Gamma spectrum separation system and method for whole body contamination (WBC) test equipment for generating artificial neural network (ANN) learning data
The gamma spectrum separation system for the WBC equipment addresses the challenge of separating single nuclide spectra from mixed nuclide spectra by calculating efficiencies and count rate contributions, resulting in improved ANN learning data and reduced RMSE values for enhanced nuclide identification.
Patent Information
- Application Number
- PCT/KR2024/003590
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-03
- Filing Date
- 2024-03-22
- Publication Date
- 2024-11-07
AI Technical Summary
The existing Whole Body Contamination Test Equipment (WBC) faces challenges in accurately separating the gamma spectrum of a single nuclide from a mixture of nuclides due to poor resolution, leading to increased Root Mean Square Error (RMSE) values, which affects the accuracy of radioactive contamination measurements in nuclear power plant workers.
A gamma spectrum separation system and method that calculates the efficiency and count rate contribution for each channel, normalizes the spectrum, and subtracts the count rate contribution of interfering nuclides to isolate the spectrum of a single nuclide, generating artificial neural network (ANN) learning data for improved nuclide identification.
This approach enables the generation of accurate synthetic spectra for ANN training, enhancing the accuracy of nuclide identification and reducing RMSE values, thereby improving the diagnostic capabilities of the WBC equipment.
Smart Images

Figure KR2024003590_07112024_PF_FP_ABST
Abstract
Description
A gamma spectrum separation system and method for generating training data for artificial neural networks (ANNs) using a whole-body contamination test device (WBC).
[0001] The present invention relates to a gamma spectrum separation system for a whole-body contamination tester (WBC) for generating artificial neural network (ANN) learning data, and more specifically, to a gamma spectrum separation system for a whole-body contamination tester (WBC) for generating artificial neural network (ANN) learning data by separating a spectrum of a single nuclide from a spectrum in which various nuclides are mixed.
[0002] The body-wide contamination detection equipment is a device that measures the level of radioactive contamination by detecting gamma rays emitted from radioactive nuclides when radiation workers in nuclear power plants are exposed to internal radiation.
[0003] Figure 1 is the FastScan from Mirion, the most widely used whole-body contamination test equipment in nuclear power plants, and has an internal size of 7.6×12.7×40.6 cm. 3 Two NaI(Tl) detectors of the same size are built in series, one above the other. Figure 2 shows the results measured in the body whole body contamination test equipment. 134 Cs, 137 Cs, 54 Mn, 57 Co, 58 Co, 60 Co is a gamma-ray spectrum.
[0004] Table 1 shows the interference nuclides in the lung area of the "V. Fission and radioactivity products" category of the "Performance Criteria for Radiobioassay" of the US ANSI / HPS N13.30-2011, which is a performance standard for whole-body contamination testing. 134 Cs, 137 When there is Cs) 54 Mn, 57 Co, 58 Co, 60 It is required that the results of measuring two random nuclides from Co satisfy RMSE (Root Mean Square Error) ≤ 0.25.
[0005] [Table 1]
[0006]
[0007] Here, RMSE is is expressed as
[0008]
[0009] Here, Among the opponents, Among the individual counterparts, is relative precision, is the Ith measured radioactivity, and N is the number of measurements.
[0010] FastScan's NaI(Tl) detector has excellent gamma-ray detection efficiency, allowing for a short measurement time of approximately 1 to 3 minutes. However, its resolution is poor, so when the target nuclide is affected by interfering nuclides, it may not satisfy the RMSE criterion. For example, in Fig. 2 54 The 835 keV peak of Mn or 58 The 811 keV peak of Co is an interfering nuclide. 134 When overlapped with the 796 keV peak of Cs, the peak analysis accuracy decreases and the RMSE value increases.
[0011] To improve this, the spectrum is analyzed using an artificial neural network, which is one of the artificial intelligence technologies. However, since it is difficult to obtain a spectrum for training the artificial neural network through actual measurements, the spectrum is sometimes obtained by synthesizing a single-nuclide spectrum.
[0012] For example, the black line in Figure 3 54 Mn, 60 Co, 134 Cs, 137 This is a spectrum synthesized from the single spectrum of each Cs nuclide at ratios of 0.1, 0.31, 0.24, and 0.35, and the red line is 54 Mn, 60 Co, 134 Cs, 137 This is an actual spectrum measured simultaneously with Cs.
[0013] [Prior Art Literature]
[0014] [Patent Document] Korean Patent No. 2249120 (April 30, 2021)
[0015] The present invention aims to solve the above-described problem by providing a whole-body contamination test (WBC) gamma spectrum separation system and method for generating training data for an artificial neural network (ANN) that detects gamma rays emitted from radioactive nuclides when a radiation worker in a nuclear power plant is exposed to radiation and measures the level of radiation contamination.
[0016] In order to achieve the above object, the present invention includes an efficiency calculation unit (100) for calculating the efficiency of channels A, B, and C of a body-wide contamination inspection device, a counting rate contribution calculation unit (200) for calculating the counting rate contribution by channel by multiplying the value obtained by multiplying the gamma-ray emission rate of each nuclide by the mixed source for generating artificial neural network learning data by the efficiency of each channel, and a normalization unit (300) for obtaining a normalized spectrum of the nuclide by subtracting the counting rate contribution by channel from the counting rate by channel measured by the nuclide and then normalizing the result.
[0017] According to the present invention, there is an effect that can be used to generate a synthetic spectrum for artificial intelligence learning by separating a spectrum of a single nuclide from a spectrum in which various nuclides are mixed.
[0018] Figure 1 is a drawing showing a body-wide contamination test device.
[0019] Figure 2 is a graph showing the gamma ray spectrum of the body's whole body contamination inspection equipment.
[0020] Figure 3 is a graph showing a comparison between a synthetic spectrum and an actual measured spectrum.
[0021] Fig. 4 is 58 Mixed in Co 57 Co, 60 This is a graph showing the Co spectrum.
[0022] Fig. 5 is 58 Co, 57 Co, 60In Co mixed nuclides 57 Co, 60 By removing Co 58 This is the result of obtaining the Co single spectrum.
[0023] FIG. 6 is a configuration diagram showing a gamma spectrum separation system for a whole body contamination test device (WBC) for generating artificial neural network (ANN) learning data according to one embodiment of the present invention.
[0024] FIG. 7 is a schematic flowchart of a method using a whole body contamination test (WBC) gamma spectrum separation system for generating artificial neural network (ANN) learning data according to one embodiment of the present invention.
[0025] Hereinafter, the present invention will be described in detail with reference to the contents described in the attached drawings. However, the present invention is not limited or restricted by the exemplary embodiments. The same reference numerals in each drawing indicate components that perform substantially the same functions.
[0026] The purpose and effects of the present invention can be naturally understood or made clearer by the following description, and the purpose and effects of the present invention are not limited solely by the following description. Furthermore, in describing the present invention, if a detailed description of known technologies related to the present invention is deemed to unnecessarily obscure the gist of the present invention, such detailed description will be omitted.
[0027] The present invention relates to a method for separating a single-nuclide spectrum from a mixed-nuclide spectrum to generate training data for an artificial neural network (ANN) of a whole-body contamination test (WBC) device. shows the purchased data for generating training data for an artificial neural network. 58 I am a sailor, but my sailor certificate says 58 Not only Co 6,570 Bq 57 Co 723 Bq and 60 Co 670 Bq is mixed.
[0028] Table 3 is 58Co indicates the radioactivity of each nuclide in the crew certificate.
[0029] [Table 3]
[0030]
[0031] Figure 4 shows the normalized results of the gamma spectrum measured by the whole-body contamination test equipment 30 days after the reference date for the nuclides in Table 3. Figure 4 58 Mixed in Co 57 Co, 60 This is a graph showing the Co spectrum.
[0032] Although the gamma peaks for each nuclide are separated, the count rates for each channel are a mixture of the three nuclides in a specific ratio.
[0033] The purpose of the present invention is 58 Co, 57 Co, 60 In the spectrum where Co is mixed 58 The technical principle of the method of separating only Co is as follows.
[0034] first 57 Co and 60 Find the efficiency of each channel of the whole-body contamination test equipment for each Co.
[0035] Here 57 Co and 60 Co nuclide is a single nuclide purchased separately from Table 3, and the radioactivity of other nuclides is not mixed.
[0036] since 58 Mixed in Co 57 Co and 60 The product of the radioactivity of Co and the gamma-ray emission rate of each nuclide 57 Co and 60 If we multiply the efficiency of each channel of Co 58 Mixed in Co 57 Co and 60 The channel-wise counting rate contribution of Co can be calculated from the channel-wise counting rates measured by the three nuclides in Table 3. 57 Co and 60After subtracting the coefficient contribution per channel of Co and normalizing it, it is necessary for learning the artificial neural network. 58 Co's single spectrum can be obtained.
[0037] A detailed description of the composition and overall operating principle of the invention is as follows.
[0038] In order to obtain the normalized spectrum of only nuclide A from the spectrum of mixed nuclides A, B, and C, the counting rate and efficiency for each channel of nuclide B and nuclide C must be obtained.
[0039] Afterwards, by subtracting the channel-specific counting rate contributions of nuclides B and C from the channel-specific counting rates of the mixed nuclides A, B, and C and then normalizing, a normalized spectrum of nuclide A can be obtained.
[0040] a) The method for calculating the channel efficiency of nuclide B is as follows.
[0041] Channel-specific efficiency of nuclide B ( ) is as follows. Here, B nuclide is 57 Co, C nuclides 60 Let's assume Co.
[0042]
[0043] Here is the net count rate per channel (cps), A is the radioactivity per nuclide (Bq), and Y is the gamma-ray emission rate per nuclide.
[0044] The whole-body contamination detection equipment measures gamma-ray energy from 0 to 2 MeV, divided into 512 or 1,024 channels. For example, single nuclide 57 If the channel 20 count rate of Co 3,826 Bq is 0.1739 cps, 57 Since the gamma ray emission rate (Y) of Co is 0.9628, 57 The efficiency of channel 20 for Co is 4.72E-5. The efficiency of channels 1 to 1,024 can be obtained in the same way.
[0045] A) The method for calculating the channel-specific count rate contribution of nuclide B in a mixed nuclide of A, B, and C is as follows.
[0046] Channel-specific count rate of nuclide B in the mixed nuclides A, B, and C ( ) The contribution can be calculated by the following formula.
[0047]
[0048] Here, is the efficiency of each channel (i) of the B nuclide, is the radioactivity of nuclide B in the mixed nuclide of A, B, and C, is the gamma-ray emission rate of the B nuclide.
[0049] In Table 3 58 Mixed in Co 57 The radioactivity of Co is 670 Bq and the efficiency () in channel 20 of the whole body contamination test equipment ) is 4.72E-5, so = 670 Bq x 0.9628 x It becomes.
[0050] By the above method 58 Mixed in Co 57 Not only Co 60 The contribution of the coefficient rate per channel of Co can be obtained.
[0051] In order to obtain the normalized spectrum of nuclide A, the channel-specific counting rate contributions of nuclides B and C must be subtracted from the channel-specific counting rate of nuclide A and then normalized. Figure 5 58 Co, 57 Co, 60 In Co mixed nuclides 57 Co, 60 By removing Co 58 This is the result of obtaining the Co single spectrum. Figure 5 is 57 Co, 60 This is a graph showing a comparison before and after Co spectrum removal.
[0052] According to the present invention, 58 Co sailor has a half-life of 70.85 days.57 Co (271.81 days) and 60 It is very short compared to Co (1,925.28 days), so as time goes by 57 Co and 60 The ratio of Co increases. Using the present invention, the spectrum of a single nuclide can be separated from a spectrum containing a mixture of various nuclides, and used to generate a synthetic spectrum for artificial intelligence learning.
[0053] FIG. 6 is a configuration diagram showing a gamma spectrum separation system for a whole body contamination test device (WBC) for generating artificial neural network (ANN) learning data according to one embodiment of the present invention.
[0054] As illustrated in FIG. 6, the gamma spectrum separation system (10) of a whole body contamination test device (WBC) for generating artificial neural network (ANN) learning data according to the present embodiment includes an efficiency calculation unit (100), a counting rate contribution calculation unit (200), and a normalization unit (300).
[0055] The body-wide contamination test (WBC) gamma spectrum separation system (10) is a device that analyzes and classifies nuclides using an artificial neural network. This system is largely divided into three parts.
[0056] The efficiency calculation unit (100) calculates the gamma ray detection efficiency for each of the A, B, and C channels of the WBC equipment. This calculated efficiency value is later used in the counting rate contribution calculation unit.
[0057] The efficiency calculation unit (100) according to the present embodiment calculates the gamma-ray detection efficiency for each of the A, B, and C channels of the WBC equipment. Through this, the gamma-ray detection efficiency to be used later in the counting rate contribution calculation unit can be calculated.
[0058] In WBC equipment, channels A, B, and C each detect gamma rays in different energy ranges. The detection efficiency of these channels is calculated by comparing measured data with actual gamma-ray emissions. Specifically, the gamma-ray emissions are calculated based on the measured values from a predetermined set of gamma rays projected onto the equipment, and the detection efficiency for each channel, A, B, and C, is calculated based on this data.
[0059] Here, to explain further, the efficiency of the A, B, and C channels of the body contaminant detection equipment (WBC) is a value that represents the efficiency of gamma rays detected in each channel, and this can be obtained using the measured count rate (cps) for each channel, the radioactivity (Bq) for each nuclide, and the gamma-ray emission rate for each nuclide.
[0060] First, the measured channel-specific count rate (cps) is a value representing the intensity of gamma rays measured by the body-body contamination test (WBC) equipment. The radioactivity (Bq) of a nuclide represents the amount of radiation energy emitted by that nuclide per unit time, and the gamma-ray emission rate by nuclide represents the number of gamma rays emitted by that nuclide per unit time.
[0061] Based on this information, the channel-specific count rate contribution is calculated by multiplying the value obtained by multiplying the gamma-ray emission rate of each nuclide by the channel-specific efficiency for the mixed source for generating artificial neural network training data. This allows us to obtain the count rate value contributed by each nuclide in the corresponding channel. Then, the channel-specific count rate contribution is subtracted from the channel-specific count rate measured by the nuclide and normalized to obtain the normalized spectrum of the nuclides. This allows us to obtain the normalized count rate value contributed by each nuclide in the corresponding channel. Finally, based on these values, the efficiencies of the whole-body contamination screening (WBC) equipment for channels A, B, and C can be calculated. The efficiency of the gamma-rays detected in each channel is calculated by dividing the net count rate (cps) for each channel by the activity (Bq) for each nuclide and the gamma-ray emission rate for each nuclide, divided by the channel-specific count rate contribution. This allows us to obtain the gamma-ray detection efficiencies in channels A, B, and C.
[0062] This efficiency calculation process is used in the counting rate contribution calculation and normalization sections of the gamma spectral separation system. The accuracy of this process is crucial for obtaining accurate results for these tasks. Therefore, the WBC gamma spectral separation system precisely performs and verifies this process.
[0063] The counting rate contribution calculation unit (200) calculates how much each channel contributes to the counting rate of a nuclide by multiplying the gamma-ray emission rate of each nuclide in the mixed source by the efficiency of each channel. This is used to analyze the counting rate of each nuclide.
[0064] The counting rate contribution calculation unit (200) according to the present embodiment is a configuration used in a gamma spectrum separation system (10) of a body-wide contamination test device (WBC). This configuration calculates the counting rate contribution of gamma rays emitted from each nuclide required for generating artificial neural network learning data.
[0065] To achieve this, the counting rate contribution for each channel is first calculated by multiplying the gamma-ray emission rate of each nuclide in the mixed source by the gamma-ray detection efficiency for channels A, B, and C calculated in the efficiency calculation unit. The counting rate contribution calculated in this way is then added for all nuclides and channels, and normalized for each nuclide to obtain the counting rate contribution. This allows the construction of gamma-ray spectral data for generating artificial neural network training data.
[0066] The count rate contribution calculation unit plays a crucial role in the WBC gamma spectrum separation system. To generate training data for an artificial neural network, the count rate contribution of gamma rays emitted by each nuclide must be accurately determined. This requires precise calculation and measurement processes. Therefore, the count rate contribution calculation unit performs precise calculations and measurements, thereby obtaining the accurate data required for generating training data for an artificial neural network.
[0067] Additional explanation of the coefficient contribution calculation section is as follows.
[0068] The gamma-ray spectrum measured by the whole-body contamination detection equipment is composed of the sum of gamma rays emitted by multiple nuclides. Therefore, to train an artificial neural network, the gamma-ray spectrum must be separated by nuclide to generate training data. The count rate contribution calculation unit calculates the count rate contribution for each channel by multiplying the gamma-ray emission rate of each nuclide by the channel efficiency, which is the product of the mixed radiation source and the gamma-ray emission rate of each nuclide. The count rate represents the degree of absorption of gamma rays when passing through a specific material, and is calculated using the cps value measured in each channel of the gamma-ray spectrum and the radioactivity (Bq) value for each nuclide. Therefore, if the count rate and cps value for each nuclide are known, the count rate contribution for each nuclide in that channel can be calculated. The count rate contribution for each nuclide thus calculated is used to generate a gamma-ray spectrum for training an artificial neural network.
[0069] The normalization unit (300) subtracts the count rate contribution calculated by the count rate contribution calculation unit from the count rate for each channel, and then normalizes it to produce a normalized spectrum of the nuclide. This is used to generate learning data and is necessary for training an artificial neural network model that classifies nuclides. Through this process, the WBC gamma spectrum separation system is used to analyze and classify nuclides, and to detect and analyze radioactive contamination of nuclides.
[0070] The normalization unit (300) according to this embodiment subtracts the counting rate contribution from the counting rate measured for each channel for each nuclide, and then performs normalization for each channel to obtain a normalized spectrum of the nuclide.
[0071] First, the counting rate contribution for each channel is subtracted from the counting rate for each channel measured in the counting rate contribution calculation unit (200). The value calculated in this way is a value that has been corrected for the counting rate contribution, and in order to obtain accurate spectrum data for each nuclide, correction reflecting the counting rate contribution is necessary.
[0072] Next, the corrected count rate values are normalized for each channel. Normalization divides the count rate value of each channel by the maximum value, reducing the differences in count rate values between nuclides. These normalized values are used to generate training data for an artificial neural network, which can be used to identify each nuclide in the WBC gamma spectrum separation system.
[0073] The normalization unit (300) plays a crucial role in generating artificial neural network training data. To obtain accurate spectral data, correction and normalization of the count rate contribution are necessary. This process generates artificial neural network training data, which can then be used to accurately identify nuclides in the WBC gamma spectrum separation system.
[0074] Additionally, the normalization section is explained as follows:
[0075] The channel-specific count rate measured in the body counting device (WBC) represents the gamma-ray absorption in each channel, and this value is composed of the sum of the influences of the radioactivity of various nuclides on the detector. Therefore, by subtracting the channel-specific count rate contribution contributed by each nuclide from the measured channel-specific count rate, the channel-specific count rate with the influence of each nuclide removed is obtained. To do this, the count rate contribution of each nuclide must be known in advance. This count rate contribution is the value calculated in the "count rate contribution calculation section" described above.
[0076] Therefore, if the contribution of each nuclide to the count rate is known and subtracted from the measured channel-specific count rate, the channel-specific count rate with the channel-specific count rate contribution of each nuclide removed is obtained. Adding all of the channel-specific count rates thus obtained yields the normalized spectrum of the nuclides. This spectrum is a value corrected for the intensity of the gamma rays emitted by each nuclide, and is a spectrum created to enable comparison between nuclides.
[0077] That is, the normalization unit subtracts the channel-specific count rate contribution from the count rate for each channel, and the calculated value is a value that has been corrected for the count rate contribution. In order to obtain accurate spectrum data for each nuclide, a correction is performed that reflects the count rate contribution, and the corrected count rate value is normalized for each channel. Through this, accurate spectrum data for each nuclide can be obtained. Therefore, it is possible to accurately analyze the characteristics of gamma rays emitted for each nuclide, and thus, data measured by the body whole-body contamination screening equipment (WBC) can be analyzed more accurately.
[0078] Through this, the gamma spectrum separation system of the body-wide contamination inspection equipment for generating artificial neural network learning data according to this embodiment can obtain more accurate and reliable spectrum data for each radionuclide, which can more accurately derive the diagnostic results of the body-wide contamination inspection equipment and be of great help in patient health management.
[0079] FIG. 7 is a schematic flowchart of a method using a whole body contamination test (WBC) gamma spectrum separation system for generating artificial neural network (ANN) learning data according to one embodiment of the present invention.
[0080] As illustrated in FIG. 7, a method using a body-wide contamination screening equipment (WBC) gamma spectrum separation system (hereinafter referred to as the system) for generating artificial neural network (ANN) learning data according to an embodiment of the present invention includes (a) a step in which the system obtains the efficiency of each channel A, B, and C of the body-wide contamination screening equipment, (b) a step in which the system multiplies the gamma-ray emission rate of each nuclide by a mixed source for generating artificial neural network learning data, and multiplies the efficiency of each channel to obtain a contribution amount of counting rates by each channel, and (c) a step in which the system subtracts the contribution amount of counting rates by each channel from the counting rates measured by the nuclide, and then normalizes the counting rates to obtain a normalized spectrum of the nuclide.
[0081] In step (a), the efficiency calculation unit calculates the gamma ray detection efficiency for each channel A, B, and C of the body-wide contamination inspection equipment.
[0082] In step (b), the counting rate contribution calculation unit calculates the counting rate contribution for each channel by multiplying the gamma-ray emission rate of each nuclide in the mixed source by the gamma-ray detection efficiency for each channel A, B, and C calculated in the efficiency calculation unit.
[0083] In step (c), the normalization unit subtracts the counting rate contribution for each channel from the counting rate for each channel measured by the counting rate contribution calculation unit, and the calculated value is a value that has been corrected for the counting rate contribution. In order to obtain accurate spectrum data for each nuclide, the corrected counting rate value is normalized for each channel by performing correction reflecting the counting rate contribution.
Claims
1. Efficiency calculation section that calculates the efficiency of each channel A, B, and C of the body contaminant test equipment. A counting rate contribution calculation unit that calculates the counting rate contribution for each channel by multiplying the value obtained by multiplying the gamma-ray emission rate of each nuclide in the mixed source for generating artificial neural network learning data by the efficiency for each channel, and A gamma spectrum separation system for a whole body contamination test (WBC) for generating artificial neural network learning data, including a normalization unit that obtains a normalized spectrum of a nuclide by subtracting the channel-specific count rate contribution from the channel-specific count rate measured by the above nuclide and then normalizing it.
2. In paragraph 1, The above efficiency calculation unit is a gamma spectrum separation system for a body-wide contamination test device (WBC) for generating artificial neural network learning data, characterized in that it calculates the gamma ray detection efficiency for each of the A, B, and C channels of the body-wide contamination test device.
3. In paragraph 1, A gamma spectrum separation system for a whole body contamination screening device (WBC) for generating artificial neural network learning data, characterized in that the above-mentioned counting rate contribution calculation unit calculates the counting rate contribution for each channel by multiplying the value obtained by multiplying the gamma ray emission rate of each nuclide in the mixed source by the gamma ray detection efficiency for each channel A, B, and C calculated by the efficiency calculation unit.
4. In paragraph 1, The above normalization unit subtracts the counting rate contribution for each channel from the counting rate for each channel measured by the counting rate contribution calculation unit, and the calculated value is a value that performs correction for the counting rate contribution, and performs correction that reflects the counting rate contribution to obtain accurate spectrum data for each nuclide, and normalizes the corrected counting rate value for each channel. A gamma spectrum separation system for a body contaminant screening equipment (WBC) for generating artificial neural network learning data.
5. In a method using a gamma spectrum separation system of a whole body contamination test equipment (WBC) for generating artificial neural network learning data, (a) A step for calculating the efficiency of each channel A, B, and C of the body-wide contamination inspection equipment of the above system, (b) a step in which the system multiplies the gamma-ray emission rate of each nuclide in the mixed source for generating artificial neural network learning data by the efficiency of each channel to calculate the contribution of the counting rate for each channel; and (c) A method for separating the gamma spectrum of a whole body contamination test equipment (WBC) for generating artificial neural network learning data, comprising a step of normalizing the channel-specific count rate contribution by subtracting the channel-specific count rate contribution from the channel-specific count rate measured by the nuclide, thereby obtaining a normalized spectrum of the nuclide.
Citation Information
Patent Citations
Analysis program and analysis system for radioactive contamination in and outside the body of the subject
KR102249120B1
Dust monitor
JP1996285944A
The monitoring method of environmental radiation monitoring system that is capable of measuring dose rate of natural and artificial radioisotope
KR1020080015497A
KR20210048843A
