CN114755711A - Alpha and beta pulse discrimination method and device based on self-encoder - Google Patents
Alpha and beta pulse discrimination method and device based on self-encoder Download PDFInfo
- Publication number
- CN114755711A CN114755711A CN202210204813.7A CN202210204813A CN114755711A CN 114755711 A CN114755711 A CN 114755711A CN 202210204813 A CN202210204813 A CN 202210204813A CN 114755711 A CN114755711 A CN 114755711A
- Authority
- CN
- China
- Prior art keywords
- layer
- pulse
- alpha
- beta
- coding
- 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.)
- Granted
Links
- 238000012850 discrimination method Methods 0.000 title claims abstract description 15
- 239000002245 particle Substances 0.000 claims abstract description 29
- 238000000034 method Methods 0.000 claims abstract description 27
- 238000010791 quenching Methods 0.000 claims abstract description 27
- 230000000171 quenching effect Effects 0.000 claims abstract description 26
- 238000012216 screening Methods 0.000 claims abstract description 17
- 210000002569 neuron Anatomy 0.000 claims description 53
- 238000005259 measurement Methods 0.000 claims description 26
- 238000012549 training Methods 0.000 claims description 24
- 239000007788 liquid Substances 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 7
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000003247 decreasing effect Effects 0.000 claims description 2
- 238000013502 data validation Methods 0.000 claims 1
- 230000008859 change Effects 0.000 abstract description 5
- 230000006870 function Effects 0.000 description 22
- 238000009826 distribution Methods 0.000 description 10
- 230000000694 effects Effects 0.000 description 10
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 6
- GRYLNZFGIOXLOG-UHFFFAOYSA-N Nitric acid Chemical compound O[N+]([O-])=O GRYLNZFGIOXLOG-UHFFFAOYSA-N 0.000 description 5
- 238000001514 detection method Methods 0.000 description 5
- 229910017604 nitric acid Inorganic materials 0.000 description 5
- 239000000243 solution Substances 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 241000282414 Homo sapiens Species 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000013524 data verification Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000005262 alpha decay Effects 0.000 description 2
- 230000005255 beta decay Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 239000012530 fluid Substances 0.000 description 2
- 238000005567 liquid scintillation counting Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000008520 organization Effects 0.000 description 2
- 239000012087 reference standard solution Substances 0.000 description 2
- 101100184147 Caenorhabditis elegans mix-1 gene Proteins 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005025 nuclear technology Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000012086 standard solution Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01T—MEASUREMENT OF NUCLEAR OR X-RADIATION
- G01T1/00—Measuring X-radiation, gamma radiation, corpuscular radiation, or cosmic radiation
- G01T1/36—Measuring spectral distribution of X-rays or of nuclear radiation spectrometry
- G01T1/38—Particle discrimination and measurement of relative mass, e.g. by measurement of loss of energy with distance (dE/dx)
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01T—MEASUREMENT OF NUCLEAR OR X-RADIATION
- G01T1/00—Measuring X-radiation, gamma radiation, corpuscular radiation, or cosmic radiation
- G01T1/16—Measuring radiation intensity
- G01T1/167—Measuring radioactive content of objects, e.g. contamination
-
- 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
- G06F17/15—Correlation function computation including computation of convolution operations
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Mathematical Physics (AREA)
- High Energy & Nuclear Physics (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Mathematical Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Algebra (AREA)
- Measurement Of Radiation (AREA)
Abstract
The invention discloses an alpha and beta pulse discrimination method and a device based on a self-encoder, wherein the method comprises the following steps: acquiring particle pulse data; wherein the particle pulse data comprises alpha and beta pulses; inputting the particle pulse data into a trained self-coding model to reconstruct the pulse data to obtain a reconstruction error of the pulse data; comparing the size of the reconstruction error, and screening the input particle pulse as alpha or beta pulse based on the size of the reconstruction error; and the reconstruction error of the trained self-coding model to the alpha pulse data is smaller than that of the beta pulse data. The method realizes the discrimination of the alpha and beta pulse data, can obtain lower alpha and beta discrimination error rate under the condition of quenching level change, and has better application prospect.
Description
Technical Field
The invention relates to the technical field of nuclear radiation detection, in particular to an alpha and beta pulse discrimination method and device based on an auto-encoder.
Background
With the development of human nuclear activities such as the wide application of nuclear technology, nuclear weapon tests, nuclear energy peaceful utilization, nuclear accidents and the like, the total amount of radionuclides entering the water environment is increased due to the human nuclear activities in the global scope. In order to guarantee the water safety of human beings, relevant standards are established for the water body radioactivity detection by the world health organization, the international standardization organization and China. In order to reduce the cost and manpower of daily aquatic radioactivity detection, total alpha and total beta measurement becomes an important screening means for aquatic radioactivity detection. The liquid scintillation counting method has become an important method for measuring the total alpha and beta of the water body due to the advantages of high detection efficiency of alpha and beta, simple sample preparation, capability of simultaneously measuring alpha and beta and the like. In order to achieve a single measurement while determining the radioactivity levels of both alpha and beta populations in a sample, it is necessary to discriminate which of the alpha or beta decays is responsible for the pulse signals acquired by the measurement, and to perform a count by classification. Therefore, alpha and beta pulse discrimination is an important content of liquid scintillation counting to achieve simultaneous measurement of total alpha and total beta radioactivity levels.
The alpha and beta pulse discrimination methods mainly comprise two methods of pulse amplitude discrimination and pulse shape discrimination, the pulse shape discrimination is more in application, and each liquid scintillation spectrometer manufacturer has a unique discrimination method. All current pulse shape discrimination methods need to determine the optimal discrimination values for alpha and beta pulse shape discrimination before measurement. Once the level of quenching of the sample changes, the optimal discrimination for alpha and beta pulse shape discrimination must be re-established. Otherwise, as the quenching level of the sample to be detected is increased or reduced, the error rate of screening alpha and beta is increased, even reaching several tens of percent, so that the measurement result of the total alpha and the total beta has larger deviation.
Disclosure of Invention
The present invention is directed to solving, at least in part, one of the technical problems in the related art.
Therefore, the invention provides an alpha and beta pulse discrimination method based on a self-encoder, which adopts a machine learning self-encoder to extract the characteristics of alpha and beta pulse signals, realizes the discrimination of the alpha and beta pulse signals and can obtain lower alpha and beta discrimination error rate under the change of quenching level.
The invention also aims to provide an alpha and beta pulse discrimination device based on an automatic encoder.
In order to achieve the above object, in one aspect, the present invention provides an α and β pulse discrimination method based on an auto-encoder, including the following steps:
acquiring particle pulse data; wherein the particle pulse data comprises alpha and beta pulses; inputting the particle pulse data into a trained self-coding model to reconstruct the pulse data to obtain a reconstruction error of the pulse data; wherein the self-coding model comprises: the system comprises a network input layer, a coding layer, an intermediate layer, a decoding layer and a network output layer; the coding layer and the decoding layer are mutually mirror-matched, a first layer of the coding layer is preset with a first plurality of neurons, a second layer of the coding layer is preset with a second plurality of neurons, a third layer of the coding layer is preset with a third plurality of neurons, a fourth layer of the coding layer is preset with a fourth plurality of neurons, a middle layer of the coding layer is preset with a fifth plurality of neurons, and a network input layer and a network output layer are preset with a first plurality of neurons; comparing the size of the reconstruction error, and screening the input particle pulse data as alpha or beta pulse based on the size of the reconstruction error; and the reconstruction error of the trained self-coding model to the alpha pulse data is smaller than that of the beta pulse data.
The alpha and beta pulse discrimination method based on the self-encoder can realize the discrimination of alpha and beta pulse signals and can obtain lower alpha and beta discrimination error rate under the condition of change of quenching level.
In addition, the method for discriminating α and β pulses based on the self-encoder according to the above embodiment of the present invention may further have the following additional technical features:
further, in an embodiment of the present invention, the method further includes: acquiring a training data set; inputting the training data set into a self-coding model for iterative training to obtain the trained self-coding model.
Further, in an embodiment of the present invention, the acquiring the training data set includes: presetting measurement time and measuring alpha standard-added samples with different quenching levels to obtain corresponding pulse data; randomly sampling a preset number of pulse data from the pulse data of the alpha standard sample to form the training data set.
Further, in an embodiment of the present invention, the activation function of the first layer of the coding layer adopts a sigmoid function, the activation function of each of the second layer of the coding layer to the intermediate layer adopts a ReLU function, and a formula of the sigmoid function and a formula of the ReLU function are respectively:
further, in an embodiment of the present invention, when the training data set is input into the self-coding model for iterative training, a mean square error between the network input layer and the network output layer is used as a loss function, and an Adam optimization algorithm is used to update the network weight.
Further, in an embodiment of the present invention, the formula of α and β pulse discriminant function PSD when comparing the reconstruction error of the pulse data is as follows:
where y represents the pulse input value,
representing the model reconstructed output values.
Further, in an embodiment of the present invention, when the training data set is input into the self-coding model for iterative training, Batch-Normalization and L2 regularization are added.
Further, in an embodiment of the present invention, the number of neurons of the first plurality of neurons to the fifth plurality of neurons is a decreasing series.
Further, in an embodiment of the present invention, the method further includes performing data verification on the α and β pulse discrimination result, including: the energy released by the decay of alpha and beta in the alpha and beta standard-adding samples with different quenching levels is acted with scintillation liquid to generate photons, the photons are detected by a photomultiplier tube, and corresponding electronic pulse data are output; taking the pulse signal data as input data of the trained self-coding model, and calculating a discrimination characteristic value PSD of the electronic pulse data with output data of the model; and reading the electronic pulse data through the trained self-coding model based on the PSD to judge the pulse, and respectively counting the judged alpha pulse and beta pulse.
In order to achieve the above object, another aspect of the present invention provides an α and β pulse discriminating device based on an auto-encoder, including:
the pulse acquisition module is used for acquiring particle pulse data; wherein the particle pulse data comprises alpha and beta pulses; the error reconstruction module is used for inputting the particle pulse data into a trained self-coding model to reconstruct the pulse data to obtain a reconstruction error of the pulse data; wherein the self-coding model comprises: the system comprises a network input layer, a coding layer, a middle layer, a decoding layer and a network output layer; the coding layer and the decoding layer are mutually mirror-matched, a first layer of the coding layer is preset with a first plurality of neurons, a second layer of the coding layer is preset with a second plurality of neurons, a third layer of the coding layer is preset with a third plurality of neurons, a fourth layer of the coding layer is preset with a fourth plurality of neurons, a middle layer of the coding layer is preset with a fifth plurality of neurons, and a network input layer and a network output layer are preset with a first plurality of neurons; the pulse discrimination module is used for comparing the reconstruction error magnitude and discriminating the input particle pulse data into alpha or beta pulses based on the reconstruction error magnitude; and the reconstruction error of the trained self-coding model to the alpha pulse data is smaller than that of the beta pulse data.
The alpha and beta pulse discrimination device based on the self-encoder provided by the embodiment of the invention realizes the discrimination of alpha and beta pulse signals and can obtain a lower alpha and beta discrimination error rate under the condition of change of a quenching level.
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
The foregoing and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a diagram illustrating a self-coding model according to an embodiment of the present invention;
FIG. 2 is a flowchart of an alpha and beta pulse discrimination method based on an auto-encoder according to an embodiment of the present invention;
FIG. 3 is a block diagram of a data verification system according to an embodiment of the present invention;
FIG. 4 is a PSD distribution histogram of the Alpha1 sample according to an embodiment of the present invention;
FIG. 5 is a histogram of the Alpha1 sample and background PSD distribution according to an embodiment of the present invention;
FIG. 6 is a PSD distribution histogram of a Beta1 sample according to an embodiment of the present invention;
FIG. 7 is a graph of a Beta1 sample and background PSD distribution histogram according to an embodiment of the present invention;
FIG. 8 is a graph comparing pulse distribution to background pulse signal subtraction before and after in Alpha1 and Beta1 samples according to an embodiment of the present invention;
FIG. 9 is a graph of pulse discrimination error rates for samples Alpha1 and Beta1 according to an embodiment of the present invention;
fig. 10 is a schematic structural diagram of an α and β pulse discriminating device based on an auto-encoder according to an embodiment of the present invention.
Detailed Description
It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present invention will be described in detail below with reference to the embodiments with reference to the attached drawings.
In order to make those skilled in the art better understand the technical solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
First, an α and β pulse discrimination method based on an auto-encoder according to an embodiment of the present invention will be described with reference to the drawings.
The invention adopts a deep learning model self-encoder for alpha and beta discrimination. The self-encoder has a good feature extraction function, can well compress data to an intermediate layer dimension by adding original data again, is usually used for data dimension reduction, feature extraction, data encryption and other works, and is illustrated in a self-encoding model shown in figure 1. The self-coding pulse identification method based on the self-coding model has the basic principle that the self-coding network is built, the self-coding model is trained by using the alpha pulses with different quenching levels, so that the self-coding model has a better reconstruction function on the alpha pulses, the reconstruction error of the self-coding on the alpha pulse data is very small through continuous iterative training, and the pulse identification is realized based on the error.
Fig. 2 is a flowchart of an alpha and beta pulse discrimination method based on an auto-encoder according to an embodiment of the present invention.
As shown in fig. 2, the method for discriminating alpha and beta pulses based on the self-encoder comprises the following steps:
step S1, acquiring particle pulse data; wherein the particle pulse data comprises alpha and beta pulses.
Step S2, inputting the particle pulse data into the trained self-coding model to reconstruct the pulse data, and obtaining the reconstruction error of the pulse data; wherein the self-coding model comprises: the system comprises a network input layer, a coding layer, a middle layer, a decoding layer and a network output layer; the coding layer and the decoding layer are mutually in mirror phase, a first layer of the coding layer is preset with a first plurality of neurons, a second layer of the coding layer is preset with a second plurality of neurons, a third layer of the coding layer is preset with a third plurality of neurons, a fourth layer of the coding layer is preset with a fourth plurality of neurons, a middle layer of the coding layer is preset with a fifth plurality of neurons, and a network input layer and a network output layer are preset with a first plurality of neurons.
Specifically, the number of neurons in the first plurality of neurons to the fifth plurality of neurons is a descending sequence.
As an example, the self-coding model network for discriminating alpha and beta pulses has 1024 neurons in the input layer, 1024 neurons are designed in the first layer of the coding layer, the number of the neurons is consistent with that of alpha/beta pulse acquisition points, and a sigmoid function is adopted as an activation function. The Sigmoid function is defined as:
512 neurons are designed at the layer 2, 256 neurons are designed at the layer 3, 128 neurons are designed at the layer 4, and 64 neurons are arranged in the middle layer. The ReLU function is adopted from the layer 2 to the middle layer activation function. The ReLU function is defined as:
the decoding layer and the coding layer are in a mirror image relationship, the final output layer is 1024 neurons, and the activation function is a sigmoid function. In order to enhance the robustness of the self-encoder model and prevent the overfitting of the network, Batch-Normalization and L2 regularization are added in the model training process. In the model training process, the mean square error of an input layer and an output layer is used as a loss function, and the Adam optimization algorithm is used for updating the network weight. To better characterize the characteristic differences of the α and β pulses, the α and β pulse discriminant function PSD is defined as:
in the formula, y represents a pulse input value,
representing the model reconstruction output values.
Step S3, comparing the reconstruction error, and screening the input particle pulse data as alpha or beta pulse based on the reconstruction error; and the reconstruction error of the trained self-coding model to the alpha pulse data is smaller than that of the beta pulse data.
Specifically, due to the difference between the β pulse shape and the α pulse, the self-coding model has a larger reconstruction error for the β pulse than for the α pulse. The reconstruction error of the self-coding model to the pulse can be used for realizing the discrimination of alpha and beta pulses. Therefore, the invention can realize the discrimination of the input pulse alpha and the input pulse beta through the discrimination of the error.
Further, in order to verify the feasibility and the effectiveness of the alpha and beta pulse discrimination method based on the self-encoder, a set of verification system is built and shown in fig. 3, and data verification is performed.
The verification system consists of an optical measuring room and a computer, wherein a high-speed acquisition card is installed in the computer, and a pulse discrimination model and corresponding software are deployed. The system works as follows: the energy released by alpha/beta decay reacts with scintillation liquid in the scintillation vial to generate photons, and the photons are converted into electrons through the photomultiplier tube and amplified to finally obtain an electric pulse signal representing the decay of the particles. The high-speed acquisition card is used for acquiring the electric pulse signals, the pulse data acquired by the two paths of photomultiplier tubes are subjected to time coincidence, and the acquired data after background interference is removed is stored in a computer. Reading pulse signals based on an alpha and beta pulse discrimination model of a self-encoder to discriminate, counting in an alpha window if the discrimination model is alpha pulse, and counting in a beta window if the discrimination model is not alpha pulse.
In the embodiment, the pulse coincidence time of the photomultiplier is set to be 10ns, the sampling frequency of the high-speed acquisition board is 1250MHz, and 1024 points are acquired by each pulse.
Acquisition of alpha and beta pulse data for different quench levels:
alpha authentication reference standard solution adoption241Am, Beta authentication reference standard solution adoption90Sr and 0.05mol/L nitric acid solution are taken as a quenching agent. A standard solution with known activity is added into a certain volume of liquid scintillation fluid, and then a nitric acid solution is added, so that alpha and beta standard adding samples with different quenching levels are prepared, and the details are shown in Table 1.
TABLE 1 alpha, beta Standard sample parameters at different quenching levels
The pulse data obtained by measuring the marked sample by the verification system is stored in a computer data file for training and verifying the model. The time for measurement and the total pulse count for the spiked samples are detailed in Table 2.
TABLE 2 measurement time and Total pulse count for the spiked samples
As an example, 10000 pieces of pulse data from the first measurement of each sample Alpha0-Alpha4 were randomly sampled to form a training data set. After the model is epoch for 50 times by using the training data set, the loss function loss of the model gradually becomes stable.
The pulse discrimination error rate verification process is described below with reference to the accompanying drawings using Alpha1 and Beta1 as second measurement data.
Counting pulse characteristic PSDs of an Alpha1 sample and a background sample with the same quenching level, drawing a pulse PSD distribution histogram, and carrying out background subtraction on the Alpha1 sample;
secondly, counting pulse characteristic PSDs of a Beta1 sample and a background sample with the same quenching level, drawing a pulse PSD distribution histogram, and carrying out background subtraction on an Alpha1 sample;
and thirdly, determining optimal values of pulse discrimination function PSDs of alpha and beta, wherein the optimal values of the pulse discrimination PSDs enable the alpha pulse discrimination error rate and the beta pulse discrimination error rate to be equal as much as possible, and the sum of the discrimination error rates is minimum.
Among these, FIG. 4, FIG. 5, FIG. 6 and FIG. 7 are the PSD distribution histograms of Alpha1 and Beta1 sample measurements and the PSD distribution histogram of background sample measurements at the same quenching level. FIG. 8 is a comparison of pulse distribution versus background pulse signal before and after subtraction for the Alpha1 sample and the Beta1 sample.
Further, alpha and beta pulse discrimination error rates under different pulse discrimination threshold values PSD are calculated, a discrimination error rate curve is drawn, and a PSD value under the condition of the minimum discrimination error rate is determined. The pulse discrimination error rate curves for samples Alpha1 and Beta1 are shown in fig. 9.
According to the above procedure, the pulse discrimination PSD and the discrimination error rate of α and β spiked samples at different quenching levels are shown in table 3.
TABLE 3 different quenching levels alpha and beta standard sample pulse discrimination PSD and discrimination error rate
As can be seen from Table 3, under the condition that the ratio of the volume of the scintillation liquid to 0.05mol/L nitric acid is higher than 1:1, the PSD values for alpha and beta pulse discrimination are basically consistent, which shows that the method can be suitable for alpha and beta pulse discrimination with a wide quenching level variation range. The screening error rates of alpha and beta pulses have a positive correlation with the quenching level, namely the screening error rates of the alpha and beta pulses are increased along with the increase of the quenching level. The ratio of scintillation liquid to nitric acid in the sample is higher than 12:8, and when the quenching indication parameter tSIE is more than or equal to 300, the discrimination error rate of alpha and beta pulses is lower than 3.5%.
Further, pulse screening is used for different alpha and beta activity ratio adaptability.
As an example, three groups of mixed standard-added samples with different alpha and beta activity ratios are configured under the quenching level of 12ml of scintillation fluid and 4ml of 0.05M nitric acid, and the adaptability of the pulse discrimination PSD under different alpha and beta activity ratios is verified. The activity ratios of alpha and beta for the mixed standards are shown in table 4.
TABLE 4 Activity ratio of Mixed spiked samples alpha and beta
The net count rate for each sample was calculated by taking three measurements per mixed sample according to the above procedure, and the net count rates for the samples were calculated as shown in the following formula, and the results of the measurements of the mixed samples Mix1-Mix3 are summarized in tables 5-7.
Wherein N issRepresenting the sample pulse count, NbRepresenting the background pulse count, tsDenotes the effective measurement time of the sample, tbThe effective measurement time of the base sample is shown.
TABLE 5 statistics of mixed sample Mix1 measurements
TABLE 6 statistics of Mix2 measurements
TABLE 7 statistics of mixed sample Mix3 measurements
The examination result shows that the alpha and beta pulse screening based on self-coding has better stability for alpha and beta standard-adding mixed samples with different proportions. As can be seen from tables 5-7, the relative error between the activity measurement value of the sample and the reference value is mainly caused by statistical error, and the use requirement of actual measurement is met.
Therefore, the PSD based on the reconstruction error of the self-coding model is used for screening alpha and beta pulses, and the measurement result of the mixed labeled sample with different quenching levels and different alpha and beta activity ratios shows that the method has better robustness, can meet the measurement precision requirement of an actual water sample, and has better application prospect. The method can be applied to alpha and beta pulse discrimination of water body liquid scintillation measurement, and can also be applied to alpha and beta pulse discrimination of other total alpha and beta measurement methods.
In order to implement the foregoing embodiment, as shown in fig. 10, in this embodiment, an α and β pulse screening apparatus 10 based on an auto-encoder is further provided, where the apparatus 10 includes: a pulse acquisition module 100, an error reconstruction module 200, and a pulse screening module 300.
A pulse acquisition module 100, configured to acquire particle pulse data; wherein the particle pulse data comprises alpha and beta pulses;
the error reconstruction module 200 is configured to input the particle pulse data into a trained self-coding model to reconstruct the pulse data, so as to obtain a reconstruction error of the pulse data; wherein the self-encoding model comprises: the system comprises a network input layer, a coding layer, a middle layer, a decoding layer and a network output layer; the coding layer and the decoding layer are mutually mirror-matched, a first layer of the coding layer is preset with a first plurality of neurons, a second layer of the coding layer is preset with a second plurality of neurons, a third layer of the coding layer is preset with a third plurality of neurons, a fourth layer of the coding layer is preset with a fourth plurality of neurons, a middle layer of the coding layer is preset with a fifth plurality of neurons, and a network input layer and a network output layer are preset with a first plurality of neurons;
the pulse discrimination module 300 is used for comparing the reconstruction error magnitude and discriminating the input particle pulse data into alpha or beta pulses based on the reconstruction error magnitude; and the reconstruction error of the trained self-coding model to the alpha pulse data is smaller than that of the beta pulse data.
According to the alpha and beta pulse discrimination device based on the self-encoder, the discrimination of alpha and beta pulse signals is realized, and a low alpha and beta discrimination error rate can be obtained under the condition of change of a quenching level.
It should be noted that the foregoing explanation on the embodiment of the method for screening for α and β pulses based on an auto-encoder is also applicable to the apparatus for screening for α and β pulses based on an auto-encoder in this embodiment, and is not described herein again.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Moreover, various embodiments or examples and features of various embodiments or examples described in this specification can be combined and combined by one skilled in the art without being mutually inconsistent.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.
Claims (10)
1. An alpha and beta pulse discrimination method based on an auto-encoder is characterized by comprising the following steps:
acquiring particle pulse data; wherein the particle pulse data comprises alpha, beta pulses;
inputting the particle pulse data into a trained self-coding model to reconstruct the pulse data, and obtaining a reconstruction error of the pulse data; wherein the self-coding model comprises: the system comprises a network input layer, a coding layer, a middle layer, a decoding layer and a network output layer; the coding layer and the decoding layer are mirror-matched with each other, a first layer of the coding layer is preset with a first plurality of neurons, a second layer of the coding layer is preset with a second plurality of neurons, a third layer of the coding layer is preset with a third plurality of neurons, a fourth layer of the coding layer is preset with a fourth plurality of neurons, a middle layer of the coding layer is preset with a fifth plurality of neurons, and the network input layer and the network output layer are preset with the first plurality of neurons;
comparing the size of a reconstruction error, and screening the input particle pulse data to be alpha or beta pulses based on the size of the reconstruction error; and the reconstruction error of the trained self-coding model to alpha pulse data is smaller than that of beta pulse data.
2. The method of claim 1, further comprising:
acquiring a training data set;
inputting the training data set into a self-coding model for iterative training to obtain the trained self-coding model.
3. The method of claim 2, wherein the obtaining a training data set comprises:
presetting measurement time and measuring alpha standard-added samples with different quenching levels to obtain corresponding pulse data;
randomly sampling a preset number of pulse data from the pulse data of the alpha labeled sample to form the training data set.
4. The method according to claim 1, wherein the activation function of the first one of the coding layers employs a sigmoid function, the activation function of each of the second one of the coding layers to the intermediate layer employs a ReLU function, and the formula of the sigmoid function and the formula of the ReLU function are respectively:
5. the method according to claim 4, wherein when the training data set is input into the self-coding model for iterative training, the mean square error of the network input layer and the network output layer is adopted as a loss function, and the network weight is updated by using an Adam optimization algorithm.
6. The method of claim 5, wherein the reconstruction error of the pulse data is compared, and the equation for the alpha and beta pulse discrimination function PSD is:
where y represents the pulse input value,
representing the model reconstructed output values.
7. The method of claim 6, wherein the training data set is input into the self-coding model for iterative training, and Batch-Normalization and L2 regularization are added.
8. The method of claim 1, wherein the number of neurons in the first plurality of neurons through the fifth plurality of neurons is a decreasing series.
9. The method according to claim 1, further comprising performing data validation on the alpha and beta pulse screening results, including:
the energy released by the decay of alpha and beta in the alpha and beta standard-adding samples with different quenching levels is acted with scintillation liquid to generate photons, the photons are detected by a photomultiplier tube, and corresponding electronic pulse data are output;
taking the pulse signal data as input data of the trained self-coding model, and calculating a discrimination characteristic value PSD of the electronic pulse data with output data of the model;
and reading the electronic pulse data through the trained self-coding model based on the PSD to judge the pulse, and respectively counting the judged alpha pulse and beta pulse.
10. An alpha and beta pulse screening device based on a self-encoder is characterized by comprising:
the pulse acquisition module is used for acquiring particle pulse data; wherein the particle pulse data comprises alpha, beta pulses;
the error reconstruction module is used for inputting the particle pulse data into a trained self-coding model to reconstruct the pulse data to obtain a reconstruction error of the pulse data; wherein the self-coding model comprises: the system comprises a network input layer, a coding layer, a middle layer, a decoding layer and a network output layer; the coding layer and the decoding layer are mirror-matched with each other, a first layer of the coding layer is preset with a first plurality of neurons, a second layer of the coding layer is preset with a second plurality of neurons, a third layer of the coding layer is preset with a third plurality of neurons, a fourth layer of the coding layer is preset with a fourth plurality of neurons, a middle layer of the coding layer is preset with a fifth plurality of neurons, and the network input layer and the network output layer are preset with the first plurality of neurons;
the pulse discrimination module is used for comparing the size of a reconstruction error and discriminating whether the input particle pulse data is alpha or beta pulse based on the size of the reconstruction error; and the reconstruction error of the trained self-coding model to the alpha pulse data is smaller than that of the beta pulse data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210204813.7A CN114755711B (en) | 2022-03-03 | 2022-03-03 | Alpha and beta pulse screening method and device based on self-encoder |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210204813.7A CN114755711B (en) | 2022-03-03 | 2022-03-03 | Alpha and beta pulse screening method and device based on self-encoder |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114755711A true CN114755711A (en) | 2022-07-15 |
| CN114755711B CN114755711B (en) | 2024-06-25 |
Family
ID=82325425
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210204813.7A Active CN114755711B (en) | 2022-03-03 | 2022-03-03 | Alpha and beta pulse screening method and device based on self-encoder |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114755711B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030163436A1 (en) * | 2002-01-11 | 2003-08-28 | Jost Seifert | Neuronal network for modeling a physical system, and a method for forming such a neuronal network |
| CN105319574A (en) * | 2015-11-16 | 2016-02-10 | 中国原子能科学研究院 | An alpha and beta pulse discriminating method |
| CN109871940A (en) * | 2019-01-31 | 2019-06-11 | 清华大学 | A kind of multilayer training algorithm of impulsive neural networks |
| CN110837098A (en) * | 2018-08-16 | 2020-02-25 | 同方威视技术股份有限公司 | Method, FPGA, device and storage medium for discriminating pulse signal |
| CN110929843A (en) * | 2019-10-29 | 2020-03-27 | 国网福建省电力有限公司 | Abnormal electricity consumption behavior identification method based on improved deep self-coding network |
| CN113721286A (en) * | 2021-08-16 | 2021-11-30 | 中国科学技术大学 | Positron annihilation life spectrometer for pulse discrimination by using support vector machine |
-
2022
- 2022-03-03 CN CN202210204813.7A patent/CN114755711B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030163436A1 (en) * | 2002-01-11 | 2003-08-28 | Jost Seifert | Neuronal network for modeling a physical system, and a method for forming such a neuronal network |
| CN105319574A (en) * | 2015-11-16 | 2016-02-10 | 中国原子能科学研究院 | An alpha and beta pulse discriminating method |
| CN110837098A (en) * | 2018-08-16 | 2020-02-25 | 同方威视技术股份有限公司 | Method, FPGA, device and storage medium for discriminating pulse signal |
| CN109871940A (en) * | 2019-01-31 | 2019-06-11 | 清华大学 | A kind of multilayer training algorithm of impulsive neural networks |
| CN110929843A (en) * | 2019-10-29 | 2020-03-27 | 国网福建省电力有限公司 | Abnormal electricity consumption behavior identification method based on improved deep self-coding network |
| CN113721286A (en) * | 2021-08-16 | 2021-11-30 | 中国科学技术大学 | Positron annihilation life spectrometer for pulse discrimination by using support vector machine |
Non-Patent Citations (2)
| Title |
|---|
| 杜进;陈云华;张灵;麦应潮;: "基于改进深度残差网络的低功耗表情识别", 计算机科学, no. 09, 15 September 2018 (2018-09-15) * |
| 王一鸣;龚军军;夏文明;陈君军;刘云鹏;: "脉冲堆积条件下n与γ脉冲波形甄别仿真分析", 核电子学与探测技术, no. 01, 20 January 2017 (2017-01-20) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114755711B (en) | 2024-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109063741B (en) | Energy spectrum analysis method based on Hilbert curve transformation and deep learning | |
| CN107229787B (en) | Gamma energy spectrum analysis method based on approximation coefficient and deep learning | |
| Mohammadian-Behbahani et al. | A comparison study of the pile-up correction algorithms | |
| CN105190357A (en) | Apparatus and method for the evaluation of gamma radiation events | |
| CN102073059A (en) | Digital pileup waveform processing method and system | |
| CN105824817A (en) | Flash pulse digitization method | |
| Carpenter et al. | Scale of spatial pattern: four methods compared | |
| Bloesch et al. | Lead-210 dating of sediments compared with accumulation rates estimated by natural markers and measured with sediment traps | |
| CN103595417A (en) | Method for mixed reconstruction of pulse-continuous nuclear signals under compressed sensing frame | |
| CN107390259A (en) | A kind of nuclide identification method based on SVD and SVM | |
| CN107728192A (en) | The nuclide identifier system and method for multi-detector | |
| CN106706127A (en) | SiPM-based multi-photon detection method | |
| CN114755711B (en) | Alpha and beta pulse screening method and device based on self-encoder | |
| Yamamuro et al. | Neutron Capture Cross Section Measurements of Nb-93, 1-127, Ho-165, Ta-181 and U-238 between 3.2 and 80keV | |
| CN108646285B (en) | Gamma ray energy spectrum measuring method, device and system | |
| Gu et al. | A new pulse pileup rejection method based on position shift identification | |
| Telesca et al. | Investigating clustering structures in time-occurrence sequences of seismic events observed in the Irpinia-Basilicata region (southern Italy) | |
| CN111239797B (en) | Collector based on radiation particle event and rapid nuclide identification method | |
| CN111273337B (en) | Nuclear energy spectrum processing method based on Monte Carlo pulse interpolation method | |
| Kobayashi et al. | Measurement and covariance analysis of californium-252 spectrum averaged cross sections | |
| Gao et al. | Study on sequential Bayesian radionuclide identification approach: Threshold and detection capability | |
| Namenson et al. | Spins of low-energy neutron resonances in 175Lu, 189Os and 187Os | |
| CN110687584A (en) | Fast nuclide identification method based on LSTM | |
| Wang et al. | Dataset for neutron and gamma-ray pulse shape discrimination | |
| CN102637243B (en) | A kind of source drive-type unknown nuclear component Multi-parameter Data Acquisition method based on higher order statistic analysis |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |