Multi-channel nuclear electronics data acquisition system based on artificial intelligence

Abstract

The invention discloses a multichannel nuclear electronics data acquisition system based on artificial intelligence, and relates to the technical field, and the system comprises the steps: carrying out the front-end preprocessing of nuclear radiation or X-ray signals inputted by each detection channel; a deep neural network is utilized to extract features of the signals after front-end preprocessing, and the features are integrated to form a multi-dimensional state vector describing a channel state; inputting the multi-dimensional state vector into a deep reinforcement learning module, and determining a sampling parameter of each channel; establishing a correlation model of the dependency relationship among the channel signals; performing cross compensation on the channel in the abnormal state, and calculating and applying a correction factor through a prediction method; based on real-time feedback data, a self-evolution multi-task optimization algorithm is utilized to carry out joint adjustment on each sampling parameter, and a closed-loop data acquisition control unit is formed. According to the system, automatic joint optimization of multi-task and multi-target parameters can be realized, and the problem that real-time performance, stability and multi-target optimization cannot be considered at the same time in a traditional method is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of nuclear radiation or X-ray radiation measurement, and particularly to a multi-channel nuclear electronics data acquisition system based on artificial intelligence. Background Art

[0002] Existing nuclear electronics data acquisition technologies mainly rely on traditional fixed parameters and manual adjustment methods to detect and collect nuclear radiation or X-ray signals. In the face of actual application scenarios such as multi-channel detection, environmental changes, and equipment aging, these methods have problems such as difficulty in flexibly adapting sampling parameters, uneven signal processing effects, and low automation. At the same time, in recent years, artificial intelligence technologies, especially the application of deep learning and reinforcement learning in complex systems, have become increasingly mature, providing a new theoretical and practical basis for accurate data acquisition and dynamic optimization in multi-task and big data environments. Therefore, using advanced intelligent algorithms to achieve adaptive and high-precision data acquisition and processing for multi-channel nuclear electronics systems has become an urgent research and development direction. Summary of the Invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] Therefore, in view of the problems of fixed sampling strategies, inflexible signal processing processes, and difficult abnormal data processing existing in the existing multi-channel nuclear electronics data acquisition system, the present invention proposes a multi-channel nuclear electronics data acquisition method based on artificial intelligence. By introducing a deep neural network to extract features from preprocessed signals, using deep reinforcement learning to achieve adaptive allocation of dynamic sampling parameters, constructing a cross-channel correlation model for cross-compensation of abnormal data, and adopting a self-evolving multi-task optimization algorithm to jointly predict and globally search for multiple key parameters such as sampling rate, integration time, signal amplification factor, energy consumption, and hardware load, the automatic joint optimization of multi-task and multi-objective parameters is realized, and the problem that traditional methods cannot simultaneously take into account real-time performance, stability, and multi-objective optimization is solved.

[0005] To solve the above technical problems, the present invention provides the following technical solution, a multi-channel nuclear electronics data acquisition system based on artificial intelligence, including: Perform front-end preprocessing on nuclear radiation or X-ray signals input to each detection channel, including noise suppression and baseline correction; Use a deep neural network to extract features such as signal-to-noise ratio, signal peak, and noise mean from the front-end preprocessed signals, and integrate them into a multi-dimensional state vector describing the channel state; Input the multi-dimensional state vector into a deep reinforcement learning module, and dynamically determine the sampling parameters of each channel according to the multi-dimensional state vector; Establish a correlation model for the dependency relationship between signals of each channel to identify abnormal signals; Implement cross - compensation for channels in abnormal states, calculate and apply correction factors through prediction methods; Based on real - time feedback data, use the self - evolving multi - task optimization algorithm to jointly adjust each sampling parameter to form a closed - loop data acquisition control unit.

[0006] As a preferred solution of the multi - channel nuclear electronics data acquisition system based on artificial intelligence according to the present invention, wherein: the front - end pre - processing includes that the front - end signal processing circuit in each detection channel is integrated with a filtering and analog signal conditioning module for noise suppression and baseline drift correction of nuclear radiation or X - ray signals.

[0007] As a preferred solution of the multi - channel nuclear electronics data acquisition system based on artificial intelligence according to the present invention, wherein: the integration to form a multi - dimensional state vector describing the channel state includes that the feature extraction of the signal after front - end pre - processing is processed by a convolutional neural network CNN, the signal obtained after front - end pre - processing is reconstructed into a matrix form of a fixed size, a convolutional layer with a 3×3 convolutional kernel is applied to perform local convolutional operations on the input matrix, and a max - pooling layer with a window size of 2×2 is used to downsample the convolutional output. The second convolutional layer and its corresponding max - pooling layer are stacked in sequence to extract deeper - level features. At the same time, the feature maps generated by multi - layer convolution and pooling are flattened and mapped into a feature vector of a fixed dimension through a fully - connected layer.

[0008] As a preferred solution of the multi - channel nuclear electronics data acquisition system based on artificial intelligence according to the present invention, wherein: the dynamic determination of the sampling parameters of each channel according to the multi - dimensional state vector includes starting from the system input data, extracting information describing the multi - channel detection state through front - end signal processing and a deep neural network, and clearly defining the action space, where the multi - dimensional state vector is specifically as follows: ; wherein, represents the multi - dimensional state vector of the detection channel at time ; is the signal - to - noise ratio; is the signal peak; NoiseMean is the noise mean; is the environmental parameter.

[0009] The action definition is specifically as follows: ; wherein, represents the combination of sampling parameters selected by the system at time ; is the discrete sampling rate; is the discrete integration time; is the discrete signal amplification factor; The state is discretely mapped into discrete categories through the clustering method: ; wherein, represents the discretization result of; is the discrete category; is the total number of categories; The Deep Q-Network (DQN) is used to model the relationship between the state and the action. The policy is updated and the parameters are optimized through the improved Actor-Critic structure. At the same time, cross-channel regularization is introduced to ensure the smoothness and consistency of the system; The Q function is defined as follows: ; wherein, represents the expected value of the cumulative discounted reward obtained after taking an action under ; are the network weight parameters; is the discount factor, which characterizes the impact of future rewards on the current decision; is the immediate reward at the future time ; represents the time; Next, the function is used to update the network parameters, as follows: DQN update loss function: ; wherein, is the loss function, which is used to measure the error between the current network prediction and the target value; is the reward actually obtained at the current time; represents the parameters of the target network, which plays a role in stabilizing the update; represents the maximum expected reward that can be obtained in the next state ; represents a placeholder for "all possible actions", which is used to search for the optimal action value in the next state; The Actor-Critic structure is adopted to refine the policy through temporal difference update. At the same time, regularization is introduced to ensure the consistency of cross-channel decision-making: The temporal difference error is as follows: ; wherein, is the temporal difference error, which measures the deviation between the current state value estimation and the actual result; is the state value function estimated by the Critic network, and is the parameter of the state value function; is the state value function of the next state; The loss function of the Critic network is defined as follows: ; where, is the mean squared error of the Critic network, used to optimize the estimation of the state value function; The update gradient of the Actor network is as follows: ; where, is the gradient for updating the policy network parameters ; is the probability distribution of the policy network selecting actions in the state; is used to ensure the smoothness of actions at consecutive time moments; is used to constrain the consistency between sampling parameters among different detection channels; and are regularization coefficients, balancing the importance of temporal smoothness and cross-channel consistency respectively; ∇ is the gradient operator, i.e., the vector operation of taking the partial derivative of the network parameter vector; To enable the system to consider data quality, energy consumption, and cross-channel consistency simultaneously, a mechanism for the auxiliary network to dynamically generate various reward weights is introduced, as follows: Dynamic weight generation: ; where, represents the dynamic weight of the th reward, calculated by the auxiliary network , where are the parameters of the auxiliary network; the normalization condition ensures that the sum of all weights is 1; The comprehensive reward function is as follows: ; where, is the comprehensive reward obtained by the system taking actions in the state; is the reward corresponding to data quality; is the energy consumption penalty term; is the reward related to cross-channel consistency.

[0010] As a preferred solution of the multi-channel nuclear electronics data acquisition system based on artificial intelligence according to the present invention, wherein: the dynamic determination of the sampling parameters of each channel according to the multi-dimensional state vector includes, in order to make full use of the spatio-temporal dependence in the multi-channel detection information, using a self-attention mechanism to achieve effective fusion of cross-channel information, so as to provide more comprehensive input features for the decision-making module; The query, key, and value are mapped as: ; Wherein, represents the state vector of the th detection channel at time ; are the linear transformation matrices for generating the query , key , and value respectively; is the total number of detection channels; The attention weight is calculated as follows: ; Wherein, is the attention weight from channel to channel ; is the dimension of, used for normalization; is the summation index for traversing all channels; represents the key of each traversed channel l; The fused state vector is defined as follows: ; Wherein, is the fused state vector, representing the comprehensive state of the th channel after considering all channel information.

[0011] As a preferred solution of the multi-channel nuclear electronics data acquisition system based on artificial intelligence according to the present invention, wherein: the correlation model for establishing the dependence relationship between the signals of each channel includes performing statistical correlation analysis on the state vectors of each detection channel, and calculating the statistical correlation between the th channel and the th channel, and the formula is: ; Wherein, represents the state vector of the th channel at time ; represents the state vector of the th channel at time th channel State vector of the path; Denote the mean of the state vector ; Denote the mean of the state vector ; Denote the standard deviation of the state vector ; Denote the standard deviation of the state vector ; is a small positive constant to avoid division by zero; For adapting to the processing of graph neural networks, a non-linear mapping is used to convert : ; wherein, is the designed non-linear mapping function, composed of a fully-connected layer or a convolutional layer; is the dimension of the original state vector, is the dimension after node embedding; is the embedded representation after mapping for the th channel; After obtaining the node embedding , the unnormalized attention score is calculated through the graph attention mechanism, and the formula is: wherein, denotes concatenating the node embedding with ; is the parameter vector to be trained; denotes obtaining a scalar by linearly combining the concatenated vectors; LeakyReLU is the linear rectifier activation function with a negative slope, and the processed denotes the preliminary attention score for ; The unnormalized attention score is normalized through the Softmax operation to calculate the attention weight between each node : ; wherein, denotes the normalized attention weight from to ; in the numerator, maps the score to the positive real number space; Denotes the unnormalized attention score calculated between channel $i$ and channel $q$ in the graph attention mechanism; the denominator normalizes all attention scores for channels to ensure that ; To simultaneously utilize the statistical correlation and the attention scores obtained by the graph neural network, a weighted average method is adopted to generate the final cross-channel dependency matrix , and the formula is: ; where, denotes the comprehensive dependency between the th channel and the th channel; is the weight parameter, and its value range is ; is the value calculated from the statistical correlation; is the normalized attention weight calculated by the graph neural network.

[0012] As a preferred solution of the multi-channel nuclear electronics data acquisition system based on artificial intelligence described in the present invention, wherein: the calculation and application of the correction factor by the prediction method includes, using the cross-channel correlation matrix to perform traditional regression prediction on the abnormal channel to obtain the predicted value based on the signals of the remaining normal channels, and the formula is: ; where, denotes the predicted signal of the abnormal channel at time obtained by the statistical regression method; denotes the actual acquisition signal of the th detection channel at time ; is the total number of detection channels, and the abnormal channel is excluded during the summation; denotes the comprehensive dependency between the th channel and the th channel; Based on the traditional regression prediction, a neural network model is designed to input the signals of all channels except the abnormal channel to predict the non-linear residuals that cannot be captured by the regression model, and the formula is: ; where, denotes the residual value predicted by the neural network at time , corresponding to the non-linear prediction part of the abnormal channel; denotes the neural network that constitutes this prediction model, and its parameters are determined by It is shown that the model can adopt a multi-layer perceptron or a convolutional neural network structure; It represents the moment at which, the set of actual signals of all channels except the abnormal channel; The traditional regression prediction error and the residual of the neural network prediction are weighted and fused in proportion to calculate the comprehensive compensation factor of the abnormal channel, and its formula is: ; wherein, It represents the moment at which the comprehensive compensation factor of the abnormal channel; is the signal based on statistical regression prediction; It represents the moment at which the abnormal channel of the actual acquired signal; It represents the regression prediction error; is the residual obtained by neural network prediction; is a weight factor between 0 and 1, used to balance the contributions of the two parts; The comprehensive compensation factor is applied to the original signal of the abnormal channel to complete data correction, and the formula is: ; wherein, It represents the moment at which the corrected abnormal channel of the signal.

[0013] As a preferred solution of the multi-channel nuclear electronics data acquisition system based on artificial intelligence described in the present invention, wherein: the self-evolving multi-task optimization algorithm includes input data , and the multi-task deep neural network outputs the predictions of multiple target parameters , constructs a joint loss function to evaluate the network output error, and performs a global optimal search on the network parameters ; uses the particle swarm optimization algorithm to update the speed and parameters to form a self-evolving parameter adjustment mechanism; The parameter update formula directly uses the updated velocity vector to correct the network parameters to form a new parameter set ; this process is repeated in each iteration until the joint loss function meets the predetermined convergence condition.

[0014] A computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the multi-channel nuclear electronics data acquisition system based on artificial intelligence.

[0015] A computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, steps of a multi-channel nuclear electronics data acquisition system based on artificial intelligence are implemented.

[0016] Advantages of the present invention: Through the technical solution proposed by the present invention, the system can achieve automation, intelligence and closed-loop dynamic optimization in multi-channel nuclear electronics data acquisition. Each detection channel can adjust sampling parameters and data processing strategies in real time according to the actual signal state and environmental changes. At the same time, abnormal data is compensated and corrected through a cross-channel correlation model, and a self-evolving multi-task optimization algorithm is used to jointly adjust multiple key target parameters to ensure that all technical indicators reach the best ratio, providing effective technical support for the precision measurement, automatic control and long-term operation of nuclear radiation and X-ray data acquisition systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 It is a schematic flowchart of a multi-channel nuclear electronics data acquisition system based on artificial intelligence provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] To make the above objects, features and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are some, but not all, embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] Many specific details are set forth in the following description in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0021] Second, the "one embodiment" or "embodiment" referred to herein means a specific feature, structure or characteristic that may be included in at least one implementation manner of the present invention. The "in one embodiment" that appears in different places in this specification does not all refer to the same embodiment, nor is it an embodiment that is separate or selectively mutually exclusive with other embodiments.

[0022] The present invention will be described in detail in conjunction with schematic diagrams. When describing the embodiments of the present invention in detail, for the convenience of explanation, the cross-sectional views showing the device structure will be enlarged locally in a non-general proportion, and the schematic diagrams are only examples and should not limit the scope of protection of the present invention here. In addition, in actual production, three-dimensional spatial dimensions of length, width and depth should be included.

[0023] At the same time, in the description of the present invention, it should be noted that the orientation or positional relationship indicated by terms such as "upper, lower, inner and outer" is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0024] Unless otherwise clearly defined and limited in the present invention, the terms "installation, connection, and connection" should be understood in a broad sense. For example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, and can also be indirectly connected through an intermediate medium, or can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0025] Example 1, referring to Figure 1 , which is the first embodiment of the present invention. This embodiment provides a multi-channel nuclear electronics data acquisition system based on artificial intelligence, including: S1: Perform front-end preprocessing on the nuclear radiation or X-ray signals input to each detection channel, including noise suppression and baseline correction.

[0026] The front-end preprocessing includes that the front-end signal processing circuit in each detection channel integrates a filtering and analog signal conditioning module for noise suppression and baseline drift correction of the nuclear radiation or X-ray signals.

[0027] S2: Use a deep neural network to extract features such as signal-to-noise ratio, signal peak value, and noise mean value from the signals after front-end preprocessing, and integrate them to form a multi-dimensional state vector describing the channel state.

[0028] The integrated multi-dimensional state vector that describes the channel state includes performing feature extraction on the pre-processed signals at the front end using a convolutional neural network (CNN). The signals obtained after front-end pre-processing are reconstructed into a matrix form of a fixed size. A convolutional layer with a 3×3 convolutional kernel is applied to perform local convolution operations on the input matrix, and a max-pooling layer with a window size of 2×2 is used to downsample the convolutional output. The second convolutional layer and its corresponding max-pooling layer are stacked in sequence to extract deeper features. At the same time, the feature maps generated by multiple layers of convolution and pooling are flattened and mapped into a feature vector of a fixed dimension through a fully connected layer. This feature vector constitutes the main part of the state vector for use by subsequent modules.

[0029] S3: Input the multi-dimensional state vector into the deep reinforcement learning module, and dynamically determine the sampling parameters of each channel according to the multi-dimensional state vector.

[0030] The dynamic determination of the sampling parameters of each channel according to the multi-dimensional state vector includes starting from the system input data, extracting information describing the multi-channel detection state through front-end signal processing and a deep neural network, and clearly defining the action space. The multi-dimensional state vector is specifically as follows: ; where, represents the multi-dimensional state vector of the detection channel at time ; is the signal-to-noise ratio; is the signal peak; NoiseMean is the noise mean; is the environmental parameter (such as temperature, voltage, etc.); The multi-dimensional state vector converts the original pre-processed signals into a numerical description that can be input to the subsequent decision-making module, providing basic information for reinforcement learning decisions.

[0031] The action definition is specifically as follows: ; where, represents the combination of sampling parameters selected by the system at time ; is the discrete sampling rate; is the discrete integration time; is the discrete signal amplification factor; To reduce the complexity of the continuous state space, the state is discretely mapped into discrete categories through a clustering method here: ; where, represents the discretization result of; is the discrete category; is the total number of categories; The relationship between states and actions is modeled using a Deep Q-Network (DQN). The policy is updated and parameters are optimized through an improved Actor-Critic structure. Meanwhile, cross-channel regularization is introduced to ensure the smoothness and consistency of the system. The Q function is defined as follows: ; where represents the expected value of the cumulative discounted reward obtained after taking an action under ; are the network weight parameters; is the discount factor, which characterizes the impact of future rewards on the current decision; is the immediate reward at the future time ; represents the time step; By defining the function, a value mapping between states and actions is established, and thus a basis for action selection can be provided. Next, the function is used to update the network parameters as follows: DQN update loss function: ; where is the loss function, which is used to measure the error between the current network prediction and the target value; is the reward actually obtained at the current time step; represents the parameters of the target network, which play a role in stabilizing the update; represents the maximum expected reward that can be obtained in the next state ; represents a placeholder for "all possible actions", which is used to search for the optimal action value in the next state; The DQN part provides state-based value estimation for the system. It adopts an Actor-Critic structure, refines the policy through temporal difference updates, and introduces regularization to ensure the consistency of cross-channel decisions: The temporal difference error is as follows: ; where is the temporal difference error, which measures the deviation between the current state value estimation and the actual result; is the state value function, estimated by the Critic network, are the state value function parameters; is the state value function of the next state; Define the loss function of the Critic network as follows: ; Among them, is the mean square error of the Critic network, used to optimize the estimation of the state value function; The update gradient of the Actor network (including cross-channel regularization) is specifically as follows: ; Among them, is the gradient for updating the policy network parameters ; is the probability distribution of the policy network selecting actions in the state; is used to ensure the smoothness of actions at consecutive moments; is used to constrain the consistency between sampling parameters among different detection channels; and are the coefficients of regularization, respectively balancing the importance of temporal smoothing and cross-channel consistency; ∇ is the gradient operator, that is, the vector operation of taking the partial derivative of the network parameter vector; The Actor-Critic structure completes the iterative update of the policy through the above formula, and after introducing regularization constraints, it can further coordinate the decisions among channels.

[0032] To enable the system to consider data quality, energy consumption, and cross-channel consistency simultaneously, a mechanism for the auxiliary network to dynamically generate various reward weights is introduced, specifically as follows: Dynamic weight generation: ; Among them, represents the dynamic weight of the th reward, calculated by the auxiliary network , where are the parameters of the auxiliary network; the normalization condition ensures that the sum of all weights is 1; The comprehensive reward function is specifically as follows: ; Among them, is the comprehensive reward obtained by the system taking actions in the state; is the reward corresponding to data quality; is the energy consumption penalty term; is the reward related to cross-channel consistency.

[0033] Through the dynamic weight generation mechanism, the reward function can be autonomously adjusted according to the actual state and actions, thereby guiding the update of reinforcement learning more flexibly.

[0034] To make full use of the spatio-temporal dependence in multi-channel detection information, a self-attention mechanism is adopted to achieve effective fusion of cross-channel information, thus providing more comprehensive input features for the decision-making module. The query, key, and value mappings are as follows: ; Among them, represents the state vector of the th detection channel at time ; are the linear transformation matrices used to generate the query , key , and value respectively; is the total number of detection channels; These mappings help capture the similarities and interaction information between different channels.

[0035] The attention weights are calculated as follows: ; Among them, is the attention weight from channel to channel ; is the dimension of for normalization; is the summation index used to traverse all channels; The fused state vector is defined as follows: ; Among them, is the fused state vector, representing the comprehensive state of the th channel after considering all channel information.

[0036] S4: Establish a correlation model for the dependencies between the signals of each channel to identify abnormal signals.

[0037] The establishment of the correlation model for the dependencies between the signals of each channel includes performing statistical correlation analysis on the state vectors of each detection channel and calculating the statistical correlation between the th channel and the th channel, with the formula: ; Among them, represents the time The state vector of the th channel; represents the moment The state vector of the th channel; represents the mean of the state vector ; represents the mean of the state vector ; represents the standard deviation of the state vector ; represents the standard deviation of the state vector ; is a small positive constant used to avoid a zero denominator; This formula uses the idea of the Pearson correlation coefficient to represent the similarity degree of the two-channel state vectors with a numerical value , providing basic statistical information for the subsequent introduction of the attention weights extracted by the graph neural network.

[0038] To adapt to the processing of the graph neural network, a non-linear mapping is used to transform into a low-dimensional embedding vector : ; Among them, is the designed non-linear mapping function, which is composed of a fully connected layer or a convolutional layer; is the dimension of the original state vector, is the dimension after node embedding; is the embedded representation of the th channel after mapping; The node embedding after mapping will be used as the input for calculating the similarity between nodes in the graph attention mechanism, laying a foundation for further constructing the attention score of the graph neural network.

[0039] After obtaining the node embedding , the unnormalized attention score is calculated through the graph attention mechanism, and the formula is: ; Among them, represents concatenating the node embedding with ; is the parameter vector to be trained; represents obtaining a scalar by linearly combining the concatenated vectors; LeakyReLU is the linear rectifier activation function with a negative slope, and the processed represents The preliminary attention score; Normalize the unnormalized attention scores through the Softmax operation and calculate the attention weights between each node : Among them, represents the normalized attention weight from to ; the in the numerator maps the score to the positive real number space; represents the unnormalized attention score calculated between channel i and channel q in the graph attention mechanism; the denominator normalizes all the attention scores under channel to ensure ; The attention weights obtained through normalization reflect the correlation information calculated between each channel through the graph neural network. In the next step, it will be combined with the statistical correlation to generate the final cross-channel dependency matrix.

[0040] In order to utilize both the statistical correlation and the attention scores obtained by the graph neural network, a weighted average method is adopted to generate the final cross-channel dependency matrix , and the formula is: ; Among them, represents the comprehensive dependency relationship between the th channel and the th channel; is the weight parameter, and its value range is ; is the value calculated from the statistical correlation; is the normalized attention weight calculated from the graph neural network.

[0041] The matrix constituted by is the cross-channel dependency matrix used in the subsequent system. The cross-channel dependency matrix provides an information expression of the mutual dependency between multiple channels and is the final output of the entire cross-channel correlation modeling process.

[0042] S5: Implement cross-compensation for channels in the abnormal state, and calculate and apply the correction factor through the prediction method.

[0043] The calculation and application of the correction factor through the prediction method include, using the cross-channel correlation matrix to perform traditional regression prediction on the abnormal channels to obtain the predicted values based on the signals of the remaining normal channels. The formula is: ; Among them, represents the moment at which the predicted signal obtained by the statistical regression method for the abnormal channel is; represents the moment at which the actual acquisition signal of the th detection channel is; is the total number of detection channels, where the summation excludes the abnormal channels; represents the th channel and the th channel between the comprehensive dependence relationship; The abnormal channel is preliminarily predicted by using the signals of other channels through the linear regression method, providing a benchmark prediction component for the subsequent introduction of the neural network for non-linear compensation.

[0044] On the basis of traditional regression prediction, a neural network model is designed to input the signals of all channels except the abnormal channel, and the non-linear residuals that cannot be captured by the regression model are predicted. The formula is: ; Among them, represents the moment at which the residual value predicted by the neural network is obtained, corresponding to the non-linear prediction part of the abnormal channel; represents the neural network that constitutes this prediction model, and its parameters are represented by , and the model can adopt a multi-layer perceptron or a convolutional neural network structure; represents the moment at which the actual signal set of all channels except the abnormal channel is; The traditional regression prediction error and the residuals predicted by the neural network are weighted and fused in proportion to calculate the comprehensive compensation factor of the abnormal channel. The formula is: ; Among them, represents the moment at which the comprehensive compensation factor of the abnormal channel is; is the signal based on statistical regression prediction; represents the moment at which the actual acquisition signal of the abnormal channel is; represents the regression prediction error (i.e., the deviation between the predicted signal and the actual signal); is the residual predicted by the neural network; is a weight factor between 0 and 1, used to balance the contributions of the two parts; its value can be determined by the validation set according to the actual situation or adjusted by an adaptive strategy.

[0045] By synthesizing the linear prediction error and the non - linear residual, a comprehensive compensation factor is formed, which reflects the relative role of each method in abnormal data correction and provides a direct basis for the final data correction.

[0046] The comprehensive compensation factor is applied to the original signal of the abnormal channel to complete data correction. The formula is: ; where represents the time and is the signal of the corrected abnormal channel at time

[0047] S6: Based on the real - time feedback data, the self - evolving multi - task optimization algorithm is used to jointly adjust each sampling parameter to form a closed - loop data acquisition control unit.

[0048] The self - evolving multi - task optimization algorithm includes: input data , and the multi - task deep neural network outputs the predictions of multiple target parameters . A joint loss function is constructed to evaluate the network output error, and the global optimal search is performed on the network parameters . The particle swarm optimization algorithm is used to update the velocity and parameters to form a self - evolving parameter adjustment mechanism; The parameter update formula directly uses the updated velocity vector to correct the network parameters to form a new parameter set ; This process is repeated in each iteration until the joint loss function meets the predetermined convergence condition.

[0049] The whole process is completed in sequence, so as to realize the joint prediction and optimization of multiple targets such as sampling rate, integration time, signal amplification factor, energy consumption and hardware load. The innovation of this model lies in combining multi - task deep prediction with the global search algorithm PSO, providing an adaptive and globally optimal self - evolving multi - objective optimization scheme.

[0050] Embodiment 2 The second embodiment of the present invention is different from the previous embodiment in that: When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0051] This application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0052] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in this computer-readable memory generate a manufactured product including an instruction device, and this instruction device implements the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0053] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0054] Although the preferred embodiments of the present application have been described, additional changes and modifications can be made to these embodiments by those skilled in the art once they learn of the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications that fall within the scope of the present application.

[0055] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these modifications and variations.

Claims

1. Multi-channel nuclear electronics data acquisition system based on artificial intelligence, characterized by: include, Perform front-end preprocessing on the nuclear radiation or X-ray signals input from each detection channel, including noise suppression and baseline correction; Use deep neural networks to extract signal-to-noise ratio, signal peak, and noise mean features from the front-end preprocessed signal, and integrate them to form a multi-dimensional state vector that describes the channel state; The multidimensional state vector is input into the deep reinforcement learning module, and the sampling parameters of each channel are dynamically determined according to the multidimensional state vector; Establish a correlation model of the dependency between signals in each channel to identify abnormal signals; Implement cross compensation for channels in abnormal state, calculate and apply correction factors through predictive methods; Based on real-time feedback data, the self-evolutionary multi-task optimization algorithm is used to jointly adjust the sampling parameters to form a closed-loop data acquisition control unit.

2. The multi-channel nuclear electronics data acquisition system based on artificial intelligence as claimed in claim 1, characterized in that: The front-end preprocessing includes that the front-end signal processing circuit in each detection channel is integrated with a filtering and analog signal conditioning module for performing noise suppression and baseline drift correction on nuclear radiation or X-ray signals.

3. The multi-channel nuclear electronics data acquisition system based on artificial intelligence as claimed in claim 2, characterized in that: The integration to form a multi-dimensional state vector describing the channel state includes: using a convolutional neural network (CNN) to extract features from the signal after front-end preprocessing, reconstructing the signal obtained after front-end preprocessing into a matrix form of a fixed size, applying a convolution layer with a 3×3 convolution kernel to perform a local convolution operation on the input matrix, and using a maximum pooling layer with a window size of 2×2 to downsample the convolution output, stacking the second convolution layer and its corresponding maximum pooling layer in sequence to extract deeper features, and flattening the feature map generated by multiple layers of convolution and pooling and mapping it to a feature vector of fixed dimension through a fully connected layer.

4. The multi-channel nuclear electronics data acquisition system based on artificial intelligence as claimed in claim 3, characterized in that: The dynamic determination of the sampling parameters of each channel according to the multi-dimensional state vector includes, starting from the system input data, extracting the information describing the multi-channel detection state through front-end signal processing and deep neural network, and clearly defining the action space, wherein the multi-dimensional state vector is as follows: ； in, Indicates time The multidimensional state vector of the next detection channel; is the signal-to-noise ratio; is the signal peak; NoiseMean is the noise mean; is the environmental parameter; The action definition is as follows: ； in, Indicates at time The sampling parameter combination selected by the system; is the discrete sampling rate; is the discrete integration time; is the discrete signal amplification factor; The states are discretized and mapped into discrete categories by clustering method: ； in, express The discretization result of ; is a discrete category; is the total number of categories; The deep Q network DQN is used to model the relationship between state and action, and the strategy update and parameter optimization are realized through the improved Actor-Critic structure. At the same time, cross-channel regularization is introduced to ensure the smoothness and consistency of the system. The Q function is defined as follows: ； in, Indicated in After taking the next action, the expected value of the cumulative discounted reward; is the network weight parameter; is the discount factor, which represents the impact of future rewards on current decisions; For the future moment Instant rewards; Indicates the moment; Next, use The function updates the network parameters as follows: DQN updates the loss function: ； in, is the loss function, which is used to measure the error between the current network prediction and the target value; The actual reward obtained at the current moment; Represents the parameters of the target network and plays a role in stabilizing updates; Indicates that in the next state The maximum expected reward that can be obtained under A placeholder for "all possible actions", used to search for the optimal action value in the next state; The Actor-Critic structure is used to refine the strategy through time difference updates, and regularization is introduced to ensure the consistency of cross-channel decisions: The timing difference error is as follows: ； in, is the temporal difference error, which measures the deviation between the current state value estimate and the actual result; is the state value function, estimated by the Critic network, is the state value function parameter; is the state value function of the next state; The Critic network loss function is defined as follows: ； in, is the mean square error of the Critic network, which is used to optimize the estimation of the state value function; The Actor network update gradient is as follows: ； in, To update the policy network parameters The gradient of For strategic networks Select in status Probability distribution of actions; Used to ensure smooth movements at consecutive moments; Used to constrain the consistency between sampling parameters of different detection channels; and is the regularization coefficient, which balances the importance of temporal smoothness and cross-channel consistency respectively; ∇ is the gradient operator, that is, the vector operation of partial derivative of the network parameter vector; In order to enable the system to take into account data quality, energy consumption and cross-channel consistency at the same time, a mechanism is introduced for the auxiliary network to dynamically generate the weights of various rewards, as follows: Dynamic weight generation: ； in, Indicates The dynamic weight of the reward is determined by the auxiliary network Calculate, where are the parameters of the auxiliary network; normalization condition Ensure that the sum of all weights is 1; The comprehensive reward function is as follows: ； in, For the system Taken under the state The overall reward for the action; To provide rewards corresponding to the quality of data; is the energy consumption penalty item; Rewards related to cross-channel consistency.

5. The multi-channel nuclear electronics data acquisition system based on artificial intelligence as claimed in claim 4, characterized in that: The dynamically determining the sampling parameters of each channel according to the multi-dimensional state vector includes: in order to make full use of the spatiotemporal dependency in the multi-channel detection information, a self-attention mechanism is used to realize the effective fusion of cross-channel information, thereby providing a more comprehensive input feature for the decision module; The query, key, and value mappings are: ； in, Indicates The detection channel is at time The state vector under They are used to generate queries ,key and numerical values The linear transformation matrix of is the total number of detection channels; The attention weight is calculated as follows: ； in, From channel To channel The attention weight of for The dimension of is used for normalization; is the sum index used to traverse all channels; Representation refers to the key of each traversed channel l; The fusion state vector is defined as follows: ； in, is the fused state vector, which represents the state vector after considering all channel information. The overall status of the channels.

6. The multi-channel nuclear electronics data acquisition system based on artificial intelligence as claimed in claim 5, characterized in that: The establishment of the correlation model of the dependency relationship between the signals of each channel includes performing statistical correlation analysis on the state vectors of each detection channel, calculating the Channel and Statistical correlation between channels , the formula is: ； in, Indicates time Next The state vector of each channel; Indicates time Next Pass The state vector of the Tao; Represents the state vector The mean of Represents the state vector The mean of Represents the state vector The standard deviation of Represents the state vector The standard deviation of is a small positive constant to avoid the denominator being zero; To adapt to graph neural network processing, nonlinear mapping is used Will Convert to low-dimensional embedding vector : ； in, The designed nonlinear mapping function is composed of fully connected layers or convolutional layers; is the dimension of the original state vector, is the dimension of the node after embedding; It is The embedded representation of the channels after mapping; Getting node embedding After that, the unnormalized attention score is calculated through the graph attention mechanism , the formula is: ； in, Indicates that the node is embedded and To splice; is the parameter vector to be trained; Indicates that the concatenated vectors are linearly combined to obtain a scalar; LeakyReLU is a linear rectification activation function with a negative slope. express right Initial attention scores; Normalize the unnormalized attention scores through the Softmax operation and calculate the attention weights between each node : ； in, Indicates from arrive The normalized attention weight of Will score Mapped to the space of positive real numbers; In the graph attention mechanism, it represents the unnormalized attention score calculated from channel i to channel q; the denominator is the same as channel All attention scores under are normalized to ensure ; In order to simultaneously utilize statistical correlation and attention scores obtained by graph neural networks, a weighted average method is used to generate the final cross-channel dependency matrix , the formula is: ； in, Indicates Channel and Comprehensive dependencies between channels; is the weight parameter, and its value range is ; is the value calculated by statistical correlation; is the normalized attention weight calculated by the graph neural network.

7. The multi-channel nuclear electronics data acquisition system based on artificial intelligence as claimed in claim 6, characterized in that: The calculation and application of the correction factor by the prediction method includes, The cross-channel correlation matrix is ​​used to perform traditional regression prediction on the abnormal channel to obtain the predicted value based on the remaining normal channel signals. The formula is: ； in, Indicates time Next abnormal channel Prediction signals obtained through statistical regression methods; Indicates time Next The actual acquisition signal of each detection channel; is the total number of detection channels, where abnormal channels are excluded when summing; Indicates Channel and Comprehensive dependencies between channels; On the basis of traditional regression prediction, a neural network model is designed to input signals of all channels except abnormal channels, and predict the nonlinear residual that cannot be captured by the regression model. The formula is: ； in, Indicates time The residual value obtained by the neural network prediction corresponds to the nonlinear prediction part of the abnormal channel; represents the neural network that constitutes this prediction model, and its parameters are given by express, The model can adopt a multi-layer perceptron or convolutional neural network structure; Indicates time Next, the actual signal set of all channels except the abnormal channel; The traditional regression prediction error and the residual error of the neural network prediction are weighted and fused in proportion to calculate the comprehensive compensation factor of the abnormal channel, and the formula is: ； in, Indicates time The comprehensive compensation factor of the lower abnormal channel; is a signal predicted based on statistical regression; Indicates time Lower abnormal channel The actual acquisition signal; represents the regression prediction error; is the residual predicted by the neural network; is a weight factor between 0 and 1, used to balance the contributions of the two parts; The comprehensive compensation factor is applied to the original signal of the abnormal channel to complete the data correction. The formula is: ； in, Indicates time The abnormal channel after correction signal.

8. The multi-channel nuclear electronics data acquisition system based on artificial intelligence as claimed in claim 7, characterized in that: The self-evolutionary multi-task optimization algorithm includes: input data , by a multi-task deep neural network Output predictions for multiple target parameters , construct a joint loss function To evaluate the network output error and adjust the network parameters Conduct global optimal search; use particle swarm optimization algorithm to update speed and parameters, forming a self-evolving parameter adjustment mechanism; The parameter update formula directly uses the updated velocity vector to correct the network parameters to form a new parameter set. ; This process is repeated in each iteration until the joint loss function The predetermined convergence conditions are met.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the multi-channel nuclear electronics data acquisition system based on artificial intelligence described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the artificial intelligence-based multi-channel nuclear electronics data acquisition system according to any one of claims 1 to 8 are implemented.

Images