Neutron-gamma pulse discrimination method and system based on accumulated power trajectory slope

Abstract

The application discloses a neutron-gamma pulse discrimination method and system based on accumulated power trajectory slope, which comprises the following steps: extracting a pulse tail sequence based on a pulse peak value; performing recursive low-pass calculation on the pulse tail sequence and obtaining a system low-frequency component; obtaining a high-frequency residual component according to the pulse tail sequence and the system low-frequency component; calculating accumulated power corresponding to the system low-frequency component and the high-frequency residual component to construct a logarithmic accumulated power trajectory; performing linear fitting on the logarithmic accumulated power trajectory, and taking a fitting slope of the fitting line as a decay tail index; and realizing pulse discrimination of neutron events and gamma events based on the accumulated power trajectory slope based on the discrimination threshold and the decay tail index. The application solves the problem that the existing method cannot effectively extract stable digital discrimination features from the dynamic decay difference of the pulse falling edge tail, and realizes the differentiation of neutron and gamma events under the conditions of limited calculation resources / storage space, noise interference and low sampling rate.

Description

Technical Field

[0001] This invention relates to the field of neutron-gamma pulse discrimination technology, and in particular to a neutron-gamma pulse discrimination method and system based on the slope of the cumulative power trajectory. Background Technology

[0002] Neutrons and gamma rays may be recorded simultaneously by detectors during radiation detection. Because the energy deposition processes of these two types of particles differ in scintillation materials, the resulting pulses exhibit different attenuation characteristics at the tail of their fall edges. Therefore, using digital pulse shapes to distinguish between neutron and gamma events is an important technique in nuclear radiation measurement, nuclear safety monitoring, radiation imaging, space radiation detection, and nuclear material testing.

[0003] Existing digital neutron-gamma discrimination methods include charge comparison, peak-to-charge ratio, gradient, and pulse gradient analysis. The charge comparison method primarily obtains discrimination parameters by calculating the ratio of the tail-integrated charge to the total integrated charge. While simple to implement, it compresses the time variation at the pulse tail into a single integration result. The peak-to-charge ratio method reduces the computational cost of integration but is sensitive to noise and pulse amplitude fluctuations. Gradient-based methods utilize the local slope of the falling edge, making them susceptible to high-frequency noise, sampling rate, and sampling phase. Machine learning-based methods typically require large training samples and high computational resources, hindering implementation in embedded or low-power devices.

[0004] Therefore, there is an urgent need for a neutron-gamma pulse discrimination method and system that can utilize the dynamic decay information of the pulse tail, has a simple calculation process, calibrable parameters, and is suitable for real-time digital processing, so as to extract stable digital discrimination features from the dynamic decay difference of the falling edge tail of neutron pulses and gamma pulses, and achieve effective differentiation between neutron events and gamma events under the conditions of limited computing resources, limited storage space, noise interference, and low sampling rate. Summary of the Invention

[0005] This invention provides a neutron-gamma pulse discrimination method and system based on the cumulative power trajectory slope to overcome the above-mentioned technical problems.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows: A neutron-gamma pulse discrimination method based on the slope of the cumulative power trajectory, specifically including the following steps: S1: Acquire the analog pulse output by the preset detector; The analog pulses output by the detector are sampled at equal time intervals according to the preset sampling rate, and the pulse amplitude at each sampling time is quantized to obtain a digital pulse sequence. S2: Perform baseline correction on the digital pulse sequence to obtain the pulse amplitude after baseline correction; The pulse peak value is determined based on the baseline-corrected pulse amplitude; the pulse tail sequence is extracted based on the pulse peak value. S3: Perform recursive low-pass calculation on the pulse tail sequence to obtain the system low-frequency component; obtain the high-frequency residual component based on the pulse tail sequence and the system low-frequency component; S4: Calculate the cumulative power of the corresponding low-frequency component and high-frequency residual component of the system respectively; and construct the logarithmic cumulative power trajectory based on the cumulative power; S5: Using the least squares method, a straight line is fitted to the logarithmic cumulative power trajectory within a preset fitting interval to obtain a fitted straight line; the fitting slope of the fitted straight line is used as the decay tail index; the decay tail index distribution of neutron events and gamma events is statistically analyzed on preset calibration data; and a discrimination threshold is defined based on the decay tail index distribution; based on the discrimination threshold and the decay tail index, pulse discrimination of neutron events and gamma events based on the slope of the cumulative power trajectory is realized.

[0007] Furthermore, the formula for obtaining the pulse tail sequence described in S2 is as follows:

[0008]

[0009] In the formula: Indicates the position of the pulse peak. The next k sampling points The tail amplitude is denoted as ; N Indicates the length of the tail analysis interval; Indicates the position of the pulse peak; This indicates the pulse amplitude after baseline correction; Indicates the first n The pulse amplitude at each sampling point and ; Represents the set of sampling points; This represents the baseline value.

[0010] Furthermore, step S3 specifically includes the following steps: S31: Recursively low-pass filter the pulse tail sequence to calculate and obtain the system low-frequency components as follows:

[0011] In the formula: Indicates the first k The recursive low-pass component of each sampling point is the low-frequency component of the system. Represents the recursion coefficient used to adjust the memory length of recursive processing; Indicates the firstk- A recursive low-pass component with one sampling point; S32: Based on the pulse tail sequence and the system's low-frequency components, the high-frequency residual components are obtained as follows:

[0012] In the formula: This represents the high-frequency residual component.

[0013] Furthermore, step S4 specifically includes the following steps: S41: Calculate the cumulative power of the corresponding low-frequency component and high-frequency residual component of the system respectively:

[0014]

[0015] In the formula: Indicates the number of pulse peaks after the peak. The recursive low-pass component corresponding to each sampling point; Indicates the number of pulse peaks after the peak. The high-frequency residual components corresponding to each sampling point; This represents the cumulative power of the low-frequency components; Represents the cumulative power of the high-frequency residual components; S42: Construct a logarithmic cumulative power trajectory based on the cumulative power, the expression of which is:

[0016] In the formula: Indicates the first k The trajectory values ​​corresponding to each sampling point; Indicates a stable term.

[0017] Furthermore, the method in S5 for pulse discrimination between events and gamma events based on the cumulative power trajectory slope, based on the discrimination threshold and the attenuation tail index, is as follows: Define the discrimination threshold as T Determine the current decay tail index. With discrimination threshold T Size; like Then the corresponding simulated pulse will be identified as a neutron pulse; like If so, the corresponding analog pulse will be identified as a gamma pulse.

[0018] A neutron-gamma pulse discrimination system based on the slope of the cumulative power trajectory includes a digital pulse sequence acquisition module, a pulse tail sequence extraction module, a system low-frequency component calculation module, a high-frequency residual component calculation module, a log-cumulative power trajectory construction module, an attenuation tail index acquisition module, a discrimination threshold acquisition module, and a pulse event discrimination module. The digital pulse sequence acquisition module is used to sample the analog pulses output by the detector at equal time intervals according to a preset sampling rate, and to quantize the pulse amplitude at each sampling time to obtain a digital pulse sequence. The pulse tail sequence extraction module is used to perform baseline correction on the digital pulse sequence, obtain the baseline-corrected pulse amplitude, confirm the pulse peak value based on the baseline-corrected pulse amplitude, and extract the pulse tail sequence based on the pulse peak value. The system low-frequency component calculation module is used to perform recursive low-pass calculation on the pulse tail sequence and obtain the system low-frequency component. The high-frequency residual component calculation module is used to obtain the high-frequency residual component based on the pulse tail sequence and the system low-frequency component; The logarithmic cumulative power trajectory construction module is used to calculate the cumulative power of the corresponding low-frequency component and high-frequency residual component of the system respectively; and construct the logarithmic cumulative power trajectory based on the cumulative power. The attenuation tail index acquisition module is used to perform linear fitting on the logarithmic cumulative power trajectory within a preset fitting interval using the least squares method to obtain a fitted straight line; and to use the fitting slope of the fitted straight line as the attenuation tail index. The discrimination threshold acquisition module is used to statistically analyze the decay tail index distribution of neutron events and gamma events on preset calibration data; and to define a discrimination threshold based on the decay tail index distribution. The pulse event discrimination module is used to discriminate between neutron events and gamma events based on the cumulative power trajectory slope, based on the discrimination threshold and the decay tail index.

[0019] This invention provides a neutron-gamma pulse discrimination method and system based on the slope of the cumulative power trajectory, with the following beneficial effects: (1) The present invention uses recursive low-pass decomposition to decompose the pulse tail sequence into low-frequency components and high-frequency residual components, which can simultaneously characterize the slow decay trend and rapid change information of the tail. (2) The present invention constructs a logarithmic cumulative power trajectory by using low-frequency cumulative power and high-frequency cumulative power, and uses the least squares method to perform linear fitting on it to obtain the trajectory slope corresponding to the logarithmic cumulative power trajectory and use it as the EDTI discrimination factor, thus avoiding the problem of traditional methods using only static integral quantities. (3) The EDTI discrimination factor of the present invention can directly determine the discrimination threshold based on the calibration data, which facilitates parameter adjustment under different detectors and different sampling conditions; (4) The present invention adopts point-by-point recursive update, two-way cumulative power update and fixed interval linear fitting to complete the tail feature calculation in a single pulse processing process. The processing device only needs to save the recursive state quantity, the cumulative power state quantity and the statistics required for fitting, which reduces the use of computing resources. (5) The present invention can reduce the direct impact of local fluctuations at a single sampling point on the cumulative power trajectory by separating the recursive low-pass component and the high-frequency residual component; and under specific noise conditions, the recursive coefficients and fitting intervals can be determined by calibration data. (6) This application can output the classification results of neutron or gamma pulses, as well as the corresponding EDTI values, discrimination thresholds and statistical performance indicators, which are suitable for online radiation monitoring and real-time neutron-gamma event identification. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of the neutron-gamma pulse discrimination method based on the cumulative power trajectory slope of the present invention; Figure 2 The waveforms of the gamma pulse and neutron pulse in this embodiment are after peak alignment, normalization and baseline correction; Figure 3 This is a graph showing the changes in low-frequency components and high-frequency residual components of the pulse tail after recursive low-pass processing in this embodiment as a function of post-peak time. Figure 4 This is a graph showing the cumulative power of the low-frequency component and high-frequency residual component at the pulse tail as a function of post-peak time in this embodiment. Figure 5 This is a diagram showing the linear fitting results of the logarithmic cumulative power trajectory constructed from low-frequency and high-frequency cumulative power, as well as the neutron and gamma trajectories, in this embodiment. Figure 6 This is a parameter analysis diagram showing the influence of the recursive coefficient alpha on the discrimination factor FoM in this embodiment; Figure 7 This is a parameter analysis graph showing the influence of the fitting starting point on the discrimination factor FoM in this embodiment; Figure 8This is a parameter analysis graph showing the effect of the fitting length on the discrimination factor FoM in this embodiment; Figure 9 This is a parameter analysis diagram showing the effect of the stability term epsilon on the discrimination factor FoM in this embodiment; Figure 10 This is an experimental distribution diagram of the discrimination factors for neutron events and gamma events in this embodiment; Figure 11 This is a comparison chart of the discrimination factor FoM between the method described in this embodiment and the gradient method, peak-to-charge ratio method, and charge comparison method. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] This embodiment provides a neutron-gamma pulse discrimination method based on the slope of the cumulative power trajectory. The method described in this embodiment processes the digital pulses output by the detector and outputs the discrimination result of a neutron event or a gamma event based on the attenuation difference between the neutron pulse and the gamma pulse at the tail of the falling edge. Figure 1 As shown, the specific steps include: S1: Acquire the analog pulses output by the preset detector; sample the analog pulses output by the detector at equal time intervals according to the preset sampling rate, and quantize the pulse amplitude at each sampling time to obtain a digital pulse sequence; Specifically, the input to the method described in this embodiment is a single pulse sequence acquired by the detector and digitally processed; the input to this embodiment is a single digital pulse sequence formed by analog-to-digital conversion of the analog pulses output by the detector; specifically, in this embodiment, a preset analog-to-digital converter is used to sample the analog pulses output by the detector at equal time intervals according to a preset sampling rate, and the pulse amplitude at each sampling moment is quantized to obtain a digital pulse sequence composed of L sampling points.

[0024] In the formula: Indicates the first n The pulse amplitude at each sampling point; L Indicates the number of sampling points contained in a single digital pulse; S2: Perform baseline correction on the digital pulse sequence to obtain the baseline-corrected pulse amplitude; confirm the pulse peak value based on the baseline-corrected pulse amplitude; extract the pulse tail sequence based on the pulse peak value; Specifically, in this embodiment, digital pulses collected by the detector are acquired, and baseline correction is performed on the digital pulses in the digital pulse sequence. The average amplitude of each sampling point within the pre-trigger interval before the pulse arrives is taken as the baseline value B, and the baseline value B is subtracted from the original pulse amplitude to obtain the baseline-corrected pulse amplitude:

[0025]

[0026] In the formula: This indicates the pulse amplitude after baseline correction; Indicates the first n The pulse amplitude at each sampling point and ; Represents the set of sampling points; Indicates the baseline value; The pulse peak position p is determined based on the pulse amplitude after baseline correction, and peak alignment is performed using a conventional sampling point translation method so that the peaks of different pulses correspond to the same reference position. Peak alignment only adjusts the relative position of the sampling points without changing the pulse amplitude. When used for waveform comparison, the peak amplitude can also be used to normalize the pulse. Figure 2 The waveforms of the gamma pulse and neutron pulse after alignment are given, as well as the time interval used for tail analysis and fitting. In this embodiment, baseline correction, peak location, peak alignment and amplitude normalization are all conventional digital signal processing techniques in this field, and will not be described in detail here. The pulse tail sequence is obtained as follows:

[0027]

[0028] In the formula: Indicates the position of the pulse peak. The later k sampling points The tail amplitude is denoted as ; N Indicates the length of the tail analysis interval; Indicates the position of the pulse peak; S3: Perform recursive low-pass computation on the pulse tail sequence to obtain the system low-frequency component; obtain the high-frequency residual component based on the pulse tail sequence and the system low-frequency component, specifically including the following steps: S31: Recursively low-pass filter the pulse tail sequence to calculate and obtain the system low-frequency components as follows:

[0029] In the formula: Indicates the first k The recursive low-pass component of each sampling point is the low-frequency component of the system. This represents the recursion coefficient used to adjust the memory length of recursive processing. The larger the recursion coefficient, the more sensitive the low-frequency components are to tail changes over a longer time range. Indicates the first k- A recursive low-pass component with one sampling point; S32: Based on the pulse tail sequence and the system's low-frequency components, the high-frequency residual components are obtained as follows:

[0030] In the formula: Represents high-frequency residual components; Figure 3 and Figure 4 Dynamic variation curves of the low-frequency component and high-frequency residual component at the tail are given respectively to illustrate the component differences between neutron pulses and gamma pulses during the tail decay process. S4: Calculate the cumulative power of the corresponding low-frequency component and high-frequency residual component of the system respectively; and construct the logarithmic cumulative power trajectory based on the cumulative power, specifically including the following steps: S41: Calculate the cumulative power of the corresponding low-frequency component and high-frequency residual component of the system respectively:

[0031]

[0032] In the formula: Indicates the number of pulse peaks after the peak. The recursive low-pass component corresponding to each sampling point; Indicates the number of pulse peaks after the peak. The high-frequency residual components corresponding to each sampling point; This represents the cumulative power of the low-frequency components; This represents the cumulative power of the high-frequency residual components; Figure 4 The variation of cumulative power with post-peak time is given for two types of cases; S42: Construct a logarithmic cumulative power trajectory based on the cumulative power, the expression of which is:

[0033] In the formula: Indicates the first k The trajectory values ​​corresponding to each sampling point; This indicates a stability term, used to avoid division by zero or numerical instability when the high-frequency cumulative power is low; S5: Using the least squares method, a straight line is fitted to the logarithmic cumulative power trajectory within a preset fitting interval to obtain the fitted straight line, the expression of which is:

[0034] In this embodiment, the least squares method determines the fitting slope by minimizing the sum of squared residuals between the fitted line and each trajectory point. and the fitting intercept b; Figure 5 Fitting results for gamma pulse trajectories and neutron pulse trajectories are presented, where... Represents the slope of the neutron trajectory and =0.0003585 / sampling point; Indicates the slope of the gamma trajectory and =0.0002027 / sampling point; Because the tail of a neutron pulse decays slowly, its low-frequency cumulative power growth process is different from that of a gamma pulse, so the trajectory slopes of the two types of pulses are different; The slope of the fitted straight line As a tail indicator of attenuation : ; The distribution of decay tail indices for neutron and gamma events is statistically analyzed on pre-set known calibration data, such as... Figure 10 As shown; and a discrimination threshold is defined based on the attenuation tail index distribution; based on the discrimination threshold and the attenuation tail index, pulse discrimination of neutron events and gamma events based on the cumulative power trajectory slope is realized. Specifically, the method for pulse discrimination of events and gamma events based on the cumulative power trajectory slope based on the discrimination threshold and the attenuation tail index is as follows: Define the discrimination threshold as T Determine the current decay tail index. With discrimination threshold T Size; like Then the corresponding simulated pulse will be identified as a neutron pulse; like If so, the corresponding analog pulse will be identified as a gamma pulse.

[0035] like Figures 6 to 9 The analysis shows the effects of recursion coefficients, fitting start point, fitting length, and stability term on the discrimination factor FoM, and based on this, determines the operating parameters suitable for the current detector and sampling conditions. Figure 11 The FoM comparison results of the method described in this embodiment with other digital pulse discrimination methods are presented.

[0036] Based on the same inventive concept, this embodiment also includes a neutron-gamma pulse discrimination system based on the cumulative power trajectory slope, including a digital pulse sequence acquisition module, a pulse tail sequence extraction module, a system low-frequency component calculation module, a high-frequency residual component calculation module, a logarithmic cumulative power trajectory construction module, an attenuation tail index acquisition module, a discrimination threshold acquisition module, and a pulse event discrimination module. The digital pulse sequence acquisition module is used to sample the analog pulses output by the detector at equal time intervals according to a preset sampling rate, and to quantize the pulse amplitude at each sampling time to obtain a digital pulse sequence. The pulse tail sequence extraction module is used to perform baseline correction on the digital pulse sequence, obtain the baseline-corrected pulse amplitude, confirm the pulse peak value based on the baseline-corrected pulse amplitude, and extract the pulse tail sequence based on the pulse peak value. The system low-frequency component calculation module is used to perform recursive low-pass calculation on the pulse tail sequence and obtain the system low-frequency component. The high-frequency residual component calculation module is used to obtain the high-frequency residual component based on the pulse tail sequence and the system low-frequency component; The logarithmic cumulative power trajectory construction module is used to calculate the cumulative power of the corresponding low-frequency component and high-frequency residual component of the system respectively; and construct the logarithmic cumulative power trajectory based on the cumulative power. The attenuation tail index acquisition module is used to perform linear fitting on the logarithmic cumulative power trajectory within a preset fitting interval using the least squares method to obtain a fitted straight line; and to use the fitting slope of the fitted straight line as the attenuation tail index. The discrimination threshold acquisition module is used to statistically analyze the decay tail index distribution of neutron events and gamma events on preset calibration data; and to define a discrimination threshold based on the decay tail index distribution. The pulse event discrimination module is used to discriminate between neutron events and gamma events based on the cumulative power trajectory slope, based on the discrimination threshold and the decay tail index.

[0037] The method described in this embodiment solves the following problems: (1) Traditional integration methods cannot fully utilize the time change information of the pulse tail; (2) High-frequency noise and local waveform disturbances affect the stability of gradient-based discrimination methods; (3) Complex algorithms are difficult to implement in real time on embedded platforms; (4) Fixed parameters are difficult to maintain stable discrimination performance under different detectors, sampling rates and noise conditions; (5) A digital discrimination method is needed that can determine the working parameters and discrimination threshold through calibration data.

[0038] This embodiment also includes the following application examples: Example 1: This example uses public Am-Be neutron-gamma pulse data to illustrate the digital neutron-gamma pulse shape discrimination method of the present invention. The pulses are acquired by an EJ299-33 plastic scintillator and a digital oscilloscope, with a sampling rate of 1 GS / s, a vertical resolution of 8 bits, a hardware trigger threshold of 500 mV, and an equivalent electron energy of approximately 1.6 MeVee. For each event, baseline correction is first performed, and the pulse peak position p is determined. Subsequently, the tail local sequence after the peak is extracted according to the steps described above, and the low-frequency component, residual component, low-frequency cumulative power, and high-frequency cumulative power are calculated based on the recursive low-pass relation given above. The logarithmic cumulative power ratio trajectory is further constructed and linear fitting is performed over a fixed interval.

[0039] In this embodiment, the recursion coefficient The value is set to 0.98. The tail fitting starts at 60 sampling points after the peak value, and the fitting length is 160 sampling points. The physical time range corresponding to the fitting interval is from 60 ns to 220 ns after the peak value. The relatively stable term... We take 10⁻⁸. All continuous trajectory points within the fitting interval participate in ordinary least squares fitting, and the trajectory slope is used as the tail dynamic feature.

[0040] In this embodiment, peak value normalization of the pulse waveform can be used for illustrated diagrams and waveform comparisons between different events, but actual feature calculations do not require readjusting the fitting starting point, fitting length, or fitting weights for each event. For each event, the stability term is relatively scaled by the square of the peak amplitude to avoid the denominator approaching zero under different pulse energies. Calculations show that the trajectory slope of the neutron pulse is greater than that of the gamma pulse, indicating that the tail dynamics can reflect the difference in tail persistence between neutron and gamma pulses.

[0041] This embodiment divides all common Am-Be events into a calibration set and a hold-out test set, with the calibration set accounting for 30% and the hold-out test set accounting for 70%. The recursion coefficients, fitting start point, fitting length, relatively stable terms, feature normalization parameters, and discrimination thresholds are all determined by the calibration set and remain unchanged during the hold-out test phase. Parameter calibration uses a single-factor analysis approach, changing only one parameter at a time while keeping the others constant, and determining the operating point based on the quality factors on the calibration set. Considering the performance of clean data, relevant noisy data, and independent datasets, the operating point is selected. =0.98, a fitting start point of 60 ns, and a fitting length of 160 ns were selected as the preferred operating points.

[0042] Example 2: This example illustrates the applicability of the method described in this example under different datasets, noise conditions, and sampling rates. For public Am-Be data, all comparison methods use the same set of events for calculation, without quantile truncation or event removal for any particular method. The set of events in the public Am-Be dataset contains 10,083 events, and the mean quality factors (FoM) for the EDTI, CC, PCPR, and LG methods are 1.44±0.02, 1.42±0.02, 1.31±0.02, and 1.16±0.01, respectively. EDTI has the highest FoM point estimate, but because the confidence intervals of EDTI and CC methods overlap, this example only illustrates that EDTI achieves a higher point estimate and does not represent this difference as a statistically significant absolute advantage.

[0043] This embodiment also uses independent Time-of-Flight (ToF) labeled data for verification. The ToF data has a sampling rate of 500 MS / s, a sampling interval of 2 ns, and each event contains 496 sampling points. 20% of the events are divided into a calibration set and 80% into a hold-out test set, with the energy window determined based on the 30th and 80th percentiles of the energy distribution in the calibration set. Calibration parameters, feature normalization parameters, and classification thresholds are all determined by the calibration set and remain unchanged during the hold-out test phase.

[0044] On the Time-of-Flight (ToF) test set of 1,772 events, EDTI achieved a FoM of 1.24, an AUC of 0.9982, and 22 / 1,772 false events, resulting in an error rate of 1.24%. In comparison, the FoM values ​​for the LG, PCPR, and CC methods were 0.79, 0.98, and 1.09, respectively. For the normalized EDTI features, under the calibration conditions of this embodiment, the discrimination threshold is approximately -0.13; the actual threshold is determined based on the neutron and gamma feature distributions in the calibration data and is not limited to the aforementioned values.

[0045] To investigate the impact of noise, Gaussian white noise, band-limited correlated noise, and dark impulse interference were added to the impulse data. For random noise conditions, 10 independent noise sets were used for each test point, and the average of the 10 results was used as the quality factor, while the standard deviation was used to represent the fluctuation of the results. The results show that band-limited correlated noise introduces low-frequency perturbations into the impulse tail, which usually causes a more significant performance degradation than white noise; when the recurrence coefficient is greater than 0.98, the separation performance on clean data may continue to improve, but the robustness under correlated noise conditions will decrease.

[0046] To examine the impact of sampling rate, direct digital decimation was performed on the original 1 GS / s data. At 1000 MS / s, the FoM of EDTI was approximately 1.44; as the equivalent sampling rate decreased, the FoM was approximately 0.42 at 20 MS / s. The downsampling process described above did not incorporate anti-aliasing filtering or sampling phase scanning; therefore, this result is only intended to illustrate that the method of this invention can still complete feature calculation at low sampling rates and is not equivalent to the performance of a complete hardware ADC sampling model.

[0047] When the sampling rate changes from 1 GS / s to 500 MS / s, according to the equivalent time constant conversion relationship given above, the recursive coefficient is converted from 0.98 to 0.9604, and the fitting interval from 60 ns to 220 ns after the peak is converted to 80 sampling points starting from 30 sampling points after the peak. If it is necessary to directly compare the slope characteristics between different sampling rates, the slope of each sampling point is converted into a slope per unit physical time; if this conversion is not performed, the normalization parameters and discrimination thresholds are determined from the calibration data for different sampling rates.

[0048] Example 3: This example illustrates the implementation of the method described in this example in an embedded digital signal processing device. The method described in this example is deployed on an STM32H750 microcontroller, which uses a 480 MHz Cortex-M7 core. For a single input pulse event, the processor sequentially executes the steps described above: baseline correction, peak localization, tail sequence extraction, recursive low-pass update, residual calculation, two-channel cumulative power update, logarithmic cumulative power ratio calculation, and fixed interval linear fitting. Since the fitting slope is calculated using a closed-form least squares formula, feature extraction can be completed without saving all trajectory points.

[0049] In the finite word length implementation, recursive updates and cumulative power calculations employ Q31 and Q15 fixed-point formats, respectively. The FoM obtained by the Q31 fixed-point implementation is the same as the reported value of the floating-point implementation, while the FoM of the Q15 fixed-point implementation is reduced by approximately 1.0% compared to the floating-point implementation. Logarithmic and division operations can be implemented using fast mathematical functions provided by the microcontroller or pre-computed lookup tables. The square accumulation process uses an accumulator with sufficient bit width, and a scaling factor is set according to the input amplitude range to reduce the risk of overflow in fixed-point operations.

[0050] During the online discrimination process, calibration data is first used to determine the recursive coefficients, fitting start point, fitting length, relative stability term, normalization parameters, and discrimination threshold, and these parameters are then written into the digital signal processing device. For each pulse to be discriminated, the processor uses fixed parameters to perform feature extraction and normalization, compares the obtained EDTI features with the discrimination threshold, and outputs either a neutron event or a gamma event. During the online processing phase, the fitting interval is not changed based on the energy of a single event, the fitting weights are not recalculated, and the discrimination threshold is not updated using online test events.

[0051] When the sampling interval or front-end response of the digital pulse changes, the recursive coefficients, fitting time interval, normalization parameters, and discrimination threshold are redefined based on the corresponding calibration data, and the updated parameters are then used for pulse discrimination. The parameter adjustments do not alter the processing structure of recursive decomposition, two-channel cumulative power calculation, and trajectory slope extraction.

[0052] The core technical solution of the method described in this embodiment is as follows: within the tail region after the peak of a single digital pulse, a low-frequency component is first obtained through recursive low-pass operation, and then a high-frequency residual component is obtained by subtracting the low-frequency component from the original tail sequence; cumulative power calculations are performed on the low-frequency component and the high-frequency residual component respectively; a tail trajectory is constructed using the logarithmic ratio of the low-frequency cumulative power with a positive stability term and the high-frequency cumulative power; and the linear fitting slope of the trajectory within a preset tail fitting interval is used as the neutron-gamma discrimination feature. The above combination of "recursive decomposition—residual calculation—two-way cumulative power—logarithmic power ratio trajectory—fitting slope determination" is the main protection content of this application. Furthermore, the above embodiment is used to illustrate the recursive tail decomposition, two-way cumulative power construction, logarithmic cumulative power trajectory fitting, and slope determination method described in this application. Except for the specific parameters given in the embodiment, the recursive coefficients, fitting start point, fitting length, stability term, and discrimination threshold can be adjusted according to calibration data without changing the above processing steps and their order. The above processing steps and their order are used to form the tail dynamic discrimination feature. The recursion coefficients, fitting start point, fitting length, stability term, and discrimination threshold are determined based on the calibration data, and the specific values ​​are not the sole limitation on the technical solution of this application.

[0053] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A neutron-gamma pulse discrimination method based on the slope of the cumulative power trajectory, characterized in that, The specific steps include: S1: Acquire the analog pulse output by the preset detector; The analog pulses output by the detector are sampled at equal time intervals according to the preset sampling rate, and the pulse amplitude at each sampling time is quantized to obtain a digital pulse sequence. S2: Perform baseline correction on the digital pulse sequence to obtain the pulse amplitude after baseline correction; Confirm the pulse peak value based on the baseline-corrected pulse amplitude; Extracting the pulse tail sequence based on the pulse peak value; S3: Perform recursive low-pass calculation on the pulse tail sequence to obtain the system low-frequency component; Based on the pulse tail sequence and the system's low-frequency components, the high-frequency residual components are obtained; S4: Calculate the cumulative power of the corresponding low-frequency component and high-frequency residual component of the system respectively; And construct a logarithmic cumulative power trajectory based on the cumulative power; S5: Using the least squares method, a straight line is fitted to the logarithmic cumulative power trajectory within a preset fitting interval to obtain a fitted straight line; the fitting slope of the fitted straight line is used as the decay tail index; the decay tail index distribution of neutron events and gamma events is statistically analyzed on preset calibration data; and a discrimination threshold is defined based on the decay tail index distribution; based on the discrimination threshold and the decay tail index, pulse discrimination of neutron events and gamma events based on the slope of the cumulative power trajectory is realized.

2. The neutron-gamma pulse discrimination method based on the cumulative power trajectory slope according to claim 1, characterized in that, The formula for obtaining the pulse tail sequence described in S2 is: In the formula: Indicates the position of the pulse peak. The next k sampling points The tail amplitude is denoted as ; N Indicates the length of the tail analysis interval; Indicates the position of the pulse peak; This indicates the pulse amplitude after baseline correction; Indicates the first n The pulse amplitude at each sampling point and ; Represents the set of sampling points; This represents the baseline value.

3. The neutron-gamma pulse discrimination method based on the cumulative power trajectory slope according to claim 2, characterized in that, S3 specifically includes the following steps: S31: Recursively low-pass filter the pulse tail sequence to calculate and obtain the system low-frequency components as follows: In the formula: Indicates the first k The recursive low-pass component of each sampling point is the low-frequency component of the system. Represents the recursion coefficient used to adjust the memory length of recursive processing; Indicates the first k- A recursive low-pass component with one sampling point; S32: Based on the pulse tail sequence and the system's low-frequency components, the high-frequency residual components are obtained as follows: In the formula: This represents the high-frequency residual component.

4. The neutron-gamma pulse discrimination method based on the cumulative power trajectory slope according to claim 3, characterized in that, S4 specifically includes the following steps: S41: Calculate the cumulative power of the corresponding low-frequency component and high-frequency residual component of the system respectively: In the formula: Indicates the number of pulse peaks after the peak. The recursive low-pass component corresponding to each sampling point; Indicates the number of pulse peaks after the peak. The high-frequency residual components corresponding to each sampling point; This represents the cumulative power of the low-frequency components; Represents the cumulative power of the high-frequency residual components; S42: Construct a logarithmic cumulative power trajectory based on the cumulative power, the expression of which is: In the formula: Indicates the first k The trajectory values ​​corresponding to each sampling point; Indicates a stable term.

5. The neutron-gamma pulse discrimination method based on the cumulative power trajectory slope according to claim 4, characterized in that, The method in S5 for pulse discrimination of events and gamma events based on the cumulative power trajectory slope, based on the discrimination threshold and attenuation tail index, is as follows: Define the discrimination threshold as T Determine the current decay tail index. With discrimination threshold T Size; like Then the corresponding simulated pulse will be identified as a neutron pulse; like If so, the corresponding analog pulse will be identified as a gamma pulse.

6. A system based on the neutron-gamma pulse discrimination method based on the cumulative power trajectory slope as described in any one of claims 1 to 5, characterized in that, It includes a digital pulse sequence acquisition module, a pulse tail sequence extraction module, a system low-frequency component calculation module, a high-frequency residual component calculation module, a log-cumulative power trajectory construction module, an attenuation tail index acquisition module, a discrimination threshold acquisition module, and a pulse event identification module; The digital pulse sequence acquisition module is used to sample the analog pulses output by the detector at equal time intervals according to a preset sampling rate, and to quantize the pulse amplitude at each sampling time to obtain a digital pulse sequence. The pulse tail sequence extraction module is used to perform baseline correction on the digital pulse sequence and obtain the pulse amplitude after baseline correction. Confirm the pulse peak value based on the baseline-corrected pulse amplitude; Extracting the pulse tail sequence based on the pulse peak value; The system low-frequency component calculation module is used to perform recursive low-pass calculation on the pulse tail sequence and obtain the system low-frequency component. The high-frequency residual component calculation module is used to obtain the high-frequency residual component based on the pulse tail sequence and the system low-frequency component; The logarithmic cumulative power trajectory construction module is used to calculate the cumulative power of the corresponding low-frequency component and high-frequency residual component of the system, respectively. And construct a logarithmic cumulative power trajectory based on the cumulative power; The attenuation tail index acquisition module is used to perform linear fitting on the logarithmic cumulative power trajectory within a preset fitting interval using the least squares method to obtain a fitted straight line; and to use the fitting slope of the fitted straight line as the attenuation tail index. The discrimination threshold acquisition module is used to statistically analyze the decay tail index distribution of neutron events and gamma events on preset calibration data; and to define a discrimination threshold based on the decay tail index distribution. The pulse event discrimination module is used to discriminate between neutron events and gamma events based on the cumulative power trajectory slope, based on the discrimination threshold and the decay tail index.

Images