Liquid scintillation TDCR multi-nuclide β spectrum analysis method and system based on artificial intelligence

Abstract

The present invention belongs to the field of nuclear radiation measurement technology and relates to an artificial intelligence-based liquid scintillator (TDCR) multi-nuclide β spectrum analysis method and system. The method comprises: using Monte Carlo technology to construct a numerical model of a liquid scintillator detector spectrometer to simulate the energy spectrum response of a single nuclide under different quenching conditions; using a parameterized data synthesis algorithm to construct a training database; constructing a multi-task mixed spectrum analysis neural network model; using the constructed database to train the model; inputting a measured spectrum, and using the trained neural network model to output the multi-nuclide absolute activity, detector efficiency, and decomposed energy spectrum contribution curve in real time. The present invention innovatively integrates the differential distribution characteristics of the β continuum energy spectrum with the characteristics of the quenching response curve to construct a neural network architecture with physical mechanism constraints, achieving decoupling of the liquid scintillator TDCR multi-nuclide energy spectrum characteristics and accurate activity solution under unknown quenching conditions.

Description

Technical Field

[0001] The present invention belongs to the technical field of nuclear radiation measurement, and relates to a liquid scintillator TDCR multi-nuclide β spectrum analysis method and system based on artificial intelligence. Background Art

[0002] Radioactive decay mainly includes three types: α, β and γ. Among them, β decay is accompanied by the momentum conservation feature of neutrino emission, and the energy of the electrons released presents a continuous energy spectrum distribution from zero to the maximum value, which is significantly different from the single energy peak of α particles and the discrete characteristic peak of γ rays. This continuous energy spectrum feature makes β spectrum measurement very useful in nuclear facility status monitoring (such as nuclear fuel cycle 90 Sr activity analysis), environmental radiation assessment ( 3 Identification of low-energy nuclides such as H) and nuclear medicine diagnosis ( 14 It is worth noting that the demand for the analysis of mixed β-nuclides systems is becoming increasingly prominent: dual-nuclides systems (such as 3 H- 14 C) has become a standard analytical target in scenarios such as biomedical tracing and simultaneous detection of multiple pollutants, and demand for mixed systems of three or more nuclides in the development of new nuclear medicines continues to grow. However, the continuous distribution of the beta energy spectra of different nuclides leads to a high degree of overlap in the energy range. This, especially under unknown quenching conditions, makes the rapid, convenient, and efficient calculation of the absolute activity of multiple nuclides a current technical bottleneck.

[0003] Currently, liquid scintillation spectrometer (LSA) is the core equipment for beta spectrum measurement. Its international mainstream models include the American PerkinElmer Tri-Carb series, the Finnish Hidex 300SL, and the domestic Shanghai Xinman LSA3000 and Hubei Fangyuan FY2700-00. These devices generally adopt a triple photomultiplier tube (PMT) coincidence measurement architecture, and realize the adaptive calibration of the single-nuclide detector efficiency through the counting ratio (TDCR) of triple coincidence (TAC) and double coincidence (DAC). The TDCR method effectively suppresses noise interference through the three-tube coincidence logic, eliminates the efficiency drift caused by the quenching effect without the need for standard source calibration, and significantly improves the efficiency of single-nuclide low-activity samples (such as environmental 3 This breakthrough has led to its inclusion in the International Bureau of Weights and Measures (BIPM) as a benchmark method for radioactivity measurement and is widely used in standards laboratories around the world.

[0004] However, the traditional TDCR method faces a fundamental limitation in the analysis of multi-nuclides mixed system: when the sample contains multiple β-decay nuclides and the quenching degree is unknown, the existing method must rely on external calibration to establish the quenching efficiency curve. 3 H- 14Taking the C dual-nuclide system as an example, the current solutions are mainly based on two mutually exclusive strategies: Exclusion Method and Inclusion Method. The Exclusion Method achieves nuclide decoupling by energy spectrum partitioning: the high energy region (such as >50 keV) is selected as 14 C characteristic window, using the energy zone count and efficiency curve to solve 14 After the C activity is determined, the H activity is inverted by the composite signal in the low energy region. The inclusion rule allows the energy regions to overlap, and by constructing a system of simultaneous equations (CPM A = H·ε hA + C·ε cA ;CPM B = H·ε hB + C·ε cB ) to simultaneously solve the dual-nuclide activity. Although these two methods have been engineered in patented technologies (such as the efficiency calibration method of CN117949998A and the energy window interpolation method of CN114488264B), the existing TDCR method has common technical defects: First, its experimental system needs to rely on external sources and standard sources of nuclides to be measured to establish a multi-energy zone quenching efficiency response curve, and the strict supervision of radioactive materials and the scene limitations of conventional laboratories and portable equipment make it difficult to routinely equip standard sources (including supporting γ reference sources), which significantly increases the complexity of the experiment and hinders the promotion of standardization; secondly, the method has a subjective dependence on energy zone division, and the fixed energy window (such as artificial setting) is manually set. 14 C 50-156 keV) leads to fragmented utilization of energy spectrum information, especially in 14 C (Emax=156 keV) and 35 In scenarios with close energy spectrum proximity or strong quenching, such as S (Emax = 167 keV), fuzzy spectral boundaries and empirical screening can easily lead to significant cross-interference and systematic errors. Furthermore, traditional solution models employ fixed-dimensional parameters to construct simultaneous equations, and their rigid mathematical framework cannot adapt to changes in spectral morphology. Under complex quenching conditions, they face problems such as no solution, local optimal solutions, or dynamic calibration failure. These shortcomings collectively restrict the accuracy and applicability of multi-nuclide beta spectroscopy.

[0005] In recent years, the outstanding performance of deep neural networks in dealing with multivariable nonlinear problems has provided a new path to break through the limitations of traditional methods. Patent CN119763688 demonstrates a convolutional neural network neutron detector efficiency correction technology, and multi-nuclide beta spectrum analysis is essentially a multi-dimensional inverse problem involving quenching parameters, energy spectrum distribution characteristics, and nuclide activity. By constructing a numerical model of detector response to generate a training data set, the neural network can establish a direct mapping between energy spectrum characteristics and nuclide activity, which can theoretically avoid the traditional method's reliance on calibration sources and artificial energy zone divisions. This data-driven strategy provides a new technical paradigm for the development of adaptive multi-nuclide analysis algorithms, marking that beta spectrum analysis technology is evolving from experience-dependent to intelligent computing. Summary of the Invention

[0006] The present invention proposes a TDCR liquid scintillator multi-nuclide β spectrum analysis method and system based on neural network (NN). The system consists of two core modules: detector response modeling and multi-task mixed spectrum analysis neural network. At the technical implementation level, the Monte Carlo numerical simulation method is first used to construct the TDCR liquid scintillator detector response model. The system simulates different nuclides (such as , , , The method uses a multi-layer perceptron neural network architecture to establish a nonlinear mapping between multi-nuclide spectral response characteristics and absolute activity through an adaptive feature extraction mechanism. The key breakthrough of this method lies in the high-precision analysis of multi-nuclide β spectrum through the fusion of numerical simulation and deep learning, which simultaneously outputs the absolute activity of each nuclide and its corresponding β spectrum response.

[0007] The present invention provides an artificial intelligence-based liquid scintillation TDCR multi-nuclide β spectrum analysis method, comprising the following steps:

[0008] Step 1: Construct a numerical model of the liquid scintillation detector spectrometer to simulate the energy spectrum response of the liquid scintillation TDCR under different quenching conditions with two-channel logic and three-channel coincidence;

[0009] Step 2: Use a parameterized data synthesis algorithm to construct a training database, including different quenching conditions and multi-component mixture compositions;

[0010] Step 3: Build a multi-task mixed spectrum parsing neural network model; use the constructed database to perform cascade multi-task training of the model;

[0011] Step 4: Input the measured spectrum and use the trained neural network model to output the multi-nuclide absolute activity, detection efficiency, and decomposition energy spectrum contribution curve in real time.

[0012] Preferably, in step one, a numerical model of a liquid scintillation detector spectrometer is constructed using Monte Carlo technology, including: detector geometry modeling, nuclide decay process simulation, liquid scintillation performance simulation, optical physics process simulation, photomultiplier tube response, and signal digitization process simulation.

[0013] Preferably, in step 2, the method for constructing the training database includes: using the constructed numerical model to randomly generate quenching values, simulating the TDCR response energy spectrum of any two tubes and three tubes under different quenching conditions of a single nuclide; secondly, dynamically adjusting the multi-nuclide concentration ratio and quenching intensity distribution to generate a multi-component mixed energy spectrum.

[0014] Preferably, in step three, the multi-task mixed spectrum analysis neural network model adopts a two-stage cascade design: in the first stage, the mixed spectrum features are extracted through a shared fully connected layer, the activity and detector efficiency of each nuclide are simultaneously predicted, and the gradient magnitude adaptive dynamic weight distribution mechanism is used to dynamically balance the optimization goals of activity and detector efficiency; in the second stage, the activity and detector efficiency prediction results are integrated with the shared features to reconstruct the original spectrum of each nuclide and the mixed spectrum of two tubes and three tubes.

[0015] Preferably, the input of the multi-task mixed spectrum analysis neural network model is the TDCR response energy spectrum of the two-tube coincidence logic and the three-tube coincidence of multi-nuclide mixture, and the output is the activity of each nuclide, the detector efficiency, the original spectrum of each nuclide, and the decomposition spectrum of the two-tube and three-tube.

[0016] The present invention also provides an artificial intelligence-based liquid scintillation detection and response (TDCR) multi-nuclide beta spectrum analysis system, which includes a liquid scintillation spectrometer Monte Carlo numerical simulation module and a training module for a multi-task mixed spectrum analysis neural network model; the liquid scintillation spectrometer Monte Carlo numerical simulation module is used to simulate the two-channel coincidence logic and three-channel coincidence energy spectrum responses of the liquid scintillation detection and response (TDCR) of a single nuclide under different quenching conditions; the training module constructs a training database based on the simulated energy spectrum responses, and uses the training database to train the multi-task mixed spectrum analysis neural network model; the multi-task mixed spectrum analysis neural network model is used to predict the input measured spectrum and output the multi-nuclide absolute activity, detection efficiency, and decomposition energy spectrum contribution curve in real time.

[0017] Preferably, the system further comprises a display module for visually displaying the output results.

[0018] The artificial intelligence-based liquid scintillator TDCR multi-nuclide β spectrum analysis method and system provided by this invention overcomes the technical bottlenecks of traditional methods in three dimensions: dependence on calibration sources, fragmentation of energy spectrum analysis, and rigid model constraints. This method innovatively integrates the differential distribution characteristics of the β continuum energy spectrum with the characteristics of the quenching response curve to construct a neural network architecture constrained by physical mechanisms. This method achieves decoupling of multi-nuclide energy spectrum characteristics and accurate activity determination under unknown quenching conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart for modeling and optimizing the numerical model of the LSA detector in an embodiment of the present invention;

[0020] Figure 2 Schematic diagram of a cascaded multi-task mixed spectrum parsing neural network model in an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of an application example of the artificial intelligence-based liquid scintillation TDCR multi-nuclide β spectrum analysis system provided by the present invention. DETAILED DESCRIPTION

[0022] In order to facilitate the understanding of the present invention, the present invention is described in more detail below with reference to the accompanying drawings and specific embodiments. The preferred embodiments of the present invention are given in the accompanying drawings. All equivalent technical variations and improvement measures derived from the core technical solution of the present invention, and all implementation plans realized by conventional experimental means are within the scope of legal protection, specifically covering but not limited to the following technical dimensions: optimization of the detection cavity structure, upgrade of the detector material components and replacement of photoelectric conversion devices at the hardware adaptation level; training sample expansion strategy, optimization of data batch processing strategy, enhancement of the filtering algorithm of the input / output energy spectrum, iteration of data cleaning and denoising technology and upgrade of parallel computing architecture in the data processing link; architecture replacement of the fully connected layer and convolution / deconvolution layer in the model architecture dimension, dynamic configuration of the number of neurons, selection of activation function, dynamic adjustment of the number of training times and early stopping mechanism, adaptive learning rate scheduling strategy and multi-task loss function optimization; reconstruction of model evaluation indicators, exploration of regularization parameter space and adjustment of Monte Carlo simulation parameters at the algorithm verification level; expansion of the nuclide database, improvement of signal processing technology and optimization of detector shielding materials at the extended application level, etc. This patent strictly adheres to the principle of central limitation. The scope of protection of the claims is not limited to the embodiments described in the specification. Any improvement scheme based on equivalent technical principles and achieved through conventional engineering optimization means, including but not limited to attention mechanism integration, residual connection topology optimization, batch normalization strategy adjustment, training data spectrum enhancement technology development, detector cavity material optimization and other technical variations, regardless of their form of expression and degree of optimization, are deemed to fall within the legal protection boundaries of the claims of this patent. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0023] Example 1 The artificial intelligence-based liquid scintillator TDCR multi-nuclide β spectrum analysis method provided by the present invention comprises the following steps:

[0024] Step 1: Numerical simulation of the liquid scintillation detector spectrometer. This paper uses a hierarchical modeling strategy to construct a full-link digital twin of the liquid scintillation spectrometer, realizing the energy spectrum response of a single nuclide under different quenching conditions. Through actual measurements, comparisons with simulations, and optimization, a numerical model of the liquid scintillation detector spectrometer is ultimately constructed.

[0025] Liquid scintillation spectrometer modeling process Figure 1 As shown, the following steps are included:

[0026] First, we build a detector physical model based on the Monte Carlo method, which includes:

[0027] 1. Detector Modeling

[0028] (1) Geometric structure: Taking a cylindrical hollow cavity as an example, the cavity has an inner diameter of 100 mm and a height of 200 mm. The inner layer of the cavity is covered with a 0.2 mm reflective layer. Three photomultiplier tubes (PMTs) with a diameter of 50 mm and a sensitive area diameter of 48 mm are installed on the side of the cavity. They are symmetrically distributed at 120°. The PMTs are coupled to the test cavity through a quartz window (refractive index 1.46). The sample is placed in the center of the test cavity.

[0029] (2) Material Properties: The test chamber material is polytetrafluoroethylene (PTFE) with a density of 0.89 g / cm³. The reflective layer within the chamber is polyethylene. A bidirectional scattering distribution function (BSDF) is defined on the reflective layer surface to simulate the mixed effects of diffuse and specular reflection. The sample vial and the PMT optical window are both molten quartz glass with a density of 2.2 g / cm³. Its refractive index is wavelength-dependent and is described by the Sellmeier equation, e.g., n = 1.470 @ 400 nm. The liquid scintillator has a refractive index of 1.59 and a light attenuation length of 35 cm.

[0030] 2. Liquid scintillator performance parameterization

[0031] (1) Light yield model

[0032] Based on the Birks quenching formula, the photon yield per unit path length is defined , where the light yield coefficient photons / Me, (Birks quenching constant), energy deposition rate Calculated by the continuous slowing down approximation (CSDA) of beta particles in liquid scintillator.

[0033] (2) Quenching nonlinearity correction

[0034] A lookup table interpolation of Scintillation Yield vs. dE / dx is introduced to cover the saturation effect in the low energy region (<100 keV), ensuring that the light yield is consistent with the experimental calibration data.

[0035] 3. Radioactive Source Sample Modeling

[0036] (1) β emission spectrum

[0037] Based on the nuclide decay database (such as NuDat 3.0), the Fermi function is called to calculate the theoretical β energy spectrum:

[0038] ;

[0039] in, is the mass of the electron, is the speed of light, is the maximum energy of the β particle, is the Coulomb correction factor for the nucleus.

[0040] (2) Spatial distribution

[0041] The radioactive source is evenly distributed in the sample bottle to avoid boundary effects; the decay times are set to times to ensure that the statistical fluctuation of the energy spectrum is <1%.

[0042] (3) Background noise

[0043] The spatiotemporal distribution of simulated environmental background (such as cosmic ray muons and natural radionuclides) is superimposed on the main signal to enhance the realism of the simulation.

[0044] 4. Refined modeling of physical processes

[0045] (1) Particle transport

[0046] Simulate the ionization loss of beta particles (electrons / positrons) during transport, the generation of delta rays, and the determination of the Cherenkov radiation threshold (triggered when the beta speed exceeds the speed of light in the liquid scintillator).

[0047] (2) Photon generation and collection

[0048] Each energy deposition event generates the number of photons according to the Birks formula, and the photon emission direction is isotropically distributed; photons experience reflection / refraction during transmission: the interface behavior is tracked through a geometric optical model; photons are absorbed / scattered in the liquid scintillator: the probability is calculated based on the absorption length and scattering length of the liquid scintillator.

[0049] The second is the readout electronics system, which includes PMT gain process simulation and electronics response simulation, specifically:

[0050] PMT gain process simulation: By establishing a photomultiplier tube (PMT) cascade amplification system and defining the PMT photoelectric conversion quantum efficiency (20% at 400 nm), the full-link conversion from single-photon pulses to PMT readout signals was simulated. The focus was on quantifying the impact of PMT gain fluctuations and dark noise interference on the energy spectrum morphology.

[0051] Signal digitization simulation: The analog pulse is converted into discrete energy spectrum data through the ADC model. The detector response function generated in this stage is combined with the physical simulation data to form a simulated energy spectrum, including the digitized signal and TDCR information.

[0052] By systematically comparing Monte Carlo simulation data with experimentally measured energy spectra and trigger coincidence information (TDCR), a parameter optimization mechanism for the detector physical model and front-end electronics system was established. After iterative calibration, the relative deviation between the simulated values ​​of the quenching response curve and detection efficiency and the experimental measured values ​​can be controlled within a 5% confidence interval. This invention constructs a universal simulation calibration framework that only requires a set of reference nuclides (such as ) experimental data can be used to calibrate the detector response characteristics, eliminating the need for measured data for all the nuclides to be measured. The calibrated Monte Carlo simulation system uses only a publicly available nuclide database to accurately reconstruct the energy spectrum characteristics of any nuclide through particle transport calculations and energy deposition models.

[0053] Step 2: Generation of multi-nuclide mixed energy spectrum data.

[0054] An enhanced training database is constructed by parameterized data synthesis algorithm. Using the verified liquid scintillation detector numerical model, the quenching parameters (light yield and Birks constant) are first randomly generated. , generate the basic quenching spectrum of each nuclide in two-tube / three-tube configuration , covering the energy range of 0.1-10 MeV, simulating the TDCR response energy spectrum of any two-tube and three-tube coincidences of a single nuclide under different quenching conditions.

[0055] On this basis, the multi-nuclide concentration ratio and quenching intensity distribution are dynamically adjusted to generate a multi-component mixed energy spectrum: Apply random activity , simulate different nuclide activity ratios; introduce PMT noise (Gaussian distribution ) and electronic baseline drift to simulate the real measurement environment. The final result is 10 7 The mixed spectrum sample database (80% for training, 10% for validation, and 10% for testing) covers two- and three-nuclide mixing scenarios, providing a high-fidelity and strong generalization training foundation for neural networks.

[0056] Step 3: Cascade multi-task mixed spectrum parsing neural network.

[0057] This paper constructs a deep learning model dedicated to mixed nuclear β spectrum decomposition. The main process of this part is shown in Figure 2 , its core architecture adopts a two-stage cascade design to achieve dual optimization from physical quantity inversion to spectral shape fine-tuning.

[0058] 1. Model Construction

[0059] Phase 1 (Physical Quantity Inversion): Input two-tube and three-tube TDCR multi-nuclide mixed spectra are passed through a shared fully connected layer (two-layer FC network with 128 and 645 neurons respectively, ReLU activation, Dropout regularization) to extract high-dimensional features and branch out independent tasks:

[0060] (1) Activity prediction: the three-layer FC network outputs the activity coefficient of each nuclide and imposes a non-negative constraint (Softplus output);

[0061] (2) Efficiency prediction: the two-layer FC network outputs the detection efficiency coefficient of each nuclide, and imposes a value constraint of 0-1 (Sigmoid output);

[0062] (3) Adopting gradient-level adaptive weight allocation: dynamically balancing the optimization weights of activity (MSE loss) and efficiency (MSE loss);

[0063] Phase 2 (spectral shape reconstruction and parameter fine-tuning): Activity, efficiency, and shared features are concatenated and fed into a spectral shape reconstructor with a lower learning rate. Two- and three-tube quenching spectra are reconstructed independently for each nuclide using an FC network (Huber loss), with non-negative constraints imposed (Softplus output), and smoothing filtering and denoising. Two- and three-tube mixed spectra are generated based on the physical mixing formula:

[0064] ;

[0065] ;

[0066] in, For the species of nuclides, N is the total number of nuclides in the mixed sample.

[0067] 2. Model training and evaluation

[0068] (1) Training strategy

[0069] Two-stage joint training: After 200 rounds of pre-training in the first stage, the parameters of the second stage are unlocked;

[0070] Early stopping mechanism: monitor the loss function value of the validation set and tolerate no improvement for 10 rounds;

[0071] Optimizer: AdamW (lr=1e-3 in the first round, 1e-4 in the second round).

[0072] During the training process, the model uses an early stopping strategy to monitor the validation set based on a mixed spectrum dataset containing different quenching conditions and concentration ratios to ensure generalization, and saves the best model based on the verification results.

[0073] (2) Evaluation system

[0074] Table 1 Evaluation parameters

[0075] index Physical meaning Target Prediction MSE (activity) Ingredient identification accuracy >0.95 MSE (efficiency) Detector response modeling accuracy <0.05 SSIM (spectral shape) Hybrid spectrum reconstruction fidelity >0.95 .

[0076] Step 4: Use the trained neural network model to output the multi-nuclide absolute activity, detection efficiency, and decomposition spectrum contribution curve in real time. For the specific process, see Figure 3 .

[0077] The user inputs a measured spectrum, and the model generates predictions. After loading the trained model, efficiency and activity predictions are performed, and a contribution spectrum is generated. The predicted activity values ​​and decomposition spectra are displayed to the user, facilitating analysis and decision-making. This process not only improves analysis efficiency but also reduces reliance on specialized knowledge.

[0078] Example 2 The present invention also provides an artificial intelligence-based liquid scintillator TDCR multi-nuclide β spectrum analysis system, which includes a liquid scintillation spectrometer Monte Carlo numerical simulation module, a training module, and a cascaded multi-task mixed spectrum analysis neural network model; the liquid scintillation spectrometer Monte Carlo numerical simulation module is used to simulate the energy spectrum response of a single nuclide under different quenching conditions; the training module constructs a training database based on the simulated energy spectrum response and uses the training database to train the cascaded multi-task mixed spectrum analysis neural network model; the cascaded multi-task mixed spectrum analysis neural network model is used to predict the input measured spectrum and output the multi-nuclide absolute activity, detector efficiency, and decomposition energy spectrum contribution curve in real time. The output results are displayed by the display module.

[0079] The online parsing engine loads the pre-trained model (best_model.pth) and outputs prediction results in real time. The human-computer interaction interface provides a spectrum comparison view, dynamic rendering of the confidence heat map (nuclide probability is mapped in color scale), and a data export module (CSV / ROOT format compatible).

Claims

1. Liquid scintillation TDCR multi-nuclide β spectrum analysis method based on artificial intelligence, characterized in that: The following steps are involved: Step 1: Construct a numerical model of the liquid scintillation detector spectrometer to simulate the energy spectrum response of the liquid scintillation TDCR under different quenching conditions, including two-channel logical sum and three-channel coincidence. Step 2: Use a parameterized data synthesis algorithm to construct a training database, including different quenching conditions and multi-component mixture compositions; Step 3: Construct a multi-task mixed spectrum analysis neural network model and use the constructed database to perform cascade multi-task training of the model; the multi-task mixed spectrum analysis neural network model adopts a two-stage cascade design: in the first stage, the mixed spectrum features are extracted through a shared fully connected layer, the activity and detector efficiency of each nuclide are simultaneously predicted, and the optimization goals of activity and detector efficiency are dynamically balanced using a gradient magnitude adaptive dynamic weight allocation mechanism; in the second stage, the activity and detector efficiency prediction results are integrated with the shared features to reconstruct the original spectrum of each nuclide and the mixed spectrum of two tubes and three tubes; the input of the multi-task mixed spectrum analysis neural network model is the TDCR response energy spectrum of the two-tube logical and three-tube coincidence of the multi-nuclide mixture, and the output is the activity of each nuclide, the detector efficiency, the original spectrum of each nuclide and the decomposition spectrum of the two tubes and three tubes; Step 4: Input the measured spectrum and use the trained neural network model to output the multi-nuclide absolute activity, detector efficiency, and decomposition energy spectrum contribution curve in real time.

2. The liquid scintillation TDCR multi-nuclide β spectrum analysis method based on artificial intelligence according to claim 1, characterized in that: In the step 1, a numerical model of a liquid scintillation detector spectrometer is constructed using Monte Carlo technology, including: detector geometry modeling, nuclide decay process simulation, liquid scintillation performance simulation, optical physics process simulation, photomultiplier tube response, and signal digitization process simulation.

3. The liquid scintillation TDCR multi-nuclide β spectrum analysis method based on artificial intelligence according to claim 1, characterized in that: In step 2, the training database construction method includes: using the constructed numerical model to randomly generate quenching values ​​to simulate the TDCR response energy spectrum of any two or three tubes under different quenching conditions for a single nuclide; secondly, dynamically adjusting the multi-nuclide concentration ratio and quenching intensity distribution to generate a multi-component mixed energy spectrum.

4. Liquid scintillation TDCR multi-nuclide β spectrum analysis system based on artificial intelligence, characterized by: The system includes a liquid scintillation spectrometer Monte Carlo numerical simulation module and a training module for a multi-task mixed spectrum analysis neural network model; the liquid scintillation spectrometer Monte Carlo numerical simulation module is used to simulate the two-channel logical and three-channel coincident energy spectrum responses of liquid scintillation TDCR of a single nuclide under different quenching conditions; the training module constructs a training database based on the simulated energy spectrum responses, and uses the training database to train the multi-task mixed spectrum analysis neural network model; the multi-task mixed spectrum analysis neural network model is used to predict the input measured spectrum and output the absolute activity, detection efficiency, and decomposition energy spectrum contribution curve of multiple nuclides in real time; when the system is in operation, the steps of the method described in any one of claims 1 to 3 are implemented.

5. The artificial intelligence-based liquid scintillator TDCR multi-nuclide β spectrum analysis system according to claim 4, characterized in that: The system also includes a display module for visually displaying the output results.

Images