificial neural network Active CN113536679B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110816233.9A CN113536679B (en) 2021-07-20 2021-07-20 Point source dose rate correction method based on artificial neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110816233.9A CN113536679B (en) 2021-07-20 2021-07-20 Point source dose rate correction method based on artificial neural network

Publications (2)

Publication Number Publication Date
CN113536679A CN113536679A (en) 2021-10-22
CN113536679B true CN113536679B (en) 2024-09-24

Family

ID=78100324

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110816233.9A Active CN113536679B (en) 2021-07-20 2021-07-20 Point source dose rate correction method based on artificial neural network

Country Status (1)

Country Link
CN (1) CN113536679B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113835111B (en) * 2021-11-26 2022-02-11 常州微亿智造科技有限公司 Unmanned aerial vehicle-based radioactive source positioning method and device

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110133703A (en) * 2019-05-15 2019-08-16 中国辐射防护研究院 A kind of gamma-rays Dosimetry based on BP neural network and Monte carlo algorithm
CN111539324A (en) * 2020-04-23 2020-08-14 重庆建安仪器有限责任公司 Novel nuclide identification method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102051576B1 (en) * 2018-03-07 2019-12-05 한국과학기술원 Apparatus and method for identifying multi-radioisotope based on plastic scintillator using Artificial Neural Network

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110133703A (en) * 2019-05-15 2019-08-16 中国辐射防护研究院 A kind of gamma-rays Dosimetry based on BP neural network and Monte carlo algorithm
CN111539324A (en) * 2020-04-23 2020-08-14 重庆建安仪器有限责任公司 Novel nuclide identification method

Also Published As

Publication number Publication date
CN113536679A (en) 2021-10-22

Similar Documents

Publication Publication Date Title
US9316753B2 (en) Neutron spectrometer
WO2017020857A1 (en) Formation elemental capture gamma ray spectrum
CN104330815A (en) Method applied to detection of air kerma conventional true value
CN107657119B (en) Measurement method for improving gamma energy spectrum data quality
JP7383114B2 (en) Method for monitoring radionuclide content in radioactive waste
CN113536679B (en) Point source dose rate correction method based on artificial neural network
CN113189636B (en) Aviation gamma energy spectrum analysis method based on multi-energy response
JP5703462B2 (en) Nuclear explosion identification method based on krypton and xenon isotopes
CN107092028A (en) A kind of computational methods of closed container nuclear material quality
Luo et al. Sequential probability ratio test using scaled time-intervals for environmental radiation monitoring
CN110687584B (en) LSTM-based rapid nuclide identification method
Shaver et al. Radiation detection computational benchmark scenarios
EP0042099A2 (en) Self-regulating neutron coincidence counter
Walt et al. Small-Angle Elastic Scattering of Fast Neutrons and the Electric Polarizability of the Neutron
Checchia et al. Muon tomography for spent nuclear fuel control
Moussa et al. Impact of Spontaneous Fission Neutron Emission Energy on Neutron Detector Response
Mengesha et al. Neutron Spectrometry for ${\rm UF} _6 $ Enrichment Verification in Storage Cylinders
Padilla et al. Prototype for Unclassified Algorithm
McFerran Surface Measurement Gamma-Ray Spectroscopy Variation Characterization of Filled Uranium Hexafluoride Cylinders
Kulisek et al. Lead Slowing-Down Spectrometry Time Spectral Analysis for Spent Fuel Assay: FY12 Status Report
Ivarsson Biebel et al. Calibration of Coincidence Gamma Spectrometry Detector GeCo
Artioli The light W-PIE neutron spectrometer for soil moisture measurements: design, characterization and preliminary measurements
CN116090180A (en) Gamma energy spectrum continuous substrate estimation method, storage medium and system
Hazzaa Applying Monte Carlo and Artificial Intelligence Techniques for 235U Mass Prediction in Samples with Different Enrichments Sameh E. Shaban1
Abd El Gawad et al. Uranium Enrichment Measurement Using Enrichment-Meter Approach

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
CB03 Change of inventor or designer information
CB03 Change of inventor or designer information

Inventor after: Wang Zeyu

Inventor after: Mo Hong

Inventor after: Tang Xiaobin

Inventor after: Gong Pin

Inventor after: Wang Peng

Inventor after: Liang Dajian

Inventor after: Zhou Cheng

Inventor after: Zhu Xiaoxiang

Inventor after: Jiang Ruocheng

Inventor before: Tang Xiaobin

Inventor before: Gong Pin

Inventor before: Mo Hong

Inventor before: Wang Peng

Inventor before: Liang Dajian

Inventor before: Wang Zeyu

Inventor before: Zhou Cheng

Inventor before: Zhu Xiaoxiang

Inventor before: Jiang Ruocheng

GR01 Patent grant
GR01 Patent grant