Nuclear radiation detection management system based on artificial intelligence

Abstract

The invention relates to the technical field of nuclear radiation detection management, and discloses a nuclear radiation detection management system based on artificial intelligence, comprising: a dynamic sensing module which detects nuclear radiation through quantum sensing to obtain radiation data, captures NV color center fluorescence intensity, analyzes to obtain gamma ray dose rate, constructs a nanoscale olfactory receptor simulation structure, and sends the radiation data to a control module; radioactive gas is detected and cooperatively monitored; the reasoning calculation module is used for analyzing space-time correlation based on radiation data, optimizing algorithm to accelerate result output, adjusting calculation complexity by calculating dynamic weight distillation, and correcting data transmission in real time; the decision management module is used for constructing a four-dimensional causal graph based on the radiation data, performing intervention simulation, evaluating a decision long-term effect, screening an optimal decision, and performing adversarial training based on historical repair case data; and the fusion cooperation module is used for establishing a cross-mechanism federated learning framework, carrying out hash operation on the radiation data, and establishing an NFT (Network Fourier Transform) on a block chain through an intelligent contract for query, verification and transaction.

Description

Technical Field

[0001] The present invention relates to the technical field of nuclear radiation detection management, and particularly relates to an artificial intelligence-based nuclear radiation detection management system. Background Art

[0002] The development of nuclear radiation detection management systems mainly stems from the need for radiation safety monitoring in fields such as nuclear power generation, medical treatment, and industry. For example, nuclear power plants need to monitor the radiation levels around the reactor in real time to ensure safe operation. In the medical field, it is necessary to calibrate the dose of radiotherapy equipment. In industry, radiation protection is required in scenarios such as uranium mining. In addition, nuclear accident emergency response drives the demand for large-scale and rapid monitoring technologies. Early nuclear radiation monitoring mainly relied on manual inspections and fixed-point monitoring. Manual inspections had low efficiency and posed a radiation risk to personnel. Although fixed-point monitoring could obtain data in real time, the traditional wired transmission method had distance limitations and signal attenuation problems. With technological progress, optical fiber communication, Beidou short message, and the Internet of Things have been introduced to improve the stability and coverage of data transmission. For example, the long-distance transmission solution based on optical fiber has solved the problem of large data loss in traditional systems, and the Beidou short message technology has enhanced the monitoring ability in remote areas. In recent years, artificial intelligence and machine learning technologies have begun to be applied to nuclear radiation detection, such as abnormal pattern recognition and data analysis. At the same time, breakthroughs in chip technology have enabled the miniaturization and low power consumption of devices, which can be integrated into intelligent devices such as mobile phones and drones to adapt to diverse scenarios. In the prior art, traditional detectors have problems such as large detection errors, inability to distinguish radiation types, and performance degradation in high-radiation environments. Some systems have a slow response speed and cannot give real-time warnings. Although the new chips have improved the response speed, they have not been fully popularized. The data formats and communication protocols of devices from different manufacturers are not unified, resulting in difficulties in system integration and making it difficult to integrate data between fixed monitoring stations and mobile devices, affecting overall decision-making. Traditional wired transmission is limited by distance, and wireless transmission has unstable signals in remote areas. In addition, the storage and analysis of massive monitoring data rely on centralized servers, which are prone to processing delays. Data collection, analysis, and alarm threshold setting still require manual operations, with low efficiency and easy errors, and cannot be adaptively adjusted according to real-time data. Most existing systems lack machine learning support, making it difficult to achieve abnormal pattern prediction and intelligent decision-making, and most systems still rely on manual experience. In view of this, there is a need to provide an artificial intelligence-based nuclear radiation detection management system. Summary of the Invention

[0003] The purpose of the present invention is to provide an artificial intelligence-based nuclear radiation detection management system. To solve the above-mentioned problems in the prior art, the present invention is achieved through the following technical solutions: An artificial intelligence-based nuclear radiation detection management system provided by an embodiment of the present invention specifically includes the following modules: Dynamic Sensing Module: Detect nuclear radiation through quantum sensing to obtain radiation data, capture the fluorescence intensity of NV color centers based on photodetectors, analyze the changes to calculate the γ-ray dose rate, construct a nanoscale olfactory receptor simulation structure, detect radioactive gases through biomimetic detection, and use drone swarms for collaborative monitoring; Inference Calculation Module: Analyze the spatio-temporal correlation based on the obtained radiation data, optimize the algorithm to accelerate the result output, adjust the calculation complexity by calculating the dynamic weight distillation, and correct the data transmission in real time through strengthening the hardware; Decision Management Module: Construct a four-dimensional causal graph based on the radiation data after correction, conduct intervention simulation based on the four-dimensional causal graph, evaluate the long-term effect of the decision and screen the optimal decision, and construct a generative adversarial network for adversarial training based on historical repair case data; Fusion and Collaboration Module: Build a cross-institutional federated learning framework, perform hash operations on the radiation data, and create NFTs on the blockchain through smart contracts for querying, verification, and trading.

[0004] Furthermore, the method for detecting nuclear radiation through quantum sensing is as follows: Using diamond as the substrate, introduce nitrogen-vacancy (NV) color center defects into the diamond lattice through ion implantation technology; Precisely control the concentration and distribution of nitrogen atoms and NV color center defects, and use electron spin resonance technology to initialize and read out the NV color centers; Adopt a low-temperature environment and microwave resonance excitation to change the electron spin state of the NV color centers and perform highly sensitive detection of γ-rays; Based on the interaction between γ-rays and diamond NV color centers, the electron spin state of the NV color centers changes; Obtain the intensity information of γ-rays by measuring the change in the spin state; By measuring the change in fluorescence intensity, collect the fluorescence signal based on a highly sensitive photodetector and convert the fluorescence signal into a fluorescence electrical signal.

[0005] Furthermore, the method for obtaining the γ-ray dose rate is as follows: Perform filtering processing on the collected fluorescence electrical signals; Through the Fourier transform signal processing algorithm, convert the time-domain signal into a frequency-domain signal, extract the characteristic parameters related to the γ-ray intensity, that is, the fluorescence intensity; Set that there is a linear relationship between the fluorescence intensity I and the γ-ray dose rate D; Based on the measured fluorescence intensity I, calculate the γ-ray dose rate D using the linear relationship between the fluorescence intensity I and the γ-ray dose rate D.

[0006] Further, the method for biomimetic detection of radioactive gas is as follows: Synthesize a protein with a similar function to mammalian olfactory receptors through genetic engineering technology and immobilize the protein on the surface of the chip; Fabricate a microfluidic channel using lithography and etching processes to enable the radioactive gas to come into full contact with the nano-scale olfactory receptor simulation structure; Based on the binding of the radioactive gas to the nano-scale olfactory receptor simulation structure on the chip surface, which causes a change in the protein conformation, and further leads to a change in the chip surface resistance and chip surface capacitance; Obtain the concentration information of the radioactive gas by measuring the changes in the chip surface resistance and chip surface capacitance; Precisely capture the weak electrical signal changes through a 24-bit Δ-Σ type analog-to-digital converter ADC; Perform normalization processing on the electrical signals and map the resistance and capacitance signal data to the [0,1] interval; Perform frame processing on the radiation data, divide the continuous signal sequence into frames of a fixed length, and extract the mean, standard deviation, and spectral energy of each frame to construct a feature vector; Based on the feature vector, construct a prediction model using a multi-layer perceptron MLP architecture; Input the feature vector of the electrical signal collected in real time and preprocessed into the prediction model, and the result output by the prediction model is the predicted value of the concentration of the radioactive gas in the current environment.

[0007] Further, the method for collaborative monitoring using a drone swarm is as follows: Carry a lightweight quantum-biological hybrid sensor on the drone. By optimizing the circuit design and structural layout of the sensor, reduce its weight and volume. The drone swarm adopts a self-organizing network communication protocol; During the flight of the drone, the radiation data is collected in real time through the carried sensor and transmitted to the ground control center through the wireless communication module; The ground control center performs data cleaning and format conversion on the received radiation data; Use a swarm intelligence algorithm to analyze the radiation data and determine the radiation hot spot area.

[0008] Further, the method for accelerating the result output is as follows: Introduce spatio-temporal sparsification decision in the attention mechanism. By analyzing the spatio-temporal correlation of the radiation data, determine the key spatio-temporal positions, perform attention calculation on the key spatio-temporal positions, ignore other redundant positions, and use pruning technology to remove unimportant connections and parameters; Organize the collected radiation data according to the time series and spatial positions to form a spatio-temporal data matrix; perform normalization processing on the spatio-temporal data matrix; The input radiation data undergoes multi-layer attention calculations and is processed by a feed-forward neural network to construct a Transformer model to obtain the output result. In the spatio-temporal sparsified Transformer, the input radiation data is X ∈ R N×T×D , where N is the number of spatial positions, T is the number of time steps, and D is the data feature dimension; through spatio-temporal sparsification decision-making, attention calculations are performed on K key spatio-temporal positions.

[0009] Furthermore, the method for adjusting the computational complexity is as follows: During the operation of the AI inference engine, the radiation intensity of the environment is monitored in real time; According to the different radiation intensities, the weight parameters of the prediction model are dynamically adjusted; If the radiation intensity is less than the preset radiation intensity threshold, a model with a higher complexity is used for accurate prediction; if the radiation intensity is greater than or equal to the preset radiation intensity threshold, it is determined as a key pollution event, and the knowledge of the complex model is transferred to the simple model through the weight distillation technique to reduce the model complexity; The environmental radiation intensity is used as an additional input feature and is input into the model together with the radiation data. During the model training stage, the optimal weight adjustment decision under different radiation intensities is learned through a reinforcement learning algorithm; Based on the obtained environmental radiation intensity R and the weight parameters of the model W, a functional relationship W = g(R) is established, where g(R) is the weight relationship function, and the weight parameter W is dynamically adjusted according to the radiation intensity R.

[0010] Furthermore, the method for performing intervention simulation based on the four-dimensional causal graph is as follows: Based on the constructed four-dimensional causal graph, counterfactual reasoning technology is used for intervention simulation; for different treatment schemes, corresponding intervention nodes are set in the four-dimensional causal graph to simulate the radiation propagation situation after the implementation of different schemes; Based on Monte Carlo simulation, different initial conditions and environmental changes are simulated multiple times to evaluate the long-term effects of the treatment schemes; During the simulation process, the parameters in the four-dimensional causal graph are adjusted according to the actual situation, and the simulation results are statistically analyzed to calculate the mean and variance of the reduction in radiation concentration and the reduction in the affected range.

[0011] Furthermore, the method for performing adversarial training is as follows: Historical repair case data is collected, and a generative adversarial network is constructed by a generator and a discriminator for adversarial training. The generator generates candidate repair schemes based on real-time radiation data and historical case data, and the discriminator distinguishes between the generated schemes and the real effective schemes; through adversarial training, the generator continuously optimizes the generated schemes; Extract and encode the features of historical repair case data, and convert the historical repair case data into a format suitable for network input; preprocess the real-time data, extract key features, and combine them with the historical case data as the input of the generator.

[0012] Further, the method for creating NFT is as follows: During the upload process, differential privacy technology is used to add noise to the radiation data. The central server receives the information uploaded by each institution, aggregates and updates it, and then distributes the updated federated learning framework to each institution; Perform a hash operation on the radiation data of key pollution events to generate a unique digital fingerprint; Obtain relevant metadata, package the digital fingerprint together with the relevant metadata, and create a non-fungible token NFT on the blockchain through a smart contract; each participating party can query, verify, and trade the NFT through the blockchain platform.

[0013] Advantages of the present invention: 1. Based on the electron spin characteristics of diamond NV centers, detect the γ-ray dose rate through the change in fluorescence intensity, break through the detection lower limit of traditional Geiger counters, simulate olfactory receptors using MEMS chips, sense the concentration of radioactive gases through resistance / capacitance changes, combine neural networks to achieve signal pattern recognition, and use a cluster of radiation-resistant drones equipped with lightweight sensors to collaboratively track radiation hotspots through swarm intelligence algorithms to improve monitoring efficiency; by analyzing the spatio-temporal correlation of radiation data, compress the number of model parameters, improve the inference speed, dynamically switch the model complexity according to the real-time radiation intensity, use silicon carbide encapsulated chips, and integrate error correction coding circuits to ensure the reliability of data processing in harsh environments; fuse multi-source data, mine the causal relationship between variables through causal discovery algorithms, construct a graph database to support counterfactual reasoning, use Monte Carlo simulation and counterfactual reasoning to evaluate the long-term effects of governance plans, and combine generative adversarial networks to generate intelligent repair plans; share model parameters across institutions instead of raw data, ensure privacy and security through noise addition processing, generate NFTs for the data of key pollution events, and use the uniqueness of hashing and the immutability of the blockchain to support traceability and liability determination; 2. Synchronously obtain multi-dimensional data of γ-ray dose rate, radioactive gas concentration, and spatial distribution, support the three-dimensional characterization of the radiation field, reduce the model calculation complexity in high-radiation environments, ensure the real-time response ability of edge devices in harsh conditions, improve the radiation resistance of silicon carbide encapsulated and error correction coding chips, and be used in extreme scenarios around nuclear facilities and radioactive pollution areas. Through the identification of key propagation paths, such as the impact of groundwater seepage on radiation diffusion, targeted blocking decisions can be made, and combined with counterfactual reasoning, the effects of the plan can be predicted in advance. Description of the Drawings

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0015] Figure 1 It is a schematic structural diagram of a nuclear radiation detection and management system based on artificial intelligence provided in Embodiment 1 of the present invention. Detailed implementation manners

[0016] To enable those skilled in the art to better understand the solutions of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0017] Embodiment 1 As Figure 1 shown, a nuclear radiation detection and management system based on artificial intelligence provided in the embodiment of the present invention specifically includes the following modules: Dynamic perception module: Detect nuclear radiation through quantum sensing to obtain radiation data, capture the fluorescence intensity of NV color centers based on a photodetector, analyze the changes to calculate the gamma-ray dose rate, construct a nanoscale olfactory receptor simulation structure, detect radioactive gases through biomimicry, and use a drone swarm for collaborative monitoring; It should be noted that the radiation data includes: radiation monitoring data, geological exploration data, meteorological observation data, and biological sample data; In a specific embodiment, the specific method for detecting nuclear radiation through quantum sensing is as follows: For quantum sensing, using diamond as the substrate, introduce nitrogen-vacancy NV color center defects into the diamond lattice through ion implantation technology; During the preparation process, precisely control the concentration and distribution of nitrogen atoms and nitrogen vacancies, and use electron spin resonance technology to initialize and read out the NV color centers; Specifically, adjust the depth distribution of nitrogen atoms by adjusting the ion energy and dose. High-temperature annealing above 1000 °C promotes the combination of vacancies and substitutional nitrogen atoms to form NV color centers; low-temperature annealing below 800 °C inhibits the migration of vacancies and retains single-vacancy defects; Adopt a low-temperature environment and microwave resonance excitation to change the electron spin state of the NV color centers and perform highly sensitive detection of gamma rays; Specifically, a low-noise microwave source is used to adjust the microwave frequency to the NV resonance point. A superconducting niobium (Nb) coplanar waveguide is fabricated on the diamond surface. Quasistatic noise is eliminated through multi-pulse refocusing. By monitoring the variation of fluorescence intensity with the microwave frequency, the change in the spin state is obtained, and the fluorescence intensity fluctuation caused by γ-rays corresponds to the detection signal. Based on the interaction between γ-rays and the diamond NV color center, the electron spin state of the NV color center changes. Radiation monitoring data, i.e., the intensity information of γ-rays, is obtained by measuring the fluorescence intensity fluctuation generated by the change in the spin state. Specifically, by measuring the change in fluorescence intensity, a high-sensitivity photodetector is used to collect the fluorescence signal and convert it into a fluorescence electrical signal. The collected electrical signal is filtered to remove noise interference. Specifically, based on the Gaussian function weight, the neighborhood sampling points are weighted and averaged to suppress the normally distributed noise. Through the Fourier transform signal processing algorithm, the time-domain signal is converted into a frequency-domain signal, and the characteristic parameter related to the γ-ray intensity, i.e., the fluorescence intensity, is extracted. Specifically, it is set that there is a linear relationship I = kD + b between the fluorescence intensity I and the γ-ray dose rate D, where k is the sensitivity coefficient, k corresponds to a γ-ray detection sensitivity of 0.01 μSv / h, i.e., the change in fluorescence intensity caused by the change in the unit dose rate, and b is the intensity offset, which is determined through a calibration experiment. Based on the measured fluorescence intensity I, the γ-ray dose rate D is calculated using the linear relationship I = kD + b between the fluorescence intensity I and the γ-ray dose rate D. It should be noted that compared with the traditional Geiger counter, the quantum sensing unit has higher sensitivity and can detect γ-rays at a lower dose rate. The response mechanism of the diamond NV color center to γ-rays is sensitive, and it can capture weak ray signal changes. Based on the microelectromechanical system (MEMS) technology, a nanoscale olfactory receptor simulation structure is constructed on the chip surface. A protein with a similar function to the mammalian olfactory receptor is synthesized through genetic engineering technology and immobilized on the chip surface. Microfluidic channels are fabricated using photolithography and etching processes to enable the radioactive gas to come into full contact with the olfactory receptor simulation structure. Based on the binding of the radioactive gas to the olfactory receptor simulation structure on the chip surface, the protein conformation changes, which in turn causes changes in the chip surface resistance and chip surface capacitance. The concentration information of the radioactive gas is obtained by measuring the changes in the chip surface resistance and chip surface capacitance. Specifically, the electrical signals of the chip surface resistance and the electrical signals of the chip surface capacitance measured are amplified and subjected to analog-to-digital conversion to convert the electrical signals of the chip surface resistance and capacitance into digital signals; Based on the neural network machine learning algorithm, pattern recognition is performed on the digital signals to establish a mapping relationship between the electrical signals and the concentration of radioactive gas; For the electrical signals of the chip surface resistance, an instrumentation amplifier is used to build a pre-amplification circuit; The instrumentation amplifier has the characteristics of high input impedance, low output impedance and high common-mode rejection ratio, suppressing the common-mode noise in the environment and converting the weak resistance change signal into a processable voltage signal; By configuring a programmable gain amplifier PGA, the amplification factor is adjusted according to the actual signal size to meet the requirements of different measurement scenarios; For the electrical signals of the chip surface resistance, a charge amplifier is used for signal conversion and amplification; It should be noted that the charge amplifier converts the charge change caused by the capacitance change into a voltage signal, and the output voltage of the charge amplifier is proportional to the input charge quantity and has nothing to do with the size of the capacitance itself, effectively avoiding the influence of the parasitic capacitance of the capacitance sensor; At the rear stage of the charge amplifier, a programmable gain amplifier is connected to further increase the signal amplitude; After the signal amplification is completed, the analog signal is converted into a digital signal for subsequent processing; The 24-bit Δ-Σ type analog-to-digital converter ADC accurately captures the weak electrical signal changes; It should be noted that to ensure the synchronous acquisition of the resistance and capacitance signals, the ADC is configured in a multi-channel synchronous sampling mode, and the two signals are sampled at a fixed frequency to convert the continuous analog signal into a discrete digital signal sequence; at the same time, an anti-aliasing filter is set at the front end of the ADC, and the filter cut-off frequency is designed according to the sampling theorem to prevent high-frequency signal aliasing and ensure the accuracy of the sampled signal; The signal is normalized to map the resistance and capacitance signal data to the [0,1] interval, eliminating the dimensional difference between different signal features and accelerating the training convergence speed of the neural network; The radiation data is framed, the continuous signal sequence is divided into frames of a fixed length, and the mean value, standard deviation and spectral energy of each frame are extracted to construct a feature vector; Gaussian noise is added to expand the number of dataset samples; Based on the feature vector, a prediction model is constructed using the multi-layer perceptron MLP architecture; Specifically, the number of input layer neurons is designed according to the dimension of the feature vector. N hidden layers are set in the middle. The number of neurons in each hidden layer is determined by tuning through the grid search method in experiments. Based on the ReLU function of the hidden layer, non-linear factors are introduced to enhance the expression ability of the model. A single neuron is set in the output layer to output the predicted value of the radioactive gas concentration, and the activation function uses a linear function to adapt to the regression task. The electrically-signal feature vectors collected in real time and preprocessed are input into the prediction model, and the result output by the prediction model is the predicted value of the radioactive gas concentration in the current environment. The change in the resistance of the chip is ΔR, and the radioactive gas concentration is C. The functional relationship f between the two is established through experiments, that is, C = f(ΔR). The fuselage of the drone is manufactured by combining carbon fiber composite materials with radiation-resistant coatings to reduce the damage of radiation to the fuselage materials. A lightweight quantum-biological hybrid sensor is carried on the drone. By optimizing the circuit design and structural layout of the sensor, its weight and volume are reduced. The drone swarm adopts a self-organizing network communication protocol to achieve real-time information interaction between nodes. During the flight of the drone, radiation data is collected in real time by the carried sensors. The radiation data includes but is not limited to: gamma-ray dose rate and radioactive gas concentration, and the radiation data is transmitted to the ground control center through a wireless communication module. The ground control center preprocesses the received radiation data, including data cleaning and format conversion. For radiation data cleaning, transmission error frames are eliminated through frame structure verification and CRC / LRC verification, missing values are processed by linear interpolation or forward filling, and sensor drift is calibrated based on standard source calibration and temperature compensation. For format conversion, the original radiation data is converted into standard physical units according to the detector energy response curve and calibration formula, and the time stamp is unified to the millisecond level. The swarm intelligence algorithm is used to analyze the radiation data to determine the radiation hot spot area. It should be noted that in the swarm intelligence algorithm, each drone is used as a particle, its position represents the coordinates in the monitoring area, and the speed represents the flight direction and speed. By continuously updating the position and speed of the particles, the drone swarm can avoid obstacles autonomously and track the radiation hot spots. By comparing with the traditional single-point deployment or simple drone formation deployment methods, it is found that due to the lack of an effective cooperation mechanism in the traditional methods, more time and resources are required when monitoring the same area, while the radiation-resistant drone swarm realizes efficient cooperative operations through the swarm intelligence algorithm. Inference calculation module: Based on the obtained radiation data, analyze the spatio-temporal correlation, optimize the algorithm to accelerate the result output, adjust the computational complexity through calculating dynamic weight distillation, and correct data transmission in real time by strengthening the hardware; Introduce spatio-temporal sparsification decision in the attention mechanism. By analyzing the spatio-temporal correlation of radiation data, determine the key spatio-temporal positions, perform attention calculation on the key spatio-temporal positions, ignore other redundant positions, and adopt pruning technology to remove unimportant connections and parameters; It should be noted that the key spatio-temporal positions are the spatio-temporal coordinate points with high contribution to the representation of radiation information; Organize the collected radiation data according to the time series and spatial positions to form a spatio-temporal data matrix; perform normalization processing on the spatio-temporal data matrix; The input radiation data is processed through multiple layers of attention calculation and feed-forward neural network to construct a Transformer model to obtain the output result. In the spatio-temporal sparse Transformer, the input radiation data is X ∈ R N×T×D , where N is the number of spatial positions, T is the number of time steps, and D is the data feature dimension; through spatio-temporal sparsification decision, perform attention calculation on K key spatio-temporal positions; During the operation of the AI inference engine, monitor the environmental radiation intensity in real time; According to the different radiation intensities, dynamically adjust the weight parameters of the prediction model; If the radiation intensity is less than the preset radiation intensity threshold, use a model with higher complexity for accurate prediction; if the radiation intensity is greater than or equal to the preset radiation intensity threshold, it is determined as a key pollution event, and transfer the knowledge of the complex model to the simple model through weight distillation technology to adjust the model complexity; Specifically, take the environmental radiation intensity as an additional input feature and input it into the model together with the radiation data. During the model training stage, learn the optimal weight adjustment decision under different radiation intensities through the reinforcement learning algorithm; Based on the obtained environmental radiation intensity R and the weight parameter W of the model, establish a functional relationship , where g(R) is the weight relationship function, is the preset weight, is the attenuation factor, with a value of 0.8, is the extreme maximum radiation threshold, is the high radiation threshold, and dynamically adjust the weight parameter W according to the radiation intensity R; It should be noted that under extreme conditions such as high radiation intensity, through dynamic weight distillation technology, reduce the computational complexity of the model, so as to reduce the required computational resources and ensure the rapid operation of the model in a harsh environment; Silicon carbide is used as chip packaging material, which has excellent radiation resistance and high temperature stability; In chip design, error correction coding circuits are integrated to detect and correct errors that occur during the transmission and storage of radiation data in real time; advanced semiconductor manufacturing processes are used to improve chip integration and performance; Specifically, the chip receives radiation data from the sensor and data processing and calculation instructions of the edge computing algorithm, and quickly processes and calculates the radiation data; during the data processing process, the error correction coding circuit is used to protect the radiation data to ensure the accuracy of the data; Decision management module: construct a four-dimensional causal graph based on the radiation data after the correction is completed, conduct intervention simulation based on the four-dimensional causal graph, evaluate the long-term effect of the decision and screen the optimal decision, and construct a generative adversarial network based on historical repair case data for adversarial training; Based on the collected radiation monitoring data, geological exploration data, meteorological observation data and biological sample data, the radiation data is analyzed using a constraint-based causal discovery algorithm to explore the causal relationship between radiation data; The radiation monitoring data, geological exploration data, meteorological observation data and biological sample data are used as nodes and the causal relationships as edges to construct a four-dimensional causal graph; Specifically, the collected radiation data are cleaned, fused, and normalized to remove noise and outliers; Convert different types of radiation data into a unified format for cause-effect analysis; Determine the causal structure by analyzing the conditional independence relationship between variables; For example, for the relationship between radiation and groundwater infiltration, the causal relationship between the two can be determined by analyzing the impact of groundwater infiltration on radiation diffusion under different geological and meteorological conditions; key transmission paths can be identified, such as the impact of groundwater infiltration on the diffusion of cesium-137, to accurately understand the mechanism of radiation transmission and provide a basis for subsequent governance decisions; Based on the constructed four-dimensional causal graph, counterfactual reasoning technology is used to conduct intervention simulation; for different governance schemes, corresponding intervention nodes are set in the causal graph to simulate the radiation propagation after the implementation of different schemes; Specifically, the Monte Carlo simulation is used to simulate different initial conditions and environmental changes multiple times to evaluate the long-term effects of the governance plan; During the simulation process, the parameters in the cause-effect diagram are adjusted according to the actual situation, and the simulation results are statistically analyzed to calculate the mean and variance of the reduction in radiation concentration and the reduction in the impact range; It should be noted that the deviation of the long-term effects of different governance solutions is obtained by calculating the mean absolute error through comparing the simulation results with the actual observed data; through counterfactual intervention simulation, the effects of the governance solutions are evaluated before the actual implementation of the governance solutions to discover potential problems in advance and select the optimal governance solution; Collect historical restoration case data, which includes but is not limited to: restoration solutions, implementation processes, and restoration effects; Build a generative adversarial network through a generator and a discriminator for adversarial training. The generator generates candidate restoration solutions based on real-time radiation data and historical case data, and the discriminator differentiates between the generated solutions and the real effective solutions; through adversarial training, the generator continuously optimizes the generated solutions; Extract features and encode the historical restoration case data to convert the historical restoration case data into a format suitable for network input; preprocess the real-time radiation data, extract key features, and combine them with the historical case data as the input of the generator; During the adversarial training process, the goal of the generator is to generate restoration solutions that bypass the discriminator, and the goal of the discriminator is to accurately distinguish between real solutions and generated solutions; Fusion collaboration module: Build a cross-institutional federated learning framework, perform hash operations on radiation data, and create NFTs on the blockchain through smart contracts for querying, verification, and trading; Establish a federated learning framework among different institutions. Each institution retains the original data locally and only uploads the model parameters or gradient information to the central server; During the uploading process, use differential privacy technology to add noise to the radiation data to ensure the privacy of the radiation data. The central server receives the information uploaded by each institution, aggregates and updates it, and then distributes the updated model to each institution; Obtain the model parameter θi of institution i, and the parameter after adding noise is θi′ = θi + ϵ, where ϵ is the noise that satisfies the Laplace distribution; Perform hash operations on the radiation data of key pollution events to generate a unique digital fingerprint; Obtain relevant metadata, which includes but is not limited to: data collection time, location, and collection equipment; Package the digital fingerprint together with the relevant metadata and create a non-fungible token NFT on the blockchain through a smart contract; each participating party queries, verifies, and trades the NFT through the blockchain platform; It should be noted that the uniqueness and immutability of radiation data are ensured through hash operations. Any minor change in radiation data will cause a huge change in the hash value. The radiation data of key pollution events are stored on the blockchain to support traceability and liability determination. In case of pollution disputes, the source and change process of the radiation data can be traced by querying the NFT information on the blockchain, and the responsibilities of all parties can be clarified.

[0018] The above has described an embodiment of the present invention in detail, but the content described is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention; the above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by software simulation of a large amount of collected radiation data to get a formula that is closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation and historical experience and can be adjusted according to the actual situation; the above is only a preferred embodiment of the present invention and is not used to limit the present invention. Any equal changes and improvements made according to the scope of the present invention application should still fall within the scope covered by the patent of the present invention.

Claims

1. An artificial intelligence-based nuclear radiation detection management system, characterized in that, It includes the following steps: Dynamic sensing module: Detect nuclear radiation through quantum sensing to obtain radiation data, capture the fluorescence intensity of NV color centers based on photodetectors, analyze the changes to calculate the gamma-ray dose rate, construct a nanoscale olfactory receptor simulation structure, detect radioactive gases through biomimetic detection, and use a swarm of drones for collaborative monitoring; Inference calculation module: Analyze the spatio-temporal correlation based on the obtained radiation data, optimize the algorithm to accelerate the result output, adjust the calculation complexity through calculating dynamic weight distillation, and correct data transmission in real time by strengthening the hardware; Decision management module: Construct a four-dimensional causal graph based on the radiation data after correction, perform intervention simulation based on the four-dimensional causal graph, evaluate the long-term effect of the decision and screen the optimal decision, and construct a generative adversarial network for adversarial training based on historical repair case data; Fusion and collaboration module: Build a cross-institutional federated learning framework, perform hash operation on the radiation data, and create NFTs on the blockchain through smart contracts for querying, verification, and trading.

2. An artificial intelligence-based nuclear radiation detection management system according to claim 1, characterized in that, The method for detecting nuclear radiation through quantum sensing is as follows: Using diamond as the substrate, introduce nitrogen-vacancy NV color center defects into the diamond lattice through ion implantation technology; Precisely control the concentration and distribution of nitrogen atoms and nitrogen-vacancy NV color center defects, and use electron spin resonance technology to initialize and read out the NV color centers; Adopt a low-temperature environment and microwave resonance excitation to change the electron spin state of the NV color centers and perform highly sensitive detection of gamma rays; Based on the interaction between gamma rays and diamond NV color centers, the electron spin state of the NV color centers changes; Obtain the intensity information of gamma rays by measuring the change in the spin state; By measuring the change in fluorescence intensity, collect fluorescence signals based on highly sensitive photodetectors and convert the fluorescence signals into fluorescence electrical signals.

3. An artificial intelligence-based nuclear radiation detection management system according to claim 2, characterized in that, The method for obtaining the gamma-ray dose rate is as follows: Perform filtering processing on the collected fluorescence electrical signals; Through the Fourier transform signal processing algorithm, convert the time-domain signal into a frequency-domain signal, and extract the characteristic parameters related to the gamma-ray intensity, that is, the fluorescence intensity; Set that there is a linear relationship between the fluorescence intensity I and the gamma-ray dose rate D; Based on the measured fluorescence intensity I, calculate the gamma-ray dose rate D using the linear relationship between the fluorescence intensity I and the gamma-ray dose rate D.

4. An artificial intelligence-based nuclear radiation detection management system according to claim 1, characterized in that, The method for biomimetic detection of radioactive gases is as follows: Synthesize proteins with similar functions to mammalian olfactory receptors through genetic engineering technology and fix the proteins on the chip surface; Use photolithography and etching processes to manufacture microfluidic channels to enable radioactive gases to fully contact the nanoscale olfactory receptor simulation structure; Based on the binding of radioactive gases to the nanoscale olfactory receptor simulation structure on the chip surface, which causes changes in the protein conformation, and further leads to changes in the chip surface resistance and chip surface capacitance; Obtain the concentration information of radioactive gases by measuring the changes in the chip surface resistance and chip surface capacitance; Precisely capture weak electrical signal changes through a 24-bit Δ-Σ type analog-to-digital converter ADC; Perform normalization processing on the electrical signals and map the resistance and capacitance signal data to the [0,1] interval; Perform frame processing on the radiation data, divide the continuous signal sequence into frames of a fixed length, extract the mean, standard deviation, and spectral energy of each frame, and construct a feature vector; Based on the feature vector, construct a prediction model using the multi-layer perceptron (MLP) architecture; Input the feature vector of the electrical signal collected in real time and preprocessed into the prediction model, and the result output by the prediction model is the predicted value of the concentration of radioactive gas in the current environment.

5. An artificial intelligence-based nuclear radiation detection management system according to claim 1, characterized in that, The method of using an unmanned aerial vehicle (UAV) cluster for collaborative monitoring is as follows: Install a lightweight quantum-biological hybrid sensor on the UAV. By optimizing the circuit design and structural layout of the sensor, reduce its weight and volume. The UAV cluster adopts a self-organizing network communication protocol; During the flight of the UAV, collect radiation data in real time through the installed sensor, and transmit the radiation data to the ground control center through a wireless communication module; The ground control center performs data cleaning and format conversion on the received radiation data; Use a swarm intelligence algorithm to analyze the radiation data and determine the radiation hot spot area.

6. The nuclear radiation detection and management system based on artificial intelligence according to claim 1, characterized in that, The method of accelerating the result output is as follows: Introduce spatio-temporal sparsification decision-making in the attention mechanism. By analyzing the spatio-temporal correlation of the radiation data, determine the key spatio-temporal positions, perform attention calculation on the key spatio-temporal positions, ignore other redundant positions, and use pruning technology to remove unimportant connections and parameters; Organize the collected radiation data according to the time series and spatial positions to form a spatio-temporal data matrix; perform normalization processing on the spatio-temporal data matrix; The input radiation data is processed through multiple layers of attention calculations and a feed-forward neural network to construct a Transformer model to obtain the output result. In the spatio-temporal sparsified Transformer, the input radiation data is X ∈ R N×T×D , where N is the number of spatial positions, T is the number of time steps, and D is the data feature dimension; Through spatio-temporal sparsification decision-making, perform attention calculation on K key spatio-temporal positions.

7. An artificial intelligence-based nuclear radiation detection management system according to claim 4, characterized in that, The method of adjusting the computational complexity is as follows: During the operation of the AI inference engine, monitor the radiation intensity of the environment in real time; According to the different radiation intensities, dynamically adjust the weight parameters of the prediction model; If the radiation intensity is less than the preset radiation intensity threshold, use a model with a higher complexity for accurate prediction; If the radiation intensity is greater than or equal to the preset radiation intensity threshold, it is determined as a key pollution event, and transfer the knowledge of the complex model to the simple model through the weight distillation technology to reduce the model complexity; The environmental radiation intensity is used as an additional input feature and input into the model together with the radiation data. During the model training stage, learn the optimal weight adjustment decision under different radiation intensities through a reinforcement learning algorithm; Based on the obtained environmental radiation intensity of R and the weight parameters of the model of W, establish a functional relationship W = g(R), where g(R) is the weight relationship function, and dynamically adjust the weight parameter W according to the radiation intensity R.

8. An artificial intelligence-based nuclear radiation detection management system according to claim 1, characterized in that, The method of performing intervention simulation based on a four-dimensional causal graph is as follows: Based on the constructed four-dimensional causal graph, use counterfactual reasoning technology to perform intervention simulation; for different treatment plans, set corresponding intervention nodes in the four-dimensional causal graph to simulate the radiation propagation situation after implementing different plans; Based on the Monte Carlo simulation, simulate different initial conditions and environmental changes multiple times to evaluate the long-term effects of the treatment plans; During the simulation process, adjust the parameters in the four-dimensional causal graph according to the actual situation, perform statistical analysis on the simulation results, and calculate the mean and variance of the reduction in radiation concentration and the reduction in the affected range.

9. The nuclear radiation detection and management system based on artificial intelligence according to claim 1, characterized in that, The method of performing adversarial training is as follows: Collect historical repair case data, construct a generative adversarial network through a generator and a discriminator for adversarial training. The generator generates candidate repair solutions based on real-time radiation data and historical case data, and the discriminator distinguishes between the generated solutions and the real effective solutions; through adversarial training, the generator continuously optimizes the generated solutions. Extract and encode the features of historical repair case data, and convert the historical repair case data into a format suitable for network input; preprocess the real-time data, extract key features, and combine them with the historical case data as the input of the generator.

10. The nuclear radiation detection and management system based on artificial intelligence according to claim 1, characterized in that The method for creating NFT is as follows: During the upload process, differential privacy technology is used to add noise to the radiation data. The central server receives the information uploaded by each institution, aggregates and updates it, and then distributes the updated federated learning framework to each institution. Perform a hash operation on the radiation data of key pollution events to generate a unique digital fingerprint. Obtain relevant metadata, package the digital fingerprint together with the relevant metadata, and create a non-fungible token NFT on the blockchain through a smart contract; each participating party queries, verifies, and trades the NFT through the blockchain platform.

Images