A method, storage medium, and system for estimating the continuous background of gamma spectra

Abstract

The present invention relates to a method for estimating the continuous background of gamma energy spectrum, which includes the steps of: generating gamma energy spectrum data for training by using Monte Carlo simulation; generating the corresponding continuous background according to the simulated energy spectrum data; normalizing the counts of each channel of the gamma energy spectrum and its continuous background; building a learning model with the counts of each channel of the gamma energy spectrum as the input and the counts of the continuous background of each channel of the corresponding gamma energy spectrum as the output; training the built model by using the stochastic gradient descent algorithm, and obtaining the count values of each channel of the continuous background of the gamma energy spectrum through the trained model. The present invention also provides a storage medium and a gamma energy spectrum continuous background estimation system. The use of a method, a storage medium and a system for estimating the continuous background of gamma energy spectrum can make up for the accuracy defects of existing methods and improve the accuracy of the calculation of the net count of the full energy peak of the gamma energy spectrum and the subsequent analysis of the content of radionuclides.

Description

Technical Field

[0001] The present invention belongs to the field of nuclear radiation detection, and particularly relates to a method for estimating a continuous background of γ energy spectrum, a storage medium, and a system. Background Art

[0002] A γ energy spectrum is a signal count statistical distribution map (arranged from small to large according to signal amplitude) formed by the energy deposited by γ rays emitted by radionuclides in a detector. Since the signal amplitude is proportional to the energy deposited by γ rays in the detector, the γ energy spectrum is a statistical distribution map of the number of γ rays according to energy. This distribution map is discrete, and its abscissa is an energy interval, or is called a "channel". Each channel is usually numbered with a natural number, and this number is called a "channel address", that is, the address of the channel. Through energy calibration, the "channel address" can be converted into an energy value.

[0003] γ energy spectrum analysis is an important way to know the types and contents of radioactive substances in a sample, and this is mainly achieved by means of full-energy peaks in the γ energy spectrum. A full-energy peak is formed by the photoelectric effect of γ rays in the detector losing all energy, and there is a definite quantitative relationship between it and the content of radionuclides in the measurement object, that is: the net count of the full-energy peak corresponds to the nuclide activity multiplied by known factors (measurement time, γ-ray branching ratio, and detection efficiency). The relationship between the net count of the full-energy peak and the count of the original γ energy spectrum can be expressed as: net count of the full-energy peak = original γ energy spectrum - continuous background. The continuous background is the count caused by the energy deposited by γ rays through Compton scattering in the detector. It can be seen that the determination of the continuous background count is the key to obtaining the net count of the full-energy peak and then realizing the quantitative analysis of the content of radionuclides. However, in the measured γ energy spectrum, the net count of the full-energy peak and the continuous background count are superimposed together and cannot be directly subtracted from the original spectrum count. An algorithm must be used to estimate the continuous background. The steps of the existing continuous background estimation method are as follows: a) Determine the full-energy peak region in the original γ energy spectrum (generally using the method of adding and subtracting 1.5 times the full width at half maximum from the peak center); b) Estimate the continuous background shape within the full-energy peak region using a linear or polynomial function; c) Based on the continuous background shape estimated in b), combined with the counts at both ends of the full-energy peak region, calculate the continuous background count of each channel within the region, and the continuous background count outside the region is directly taken as the original spectrum count value.

[0004] However, since the width of the full-energy peak region varies with the change of the peak shape, there are often errors in estimating the width using the full width at half maximum method with fixed parameters. At the same time, the continuous background shape within the full-energy peak region is extremely complicated due to the influence of multiple Compton scatterings, peak overlap effects, etc., and it cannot be accurately fitted using a linear or polynomial function. For the above reasons, the existing continuous background estimation method has a large error, which in turn causes inaccurate analysis of the content of radionuclides. Summary of the Invention

[0005] Aiming at the defects existing in the prior art, the purpose of the present invention is to provide a γ-ray spectrum continuous background estimation method, a storage medium and a system, to make up for the accuracy defects of the existing methods and improve the accuracy of the net count calculation of the full-energy peak and the subsequent analysis of the radionuclide content.

[0006] To achieve the above object, the technical solution adopted by the present invention is: a γ-ray spectrum continuous background estimation method, including the steps of: generating γ-ray spectrum data for training by Monte Carlo simulation; generating the corresponding continuous background according to the simulated spectrum data; normalizing the counts of each channel of the γ-ray spectrum and its continuous background; building a learning model with the counts of each channel of the γ-ray spectrum as the input and the counts of the continuous background of each channel of the corresponding γ-ray spectrum as the output; training the built model by using the stochastic gradient descent algorithm, and obtaining the count values of each channel of the γ-ray spectrum continuous background through the trained model.

[0007] Further, the step of generating the corresponding continuous background according to the simulated spectrum data includes: obtaining the unbroadened γ-ray spectrum; replacing the channels corresponding to the energies of the γ-ray source particles in the obtained unbroadened γ-ray spectrum; calculating the counts of each channel of the continuous background of the replaced γ-ray spectrum.

[0008] Further, the way of replacing the channels corresponding to the energies of the γ-ray source particles in the obtained unbroadened γ-ray spectrum is:

[0009] x i =(x i-1 +x i+1 ) / 2

[0010] where x i represents the count of the channel corresponding to the energy of the γ-ray source particle in the γ-ray spectrum, and i represents the channel address of this channel.

[0011] Further, the way of calculating the counts of each channel of the continuous background of the replaced γ-ray spectrum is:

[0012]

[0013]

[0014]

[0015]

[0016] where y i is the count of each channel of the continuous background, E i is the energy corresponding to the i-th channel, with the unit of keV; f is the energy resolution of the detector; W is the channel width, with the unit of keV.

[0017] Further, the method for normalizing the counts of each channel of the γ energy spectrum and its continuous background is as follows:

[0018] x′ i = x i / max i (x i ) i = 1, …, n

[0019] y′ i = y i / max i (x i ) i = 1, …, n

[0020] where x i , x’ i represent the original γ energy spectrum and the normalized count values of each channel respectively, and y i , y’ i represent the original continuous background and the normalized count values of each channel respectively, and max represents the maximum value function.

[0021] Further, the learning model is a fully connected neural network.

[0022] Further, the total number of layers of the fully neural network is 5. The number of neurons in the input layer is equal to the number of channels of the γ energy spectrum, and the neuron values are equal to the normalized count values of each channel of the γ energy spectrum. The number of neurons in the output layer is equal to the number of channels of the γ energy spectrum, and the neuron values are equal to the normalized count values of each channel of the continuous background. The activation function of each layer is the Relu function; the normalization function of the output layer is the Sigmoid function; the loss function of the learning machine is the sum of squares function.

[0023] Further, when using the stochastic gradient descent algorithm to train the constructed model, the batch training size is 1000, the learning rate is 0.001, and the number of iterations is 30.

[0024] The present invention also provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, a method for estimating the continuous background of the γ energy spectrum is implemented.

[0025] The present invention also provides a system for estimating the continuous background of the γ energy spectrum, including: a simulation unit for generating γ energy spectrum data for training by using Monte Carlo simulation; a continuous background generation unit for generating the corresponding continuous background according to the simulated energy spectrum data; a processing unit for normalizing the counts of each channel of the γ energy spectrum and its continuous background; a model construction unit for constructing a learning model with the counts of each channel of the γ energy spectrum as the input and the counts of each channel of the continuous background of the corresponding γ energy spectrum as the output; and a training execution unit for training the constructed model by using the stochastic gradient descent algorithm and obtaining the count values of each channel of the continuous background of the γ energy spectrum through the trained model.

[0026] The effects of the present invention are as follows: By using Monte Carlo simulation to generate gamma ray energy spectrum data for training, and generating its corresponding continuous base according to the simulated energy spectrum data, using the count of each channel of the gamma ray energy spectrum as the input and the count of the continuous base of each channel of the corresponding gamma ray energy spectrum as the output, a learning model is built, thereby making up for the accuracy defects of existing methods and improving the accuracy of the calculation of the net count of the full energy peak of the gamma ray energy spectrum and the subsequent analysis of the content of radionuclides. Description of the Drawings

[0027] Figure 1 It is a flow chart of the steps of a method for estimating the continuous base of a gamma ray energy spectrum in the present invention. Detailed Embodiment

[0028] The present invention will be further described below in conjunction with the drawings and specific embodiments.

[0029] As Figure 1 shown, the present invention provides a method for estimating the continuous base of a gamma ray energy spectrum, including the steps:

[0030] S1. Use Monte Carlo simulation to generate gamma ray energy spectrum data for training;

[0031] Specifically, for a specific gamma ray energy spectrum measurement scenario in practical applications, use Monte Carlo simulation software to model it and generate a gamma ray energy spectrum through simulation.

[0032] In this embodiment, a random integer (range 1-10) is used to determine the total number of types of gamma rays. For each type of gamma ray, a random floating point number (range 0.01-3.00, unit MeV) is used to determine its specific energy, and a random floating point number (range 1.00-10.00) is used to determine its relative sampling probability. The reason for this is that in the conventional actual measurement in the field of radiation detection, the types of radionuclides are generally 1-5, and the gamma rays emitted by them generally do not exceed 10. For common nuclides such as 137 Cs, 60 Co, etc., their energies are all between 0.01-3.00 MeV. In addition, among different gamma rays, if the quantity of one type is less than 1 / 10 of that of others, then this ray can generally be ignored. Therefore, the gamma ray energy spectrum samples generated in this way can better cover the diversity factors of the types of radionuclides in actual measurement.

[0033] It should be noted that the detection scenario simulated by Monte Carlo is a φ3×10 cm NaI(Tl) detector detecting a 1 MeV gamma ray point source 40 cm away from its end face without surrounding medium.

[0034] It should be noted that the computer Monte Carlo simulation uses MCNP software.

[0035] S2. Generate the corresponding continuous base according to the simulated energy spectrum data.

[0036] Specifically, the process of generating the continuous base is divided into:

[0037] First, obtain the unbroadened γ energy spectrum.

[0038] In this embodiment, the unbroadened energy spectrum is generated by using the same computer Monte Carlo simulation method as in step S1 but without energy spectrum broadening.

[0039] Secondly, replace the channels corresponding to the γ-ray source particle energies in the obtained unbroadened γ energy spectrum.

[0040] In this embodiment, the replacement formula is:

[0041] x i =(x i-1 +x i+1 ) / 2

[0042] where x i represents the channel count corresponding to the γ-ray source particle energy in the γ energy spectrum, and i represents the channel address of this channel.

[0043] Finally, calculate the channel count of each channel of the continuous base of the replaced γ energy spectrum.

[0044] In this embodiment, the calculation method is:

[0045]

[0046]

[0047]

[0048]

[0049] where y i is the channel count of each channel of the continuous base, E i is the energy corresponding to the i-th channel, with the unit of keV; f is the energy resolution of the detector; W is the channel width, with the unit of keV.

[0050] It can be understood that the real reason why the continuous background cannot be directly subtracted in actual measurement is that each channel of the γ spectrum is broadened by the detector, so that the broadened full-energy peak region is superimposed on the complex continuous background; assuming that the spectrum is not broadened, the full-energy peak region only occupies a single channel, and the continuous background of this channel can be conveniently calculated from the average value of the left and right channels. In actual measurement, due to the inherent characteristics of the detector, the γ spectrum is always broadened, so the unbroadened spectrum cannot be obtained; however, it is different when using computer Monte Carlo simulation, and the broadening state of the spectrum can be conveniently changed by controlling the simulation parameters. Therefore, the above method can obtain the continuous background count of the full spectrum by simulating the unbroadened spectrum, estimating the continuous background of the full-energy peak region with the average value of the left and right channels, and finally manually re-broadening the spectrum.

[0051] S3. Normalize the count of each channel of the γ spectrum and its continuous background;

[0052] Specifically, the processing method is as follows:

[0053] x i ′ = x i / max i (x i ) i = 1, …, n

[0054] y i ′ = y i / max i (x i ) i = 1, …, n

[0055] Among them, x i , x’ i respectively represent the count values of each channel of the original γ spectrum and its normalized values, y i , y’ i respectively represent the count values of each channel of the original continuous background and its normalized values, and max represents the maximum value function.

[0056] S4. Build a learning model with the count of each channel of the γ spectrum as the input and the count of the continuous background of each corresponding channel of the γ spectrum as the output;

[0057] Specifically, use computer code to construct a fully connected neural network: the total number of layers is 5 (including the input layer and the output layer), the number of neurons in the input layer is equal to the number of channels of the original γ spectrum, the neuron values are equal to the normalized count values of each channel of the original γ spectrum, the number of neurons in the output layer is equal to the number of channels of the original γ spectrum, the neuron values are equal to the normalized count values of each channel of the continuous background, and the activation function of each layer is the Relu function; the normalization function of the output layer is the Sigmoid function; the loss function of the learning machine is the sum of squares function.

[0058] It can be understood that the fully-connected neural network model is used because there is a global correlation between the original γ energy spectrum and the channel counts of its continuous background, and the fully-connected neural network is suitable for global correlation representation learning. According to the non-negative feature of the channel counts of the γ energy spectrum, the Relu function is used as the activation function for each layer; according to the characteristic that the channel counts of the output layer are between [0, 1], the Sigmoid function is used as the normalization function; to evaluate the overall closeness between the predicted continuous background counts and the true values, the sum of squares function is used as the loss function of the learner.

[0059] S5. The constructed model is trained using the stochastic gradient descent algorithm, and the channel counts of the continuous background of the γ energy spectrum are obtained through the trained model.

[0060] Specifically, during training, the stochastic gradient descent algorithm is adopted. In this embodiment, the batch training size is taken as 1000, the learning rate is taken as 0.001, and the number of iterations is taken as 30.

[0061] After training is completed, the γ energy spectrum normalized in step S3 is input into the model, and the channel counts of its continuous background can be obtained.

[0062] It should be noted that in this embodiment, the computer code for training and recognition of the learner is written in python.

[0063] The present invention also provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of implementing a method for estimating the continuous background of a γ energy spectrum are realized.

[0064] It should be noted that the storage medium shown in this application can be a computer-readable signal medium or a storage medium or any combination of the two. The storage medium can be, for example - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the storage medium can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this application, the storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, apparatus, or device. And in this application, the storage medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The storage medium can also be any computer-readable medium other than the storage medium, and this computer-readable medium can send, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0065] The present invention also provides a gamma energy spectrum continuous background estimation system, which includes:

[0066] An analog unit for generating gamma energy spectrum data for training by using Monte Carlo simulation;

[0067] A continuous background generation unit for generating its corresponding continuous background according to the simulated energy spectrum data;

[0068] A processing unit for normalizing the count of each channel of the gamma energy spectrum and its continuous background;

[0069] A model building unit for building a learning model with the count of each channel of the gamma energy spectrum as the input and the count of each channel of the continuous background of the corresponding gamma energy spectrum as the output;

[0070] A training execution unit for training the built model by using the stochastic gradient descent algorithm and obtaining the count value of each channel of the gamma energy spectrum continuous background through the trained model.

[0071] As can be seen from the above embodiments, the beneficial effects of the present invention are as follows: By using Monte Carlo simulation to generate gamma (γ) energy spectrum data for training, and based on the simulated energy spectrum data, generating its corresponding continuous baseline, using the counts of each channel of the γ energy spectrum as input and the counts of the continuous baseline of each channel of the corresponding γ energy spectrum as output, a learning model is built, thereby making up for the accuracy defects of existing methods and improving the accuracy of the net count calculation of the full-energy peak of the γ energy spectrum and the subsequent analysis of the content of radionuclides.

[0072] The present invention is not limited to the embodiments described in the specific embodiments. Those skilled in the art can obtain other embodiments according to the technical solution of the present invention, which also belong to the scope of the technical innovation of the present invention.

Claims

1. A method for estimating the continuous background of gamma energy spectrum, characterized in that, Including: Generating γ energy spectrum data for training by using Monte Carlo simulation; Generating its corresponding continuous base according to the simulated energy spectrum data; Normalizing the count of each channel of the γ energy spectrum and its continuous base; Constructing a learning model with the count of each channel of the γ energy spectrum as the input and the count of each channel of the continuous base of the corresponding γ energy spectrum as the output; Training the constructed model by using the stochastic gradient descent algorithm, and obtaining the count value of each channel of the continuous base of the γ energy spectrum through the trained model.

2. A method for estimating the continuous base of γ energy spectrum according to claim 1, characterized in that: The generating its corresponding continuous base according to the simulated energy spectrum data includes the steps of: Obtaining the unbroadened γ energy spectrum; Replacing the channels corresponding to the energies of γ-ray source particles in the obtained unbroadened γ energy spectrum; Calculating the count of each channel of the continuous base of the γ energy spectrum after replacement.

3. A method for estimating the continuous base of γ energy spectrum according to claim 2, characterized in that: The way of replacing the channels corresponding to the energies of γ-ray source particles in the obtained unbroadened γ energy spectrum is: x i = (x i-1 + x i+1 ) / 2 where x i represents the channel count corresponding to the energy of the γ-ray source particles in the γ energy spectrum, and i represents the channel address of this channel.

4. A method for estimating the continuous base of γ energy spectrum according to claim 2, characterized in that: The way of calculating the count of each channel of the continuous base of the γ energy spectrum after replacement is: Among them, y i is the count of each channel of the continuous base, E i is the energy corresponding to the i-th channel, with the unit of keV; f is the detector energy resolution; W is the channel width, with the unit of keV.

5. A method for estimating the continuous base of γ energy spectrum according to claim 1, characterized in that: The way of normalizing the count of each channel of the γ energy spectrum and its continuous base is: x' i = x i / max i (x i ) for i = 1, …, n y′ i = y i / max i (x i ) i = 1, …, n where x i , x' i represent the original γ energy spectrum and the count values of each channel after its normalization respectively, y i , y' i represent the original continuous background and the count values of each channel after its normalization respectively, and max represents the maximum value function.

6. A method for estimating the continuous base of γ energy spectrum according to claim 1, characterized in that: The learning model is a fully connected neural network.

7. A method for estimating the continuous base of γ energy spectrum according to claim 6, characterized in that: The total number of layers of the fully neural network is 5. The number of neurons in the input layer is equal to the number of channels of the γ energy spectrum, and the neuron value is equal to the normalized count of each channel of the γ energy spectrum. The number of neurons in the output layer is equal to the number of channels of the γ energy spectrum, and the neuron value is equal to the normalized count of each channel of the continuous base. The activation function of each layer is the Relu function; the normalization function of the output layer is the Sigmoid function; the loss function of the learning machine is the sum of squares function.

8. A method for estimating the continuous base of γ energy spectrum according to claim 1, characterized in that: When training the constructed model by using the stochastic gradient descent algorithm, the batch training size is taken as 1000, the learning rate is taken as 0.001, and the number of iterations is taken as 30.

9. A storage medium, on which a computer program is stored, characterized in that: The computer program, when executed by a processor, implements a method for estimating the continuous base of γ energy spectrum according to any one of claims 1 to 8.

10. A γ-ray energy spectrum continuous background estimation system, characterized in that, Including: A simulation unit for generating γ energy spectrum data for training by using Monte Carlo simulation; A continuous base generation unit for generating its corresponding continuous base according to the simulated energy spectrum data; A processing unit for normalizing the count of each channel of the γ energy spectrum and its continuous base; A model construction unit for constructing a learning model with the count of each channel of the γ energy spectrum as the input and the count of each channel of the continuous base of the corresponding γ energy spectrum as the output; A training execution unit for training the constructed model by using the stochastic gradient descent algorithm, and obtaining the count value of each channel of the continuous base of the γ energy spectrum through the trained model.

Images