Stacked pulse waveform discrimination method based on charge integration method and KNN algorithm

Abstract

The invention relates to a stacked waveform discriminating method, in particular to a stacked pulse waveform discriminating method based on a charge integration method and a KNN algorithm, and solves the technical problem that an existing stacked pulse discriminating method easily causes repeated selection of single-particle waveforms with small difference or even no obvious difference, so that invalid calculation is increased. The stacked pulse waveform discrimination method based on the charge integration method and the KNN algorithm comprises the following steps: S1, obtaining a neutron gamma single-wave signal sample library U1 by using the charge integration method; s2, obtaining a data set M1 to be discriminated; s3, obtaining a neutron gamma single-wave signal sample set U2 by using a charge integration method; s4, constructing a screening calculation data set M2; s5, calculating the similarity and sorting, and selecting K stacked samples with the highest similarity from the sorting by using a KNN algorithm; and taking the category with the largest number in the K stacked samples as the category of the data set M1 to be discriminated, and completing stacked pulse waveform discrimination.

Description

Technical Field

[0001] The invention relates to a pile-up waveform identification method, and in particular to a pile-up pulse waveform identification method based on a charge integration method and a KNN algorithm. Background Art

[0002] From radiation measurement to particle physics experiments, effective pulse shape discrimination (PSD) technology is essential in many applications. Traditional PSD techniques, such as fast Fourier transform, wavelet transform, charge integration method, rise time method, etc., can achieve single-particle waveform discrimination of neutron gamma based on the difference in the trailing edge decay constant of neutron gamma single wave signal.

[0003] When the detector is working under high counting rate conditions, when a radiation particle enters the detector to generate a pulse waveform that has not returned to the baseline, another radiation particle enters the detector to generate a pulse signal, resulting in a pulse signal pileup event. The pileup pulse structure is quite different from the single particle waveform, making the traditional single particle waveform PSD technology invalid. When the proportion of pileup events is low, the identification and abandonment method can be used. However, when the proportion of pileup waveforms is high, in order to reduce the loss of pileup events, it is particularly important to identify the characteristic information of neutron gamma from the pileup events. Therefore, it is necessary to develop a pileup pulse identification method.

[0004] In the existing identification methods, neutron gamma single wave signals are often randomly selected to artificially construct pile-up pulses. However, the database of neutron gamma single wave signals often contains a large amount of waveform data. Randomly selecting neutron gamma single wave signals can easily lead to neutron gamma single wave signals with small or no obvious differences being used to process pile-up events, increasing the invalid calculation of the identification method. Therefore, it is necessary to develop new methods to prevent single particle waveforms with small or no obvious differences from being repeatedly selected, while improving the calculation efficiency of the identification method. Summary of the invention

[0005] The purpose of the present invention is to solve the technical problem that the existing pile-up pulse identification method easily leads to repeated selection of single-particle waveforms with small or even no obvious differences, resulting in increased invalid calculations, and to provide a pile-up pulse waveform identification method based on charge integration method and KNN algorithm.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A method for identifying piled-up pulse waveforms based on a charge integration method and a KNN algorithm is special in that it includes the following steps:

[0008] S1, collecting neutron gamma single waveform mixed signals, using the charge integration method to identify the neutron single wave signal and the gamma single wave signal in the neutron gamma single waveform mixed signal, and constructing a neutron gamma single wave signal sample library U1;

[0009] S2, processing the single wave signal in the neutron gamma single wave signal sample library U1 into a data set M1 to be identified; the stacked waveform categories of the data set M1 to be identified are n+n, n+γ, γ+n, γ+γ, where n is a neutron single wave signal and γ is a gamma single wave signal;

[0010] S3, using the charge integration method to screen the neutron single wave signal and the gamma single wave signal in the neutron gamma single wave signal sample library U1, to obtain the neutron gamma single wave signal sample set U2;

[0011] S4, constructing a discrimination calculation data set M2 based on the to-be-discriminated data set M1 and the neutron gamma single wave signal sample set U2;

[0012] S5. Calculate the similarity between the piled waveform to be identified in the dataset M1 and the piled waveform in the corresponding dataset M2, and sort them according to the similarity. Use the KNN algorithm to select the K piled samples with the highest similarity in the sorting; take the category with the largest number of K piled samples as the category of the dataset M1 to be identified, and complete the identification of the piled pulse waveform.

[0013] Furthermore, step S1 is specifically as follows:

[0014] 1.1. Collect neutron gamma single waveform mixed signal;

[0015] 1.2. Define the amplitude integral between the starting time t1 and the end point t of the neutron gamma single waveform mixed signal as the long window integral value Q1 of the charge integration method; the amplitude integral between the 50% peak time t2 of the trailing edge of the neutron gamma single waveform mixed signal and the end point t as the short window integral value Q2 of the charge integration method; and use the ratio of the short window integral value Q2 to the long window integral value Q1 as the discrimination factor Y;

[0016] 1.3. According to the discrimination factor Y, the neutron-gamma single-wave mixed signal is discriminated, recorded and marked, and the marked neutron single-wave signal and gamma single-wave signal are combined into a neutron-gamma single-wave signal sample library U1.

[0017] Furthermore, step S2 is specifically as follows:

[0018] 2.1. Randomly select a neutron single wave signal as the first waveform and a gamma single wave signal as the second waveform in the neutron gamma single wave signal sample library U1;

[0019] 2.2. After aligning the peaks of the first waveform and the second waveform, shift the second waveform backward and record the shift time T;

[0020] 2.3. Add the first waveform and the second waveform after the back shift to obtain an n+γ type stacking waveform; the combination categories of the stacking waveform also include n+n, γ+n, γ+γ, where n is a neutron single wave signal and γ is a gamma single wave signal;

[0021] 2.4. Follow the methods of steps 2.1 to 2.3 until all n+γ stacking waveforms are obtained;

[0022] 2.5. Process the single-wave signals in the neutron gamma single-wave signal sample library U1 according to the methods of steps 2.1 to 2.4 for the combination categories of n+n, γ+n, and γ+γ, and use them as the data set M1 to be identified.

[0023] Furthermore, step S3 is specifically as follows:

[0024] 3.1. Using the long window integral value Q1 and the short window integral value Q2 as parameters, the neutron single wave signal and the gamma single wave signal in the neutron gamma single wave signal sample library U1 are screened;

[0025] 3.2. Divide the long window integral value Q1 into n intervals, with an interval ΔQ1 = Q1 / n, where n is greater than or equal to 40;

[0026] 3.3. Take the short window integral value Q2 corresponding to the interval ΔQ1, and divide the short window integral value Q2 into m intervals, where the interval ΔQ2 = Q2 / m, and the value of m is greater than or equal to 4;

[0027] 3.4. Select a single wave signal in each of the n*m ​​intervals to form the neutron gamma single wave signal sample set U2.

[0028] Furthermore, step S4 is specifically as follows:

[0029] 4.1. Select any stacked waveform from the dataset M1 to be identified as a reference waveform;

[0030] 4.2. Randomly select a neutron single wave signal as the first waveform and a gamma single wave signal as the second waveform from the neutron gamma single wave signal sample set U2;

[0031] 4.3. Multiply the first waveform by the peak factor K1, multiply the second waveform by the peak factor K2, and shift the second waveform backward by the same time interval as the reference waveform; then add the shifted first waveform to the second waveform to obtain an n+γ type stacking waveform;

[0032] 4.4. Compare the peak value of the n+γ type stacked waveform obtained in step 4.3 with the peak value of the reference waveform obtained in step 4.1 to determine whether the difference between the peak value of the stacked waveform and the peak value of the reference waveform is less than or equal to 1%. If so, execute step 4.5; if not, adjust the peak factors K1 and K2 by step size and return to step 4.3;

[0033] 4.5. Repeat the method of steps 4.2 to 4.4, so that the first waveform traverses all neutron waveforms of the neutron gamma single wave signal sample set U2 once, and the second waveform traverses all gamma waveforms of the neutron gamma single wave signal sample set U2 once, obtain the discrimination calculation subset corresponding to the reference waveform, and mark it;

[0034] 4.6. Repeat the methods of steps 4.1 to 4.5 according to the combination categories of n+n, γ+n, and γ+γ, traverse all the stacked waveforms in the data set M1 to be identified, process the identification calculation set for all the stacked waveforms in the data set M1 to be identified, and use it as the identification calculation set M2.

[0035] Furthermore, step S5 is specifically as follows:

[0036] 5.1. Use the Euclidean distance to calculate the similarity between any stacked waveform in the to-be-identified data set M1 and its corresponding stacked waveform in the identification calculation data set M2, and sort them from large to small;

[0037] 5.2. Use the KNN algorithm to select K stacking samples from the sorted discrimination calculation data set M2;

[0038] 5.3. Repeat steps 5.1 and 5.2 to select K most similar stacked samples in the identification calculation data set M2 for each stacked waveform in the data set M1 to be identified; take the most common category among the K stacked samples as the category of the data set M1 to be identified, and complete the identification of the stacked pulse waveform.

[0039] Furthermore, in step 1.1, a detection system of stilbene crystal + photomultiplier tube + oscilloscope is used to collect the neutron gamma single waveform mixed signal, and sampling points greater than two pulse time scales are recorded when collecting the neutron gamma single waveform mixed signal.

[0040] Furthermore, in step 4.4, the peak factors K1 and K2 are adjusted by step size, specifically:

[0041] After the peak factors K1 and K2 are added to the step length respectively, new peak factors K1* and K2* are obtained, and the peak factors K1=K1* and K2=K2* are updated.

[0042] Further, in step 2.2, the delay time T is 10 to 40 ns;

[0043] In step 3.2, n=50;

[0044] In step 3.3, m=4.

[0045] Further, in step 4.3, the initial values ​​of the peak factors K1 and K2 are 1;

[0046] In step 4.4, the step size is 0.001;

[0047] In step 5.2, K=11.

[0048] Beneficial effects of the present invention:

[0049] 1. The present invention discloses a method for distinguishing a piled-up pulse waveform based on a charge integration method and a KNN algorithm. The charge integration method is used to distinguish a neutron single-wave signal from a gamma single-wave signal. Based on the difference between a long-window integral value and a short-window integral value of the charge integration method, the long-window integral value and the short-window integral value are used as characteristic parameters for selecting a single-particle waveform (i.e., a neutron single-wave signal or a gamma single-wave signal), thereby preventing single-particle waveforms with small or no obvious differences from being repeatedly selected for processing piled-up waveforms.

[0050] 2. The present invention provides a method for identifying a piled-up pulse waveform based on a charge integration method and a KNN algorithm. The method uses a selected neutron gamma single wave signal sample set U2 to process and identify a calculation data set M2 for each piled-up waveform in the data set M1 to be identified. The piled-up waveforms in the data set M1 to be identified are compared for similarity with all categories of the corresponding identification calculation data set M2 in turn, and the K piled-up samples that are most similar to the piled-up waveforms in the data set M1 to be identified are selected. If most of the K most similar piled-up samples belong to a certain category, the data set M1 to be identified is classified into that category, thereby reducing meaningless repeated calculations. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of an embodiment of a method for identifying piled-up pulse waveforms based on a charge integration method and a KNN algorithm of the present invention;

[0052] Figure 2 It is a schematic diagram of using the discrimination factor Y of the charge integration method to discriminate a single wave signal in an embodiment of a method for discriminating a piled pulse waveform based on a charge integration method and a KNN algorithm of the present invention. Figure 2 The left side of the red line in the figure represents the gamma single wave signal, and the right side represents the neutron single wave signal;

[0053] Figure 3This is a schematic diagram of an embodiment of a method for identifying piled-up pulse waveforms based on a charge integration method and a KNN algorithm of the present invention, in which a single-wave signal is screened using a long-window integral value Q1 and a short-window integral value Q2 of the charge integration method. In the figure, black represents a neutron single-wave signal, and red represents a gamma single-wave signal. DETAILED DESCRIPTION

[0054] like Figure 1 As shown, a pile-up pulse waveform identification method based on charge integration method and KNN algorithm includes the following steps:

[0055] S1, using the detection system of stilbene crystal + photomultiplier tube + oscilloscope to collect neutron gamma single waveform mixed signal, using the charge integration method to distinguish the neutron single wave signal and gamma single wave signal in the neutron gamma single waveform mixed signal, and constructing the neutron gamma single wave signal sample library U1;

[0056] 1.1. Collect neutron gamma single waveform mixed signal;

[0057] 1.2. Define the amplitude integral between the starting time t1 and the end point (base point) t of the neutron gamma single waveform mixed signal as the long window integral value Q1 of the charge integration method; the amplitude integral between the 50% peak time t2 of the trailing edge of the neutron gamma single waveform mixed signal and the end point (long window integral end point) t is the short window integral value Q2 of the charge integration method; the ratio of the short window integral value Q2 to the long window integral value Q1 is used as the discrimination factor Y; the long window integral value Q1 and the short window integral value Q2 obtained after integrating a single waveform are the same value, and the long window integral value Q1 and the short window integral value Q2 obtained after integrating different waveforms are different.

[0058] 1.3, such as Figure 2 As shown, the neutron-gamma single-wave mixed signal is discriminated, recorded and marked according to the discrimination factor Y, and the marked neutron single-wave signal and gamma single-wave signal form a neutron-gamma single-wave signal sample library U1.

[0059] In this embodiment, when collecting neutron gamma single waveform mixed signals, sampling points greater than two pulse time scales are recorded. When processing two single wave signals into a piled waveform, one of the single wave signals needs to be shifted back. Sampling points greater than two pulse time scales are recorded to ensure that after one of the single wave signals is shifted back, there are sufficient sampling points for processing the piled waveform.

[0060] S2, processing the single wave signal in the neutron gamma single wave signal sample library U1 into the data set M1 to be identified;

[0061] 2.1. Randomly select a neutron single wave signal as the first waveform and a gamma single wave signal as the second waveform in the neutron gamma single wave signal sample library U1;

[0062] 2.2. After aligning the peaks of the first waveform and the second waveform, shift the second waveform backward and record the shift time T as 30ns;

[0063] 2.3. Add the first waveform and the second waveform after the backward shift to obtain an n+γ type stacking waveform; the combination categories of the stacking waveform also include n+n, γ+n, γ+γ, where n is a neutron single wave signal and γ is a gamma single wave signal;

[0064] 2.4. Follow the methods of steps 2.1 to 2.3 until 1000 n+γ stacking waveforms are obtained.

[0065] 2.5. According to the methods of steps 2.1 to 2.4, the combination categories of n+n, γ+n, and γ+γ are processed in turn to process the single wave signals in the neutron gamma single wave signal sample library U1, and the combination categories of n+n, γ+n, and γ+γ type stacking waveforms are obtained, and 1000 n+γ type stacking waveforms, 1000 n+n type stacking waveforms, 1000 γ+n type stacking waveforms, and 1000 γ+γ type stacking waveforms are used as the data set M1 to be identified, and the corresponding shift time of each stacking waveform in the data set M1 to be identified is recorded as T;

[0066] S3, such as Figure 3 As shown, the charge integration method is used to screen the neutron single wave signal and the gamma single wave signal in the neutron gamma single wave signal sample library U1 to obtain the neutron gamma single wave signal sample set U2;

[0067] 3.1. Using the long window integral value Q1 and the short window integral value Q2 as parameters, the neutron single wave signal and the gamma single wave signal in the neutron gamma single wave signal sample library U1 are screened;

[0068] 3.2. Divide the long window integral value Q1 into 50 intervals, with an interval ΔQ1 = Q1 / 50;

[0069] 3.3. Take the short window integral value Q2 corresponding to the interval ΔQ1, and divide the short window integral value Q2 into 4 intervals, whose interval ΔQ2 = Q2 / 4;

[0070] 3.4. Select a single wave signal in each of the 200 intervals, select a total of 200 neutron single wave signals and 200 gamma single wave signals, and record and mark them to form a neutron gamma single wave signal sample set U2;

[0071] S4, constructing a discrimination calculation data set M2 based on the to-be-discriminated data set M1 and the neutron gamma single wave signal sample set U2; since the first peak value, the second peak value, and the peak distance of each to-be-discriminated piled waveform in the to-be-discriminated data set M1 are different, it is necessary to process each to-be-discriminated piled waveform for the discrimination calculation sample set M2;

[0072] 4.1. Select the i-th stacked waveform from the dataset to be identified M1 as the reference waveform W i ;

[0073] 4.2. Select any two single-wave signals from the neutron-gamma single-wave signal sample set U2, wherein the first waveform of the two selected single-wave signals is a neutron single-wave signal, and the second waveform is a gamma single-wave signal;

[0074] 4.3. Multiply the first waveform by the peak factor K1, multiply the second waveform by the peak factor K2, and shift the second waveform back by the same time interval as the reference waveform; then add the first waveform after the shift to the second waveform to obtain an n+γ type stacked waveform; add the two single wave signals after the shift to obtain an n+γ type stacked waveform; in this embodiment, the initial values ​​of the peak factors K1 and K2 are 1, and the second waveform is shifted back by the same time interval as the reference waveform W. i The same time interval T i ;

[0075] 4.4. Compare the two peaks of the n+γ-type stacking waveform obtained in step 4.3 with the reference waveform W obtained in step 4.1. i By comparing the two peak values ​​of the stacked waveform with the reference waveform, it is determined whether the difference between the peak value of the stacked waveform and the peak value of the reference waveform is less than or equal to 1%. If so, step 4.5 is executed; if not, the peak factors K1 and K2 are adjusted by a step size of 0.001, so that the new peak factors K1* and K2* are K1*=K1+0.001 and K2*=K2+0.001 respectively, and the peak factors K1=K1* and K2=K2* are updated, and the process returns to step 4.3;

[0076] 4.5. Repeat the method of steps 4.2 to 4.4, so that the first waveform traverses all neutron waveforms of the neutron gamma single wave signal sample set U2 once, and the second waveform traverses all gamma waveforms of the neutron gamma single wave signal sample set U2 once, and obtains 40,000 discrimination calculation subsets (n+γ type stacking waveforms) corresponding to the reference waveform, and marks them;

[0077] 4.6. Repeat steps 4.1 to 4.5 according to the combination categories of n+n, γ+n, and γ+γ, and traverse all the stacked waveforms in the dataset M1 to be identified to complete the reference waveform W. i The discrimination calculation data set M 2i Processing, identification and calculation of data set M 2i The combination categories include 40,000 n+n, γ+n, and γ+γ type stacked waveforms, which are processed into a discrimination calculation set for all the stacked waveforms in the dataset M1 to be discriminated, and are used as the discrimination calculation set M2; the discrimination calculation set M2 consists of each reference waveform W i Corresponding discrimination calculation data set M2 i constitute;

[0078] S5, calculate the similarity between the piled waveform to be identified in the dataset M1 to be identified and the piled waveform in the corresponding dataset M2 to be identified, and sort them according to the similarity, and use the KNN algorithm to select the K piled samples with the highest similarity in the sorting; take the category with the largest number of K piled samples as the category of the dataset M1 to be identified, and complete the piled pulse waveform identification, specifically:

[0079] 5.1. Use the Euclidean distance to calculate the similarity between any stacked waveform in the to-be-identified data set M1 and its corresponding stacked waveform in the identification calculation data set M2, and sort them from large to small;

[0080] 5.2. Use the KNN algorithm to select 11 stacking samples from the sorted discrimination calculation data set M2;

[0081] 5.3. In steps 5.1 and 5.2, for each stacked waveform in the data set M1 to be identified, select the 11 most similar stacked samples with the highest similarity in the identification calculation data set M2; take the category with the largest number of the 11 stacked samples as the category of the data set M1 to be identified, until all waveforms in the data set M1 to be identified are identified, and the identification of the stacked pulse waveform is completed.

Claims

1. A method for identifying pile-up pulse waveforms based on charge integration method and KNN algorithm, characterized in that: The following steps are involved: S1, collecting neutron gamma single waveform mixed signals, using the charge integration method to identify the neutron single wave signal and the gamma single wave signal in the neutron gamma single waveform mixed signal, and constructing a neutron gamma single wave signal sample library U1; S2, processing the single wave signal in the neutron gamma single wave signal sample library U1 into a data set M1 to be identified; the stacked waveform categories of the data set M1 to be identified are n+n, n+γ, γ+n, γ+γ, where n is a neutron single wave signal and γ is a gamma single wave signal; S3, using the charge integration method to screen the neutron single wave signal and the gamma single wave signal in the neutron gamma single wave signal sample library U1, to obtain the neutron gamma single wave signal sample set U2; S4, constructing a discrimination calculation data set M2 based on the to-be-discriminated data set M1 and the neutron gamma single wave signal sample set U2; S5. Calculate the similarity between the piled waveform to be identified in the dataset M1 and the piled waveform in the corresponding dataset M2, and sort them according to the similarity. Use the KNN algorithm to select the K piled samples with the highest similarity in the sorting; take the category with the largest number of K piled samples as the category of the dataset M1 to be identified, and complete the identification of the piled pulse waveform.

2. The method for identifying piled-up pulse waveforms based on charge integration method and KNN algorithm according to claim 1, characterized in that: Step S1 is specifically as follows: 1.

1. Collect neutron gamma single waveform mixed signal; 1.

2. Define the amplitude integral between the starting time t1 and the end point t of the neutron gamma single waveform mixed signal as the long window integral value Q1 of the charge integration method; the amplitude integral between the 50% peak time t2 of the trailing edge of the neutron gamma single waveform mixed signal and the end point t as the short window integral value Q2 of the charge integration method; and use the ratio of the short window integral value Q2 to the long window integral value Q1 as the discrimination factor Y; 1.

3. According to the discrimination factor Y, the neutron-gamma single-wave mixed signal is discriminated, recorded and marked, and the marked neutron single-wave signal and gamma single-wave signal are combined into a neutron-gamma single-wave signal sample library U1.

3. The method for identifying piled-up pulse waveforms based on charge integration method and KNN algorithm according to claim 2, characterized in that: Step S2 is specifically as follows: 2.

1. Randomly select a neutron single wave signal as the first waveform and a gamma single wave signal as the second waveform in the neutron gamma single wave signal sample library U1; 2.

2. After aligning the peaks of the first waveform and the second waveform, shift the second waveform backward and record the shift time T; 2.

3. Add the first waveform and the second waveform after the back shift to obtain an n+γ type stacking waveform; the combination categories of the stacking waveform also include n+n, γ+n, γ+γ, where n is a neutron single wave signal and γ is a gamma single wave signal; 2.

4. Follow the methods of steps 2.1 to 2.3 until all n+γ stacking waveforms are obtained; 2.

5. Process the single-wave signals in the neutron gamma single-wave signal sample library U1 according to the methods of steps 2.1 to 2.4 for the combination categories of n+n, γ+n, and γ+γ, and use them as the data set M1 to be identified.

4. The method for identifying piled-up pulse waveforms based on charge integration method and KNN algorithm according to claim 3 is characterized in that: Step S3 is specifically as follows: 3.

1. Using the long window integral value Q1 and the short window integral value Q2 as parameters, the neutron single wave signal and the gamma single wave signal in the neutron gamma single wave signal sample library U1 are screened; 3.

2. Divide the long window integral value Q1 into n intervals, with an interval ΔQ1 = Q1 / n, where n is greater than or equal to 40; 3.

3. Take the short window integral value Q2 corresponding to the interval ΔQ1, and divide the short window integral value Q2 into m intervals, where the interval ΔQ2 = Q2 / m, and the value of m is greater than or equal to 4; 3.

4. Select a single wave signal in each of the n*m ​​intervals to form the neutron gamma single wave signal sample set U2.

5. The method for identifying piled-up pulse waveforms based on charge integration method and KNN algorithm according to claim 4, characterized in that: Step S4 is specifically as follows: 4.

1. Select any stacked waveform from the dataset M1 to be identified as a reference waveform; 4.

2. Randomly select a neutron single wave signal as the first waveform and a gamma single wave signal as the second waveform from the neutron gamma single wave signal sample set U2; 4.

3. Multiply the first waveform by the peak factor K1, multiply the second waveform by the peak factor K2, and shift the second waveform backward by the same time interval as the reference waveform; then add the shifted first waveform to the second waveform to obtain an n+γ type stacking waveform; 4.

4. Compare the peak value of the n+γ type stacked waveform obtained in step 4.3 with the peak value of the reference waveform obtained in step 4.1 to determine whether the difference between the peak value of the stacked waveform and the peak value of the reference waveform is less than or equal to 1%. If so, execute step 4.5; if not, adjust the peak factors K1 and K2 by step size and return to step 4.3; 4.

5. Repeat the method of steps 4.2 to 4.4, so that the first waveform traverses all neutron waveforms of the neutron gamma single wave signal sample set U2 once, and the second waveform traverses all gamma waveforms of the neutron gamma single wave signal sample set U2 once, obtain the discrimination calculation subset corresponding to the reference waveform, and mark it; 4.

6. Repeat the methods of steps 4.1 to 4.5 according to the combination categories of n+n, γ+n, and γ+γ, traverse all the stacked waveforms in the data set M1 to be identified, process the identification calculation set for all the stacked waveforms in the data set M1 to be identified, and use it as the identification calculation set M2.

6. The method for identifying piled-up pulse waveforms based on charge integration method and KNN algorithm according to claim 5, characterized in that: Step S5 is specifically as follows: 5.

1. Use the Euclidean distance to calculate the similarity between any stacked waveform in the to-be-identified data set M1 and its corresponding stacked waveform in the identification calculation data set M2, and sort them from large to small; 5.

2. Use the KNN algorithm to select K stacking samples from the sorted discrimination calculation data set M2; 5.

3. Repeat steps 5.1 and 5.2 to select K most similar stacked samples in the identification calculation data set M2 for each stacked waveform in the data set M1 to be identified; take the most common category among the K stacked samples as the category of the data set M1 to be identified, and complete the identification of the stacked pulse waveform.

7. The method for identifying piled-up pulse waveforms based on charge integration method and KNN algorithm according to claim 6, characterized in that: In step 1.1, a detection system of stilbene crystal + photomultiplier tube + oscilloscope is used to collect the neutron gamma single waveform mixed signal, and sampling points greater than two pulse time scales are recorded when collecting the neutron gamma single waveform mixed signal.

8. The method for identifying piled-up pulse waveforms based on charge integration method and KNN algorithm according to claim 7, characterized in that: In step 4.4, the peak factors K1 and K2 are adjusted by step size, specifically: After the peak factors K1 and K2 are added to the step length respectively, new peak factors K1* and K2* are obtained, and the peak factors K1=K1* and K2=K2* are updated.

9. The method for identifying piled-up pulse waveforms based on charge integration method and KNN algorithm according to claim 8, characterized in that: In step 2.2, the delay time T is 10 to 40 ns; In step 3.2, n=50; In step 3.3, m=4.

10. The method for identifying piled-up pulse waveforms based on charge integration method and KNN algorithm according to claim 9, characterized in that: In step 4.3, the initial values ​​of the peak factors K1 and K2 are 1; In step 4.4, the step size is 0.001; In step 5.2, K=11.

Images