Radioactive large tank demolition scheme optimization method and system based on digital twinning

Abstract

The invention relates to the technical field of digital twinning technology and radioactive waste treatment, and discloses a digital twinning-based radioactive large tank dismantling scheme optimization method and system. The method comprises the steps of obtaining multi-working-condition operation data and operator behavior data to generate a digital twinborn model basic data set, generating a system risk state quantitative index by using a system critical state evaluation model, identifying a potential operation error type based on an error pattern library, generating a risk coupling index by using a system-human factor risk coupling model, and establishing a twinborn model based on the risk coupling index. And generating a risk pre-control intervention strategy based on the risk coupling index. According to the method, the technical problem that system risks and human risks are separated in a traditional method is solved, and cooperative prevention of composite risks is achieved.

Description

Technical Field

[0001] This invention relates to the fields of digital twin technology, radioactive waste treatment technology, and risk control technology. More specifically, it relates to an optimization method for the dismantling of large radioactive containers based on digital twins. Background Technology

[0002] Currently, the dismantling of large radioactive storage tanks mainly employs traditional risk control methods. These methods typically treat system risk assessment and human operational risk analysis as two separate modules. System risk assessment focuses on equipment condition monitoring and parameter threshold control, while human risk analysis primarily focuses on the implementation of operating procedures and personnel training management.

[0003] However, traditional risk control methods have significant flaws: they treat systemic risks and human-caused risks separately, ignoring the interaction and coupled evolutionary patterns between the two. In actual dismantling operations, the system's critical state can affect the operators' psychological state and the difficulty of the operation, while human error can, in turn, exacerbate systemic risks, potentially creating a risk amplification effect. This fragmented risk management approach leads to incomplete risk prevention and control, making it difficult to cope with complex risk events and posing safety hazards to the dismantling of radioactive tanks. Summary of the Invention

[0004] This invention provides an optimization method for the dismantling of large radioactive tanks based on digital twins, which solves the technical problem in related technologies that separates system risks and human risks, ignoring their interaction and coupling evolution.

[0005] This invention discloses an optimization method for the dismantling of radioactive large tanks based on digital twins, comprising the following steps: acquiring multi-condition operation data and operator behavior data of the radioactive large tanks to generate a basic dataset for the digital twin model; analyzing the multi-condition operation data using a system critical state assessment model to generate a quantitative index of system risk status; analyzing operator behavior data based on an error pattern library to identify potential operational error types; analyzing the interaction between system risk status and operational error types using a system-human risk coupling model to generate a risk coupling index; and generating risk prevention and control intervention strategies based on the risk coupling index and system state prediction results. The quantitative index of system risk status is calculated by accumulating the weighted risk values ​​of each system parameter, and the risk value of each parameter is converted into a normalized risk value using a risk transformation function.

[0006] This invention discloses a system for executing the aforementioned optimization method for the dismantling of radioactive large tanks based on digital twins, comprising: a data acquisition module for acquiring multi-condition operating data and operator behavior data of the radioactive large tanks; a system risk assessment module for generating quantitative indicators of system risk status; an operation error identification module for identifying potential operation error types; a risk coupling analysis module for generating a risk coupling index; and a pre-control strategy generation module for generating risk pre-control intervention strategies.

[0007] Furthermore, in the data acquisition step, the acquired data is preprocessed, including outlier detection and handling, data normalization, operation behavior data encoding, and time series alignment.

[0008] Furthermore, the system critical state assessment model is an adaptive multi-layer parameter analysis model, which consists of a feature extraction layer, a parameter correlation layer, and a risk quantification layer. The risk quantification layer quantifies the impact of each parameter on system stability into risk indicators, and obtains the system risk state quantification index by weighting and summing the calculation results of the risk transformation function of each parameter.

[0009] Furthermore, based on the error pattern library, the analysis of operator behavior data employs a semantic distance-based pattern recognition algorithm. By calculating the similarity between the current operational behavior and the error patterns in the library, when the similarity exceeds a preset threshold, it is determined to be a potential error.

[0010] Furthermore, the system-human risk coupling model is constructed based on an improved Bayesian network, and the risk coupling index calculation considers the product of three factors: system risk index, human risk index, and system-human coupling coefficient.

[0011] Furthermore, the method also includes: using a multi-sensor data fusion algorithm to process operational behavior monitoring data and identify abnormal operational behaviors in real time; the data fusion algorithm adopts a hierarchical adaptive fusion strategy, including data layer fusion, feature layer fusion and decision layer fusion.

[0012] Furthermore, the method also includes: using an error consequence propagation simulator to process operational error data and system parameters, and predicting system state changes caused by erroneous operations; the error consequence propagation simulator consists of a state characterization module, a disturbance propagation module, and a state prediction module.

[0013] Furthermore, the method also includes: using a risk coupling evolution prediction algorithm to analyze historical data and current state to identify potential risk amplification paths; the risk coupling evolution prediction algorithm adopts a spatiotemporal graph neural network model and uses Monte Carlo simulation method to predict possible future risk evolution paths.

[0014] Furthermore, the risk prevention and control intervention strategies include system parameter adjustment suggestions and operational behavior guidance suggestions, which are generated through a weighted multi-objective adaptive optimization algorithm. The optimization objective function takes into account the weighted sum of the risk coupling index and intervention costs.

[0015] The present invention provides an optimization method for the dismantling of radioactive large tanks based on digital twins. By establishing a system-human risk coupling model, it achieves quantitative analysis of the interaction between the system's critical state and human operational errors. This solves the technical problem of traditional risk prevention and control methods that separate system risk and human risk, ignoring the interaction and coupling evolution law between the two, and achieves the following technical effects: By using a system critical state assessment model under multiple operating conditions, the risk of the radioactive large tank system was accurately quantified, and the accuracy of risk identification was improved. By constructing an error pattern library and using a semantic distance-based pattern recognition algorithm, early identification of potential operational errors was achieved, enhancing the ability to prevent and control human-related risks. By introducing a system-human factor risk coupling model, the impact of system state on operational difficulty and the impact of operational errors on system state are revealed, realizing coupled risk assessment and overcoming the problem of traditional methods neglecting risk interaction. By generating risk prevention and intervention strategies, targeted system adjustments and operational guidance suggestions are provided, effectively preventing the occurrence of complex risk events and significantly improving the safety and reliability of demolition operations; By combining data-driven approaches and model analysis, comprehensive pre-control of complex risks during the dismantling of radioactive tanks was achieved, providing safe and efficient technical support for nuclear facility decommissioning and radioactive waste disposal. Attached Figure Description

[0016] Figure 1 This is an overall flowchart of the optimization method for the dismantling of large radioactive tanks based on digital twins, which shows the main steps from data acquisition to the generation of risk prevention and control intervention strategies. Detailed Implementation

[0017] The method of this embodiment includes the following steps; Step 100: Obtain multi-condition operation data and operator behavior data of the large radioactive tank to generate the basic dataset for the digital twin model; In step 100, system parameter data for the radioactive container under different working conditions, operating environments, and dismantling stages are acquired, including but not limited to: internal pressure, temperature distribution, radiation levels, structural stress, and other system parameters; simultaneously, behavioral data of operators during the dismantling process is acquired, including operation trajectories, operation times, and operation sequences. The acquired data will be used for subsequent system criticality assessment and human factor risk analysis.

[0018] Before using the acquired raw data, the following data preprocessing operations must be performed: Outlier detection and handling: Outliers are detected using the 3-sigma rule, and outlier data points are replaced or removed using a moving average. Furthermore, the implementation details of the 3-sigma rule are as follows: for any system parameter sequence {x_1, x_2, …, x_n}, where , … Let x_i, x_i, x_i, ..., x_n represent the 1st, 2nd, ..., nth system parameter values, respectively. First, calculate their mean μ and standard deviation σ. Then, for any data point x_i, if |x_i - μ| > 3σ, it is considered an outlier. The moving average window size used for replacement is 5, that is, (x_{i-2} + x_{i-1} + x_{i+1} + x_{i+2}) / 4 replaces the outlier x_i. Data normalization: Since the system parameters have different physical dimensions (such as pressure in Pa, temperature in °C, radiation level in Sv / h, etc.), all system parameter data are normalized using Min-Max. Operation behavior data encoding: Non-numerical operation behavior data (such as operation type, operation object, etc.) are converted into numerical data vectors through one-hot encoding for subsequent model processing; Furthermore, for the complex data encoding of operational behaviors, the implementation details are as follows: First, construct an operation type dictionary D = {operation type 1, operation type 2, …, operation type m}, where operation type 1, operation type 2, …, operation type m represent the first, second, …, and m different operation behavior types, respectively. For any operation type i, encode it as an m-dimensional vector v, where the i-th element is 1 and the remaining elements are 0. For multi-attribute operational behaviors, one-hot encoding of different attributes is used to concatenate them to form a joint feature vector. Time series alignment: Time synchronization processing is performed on sensor data with different sampling frequencies, and cubic spline interpolation is used to align all data to a unified timestamp sequence.

[0019] Step 200: Analyze multi-condition operating data using the system critical state assessment model to generate quantitative indicators of system risk status; In step 200, the multi-condition operating data obtained in step 100 is input into the system critical state assessment model to calculate the quantitative index of the system risk state of the radioactive tank under different operating conditions. The system critical state assessment model is an adaptive multi-layer parameter analysis model, whose structure consists of a feature extraction layer, a parameter correlation layer, and a risk quantification layer.

[0020] The mathematical representation of the feature extraction layer is: ; in, For input multi-condition operation data, This is the weight matrix of the feature extraction layer. For bias vectors, This is an activation function (such as the ReLU function).

[0021] Furthermore, the network structure of the system critical state evaluation model is specifically configured as follows: the feature extraction layer uses a three-layer fully connected neural network, with 128, 256, and 128 neurons in each layer, respectively. The activation function is ReLU. To prevent overfitting, a Dropout layer is added between each layer, with a dropout rate of 0.2. The input layer dimension is determined based on the number of system parameters, typically 50-100 dimensions. The initialization method uses the Xavier method to ensure that the variances of the input and output of each layer are similar, which helps to stabilize signal transmission. Batch processing is used during training, with a batch size of 64 and an initial learning rate of 0.001. The mathematical representation of the parameter association layer is: ; in, The weight matrix of the parameter association layer. This is the bias vector.

[0022] Furthermore, the parameter correlation layer employs a two-layer fully connected neural network, with 64 and 32 neurons in each layer, respectively. The activation function used is Leaky ReLU (with a negative slope of 0.01), which avoids the "dead neuron" problem in the negative region of ReLU. The parameter correlation layer also introduces an attention mechanism, which calculates the correlation weight matrix between different system parameters. This enables the network to adaptively learn the interaction relationships between parameters: ,in and For learnable query and key matrices, This indicates transpose. This structural design enables the model to capture complex nonlinear dependencies between different system parameters; The risk quantification layer quantifies the impact of each parameter on system stability into risk indicators. : ; in, The number of system parameters, For the first One system parameter, The weighting coefficients for this parameter (the weighting coefficients satisfy...) and ), This is the risk conversion function corresponding to this parameter.

[0023] Furthermore, parameter weighting coefficients The method for determining the importance of each parameter is the Analytic Hierarchy Process (AHP), which involves: first, domain experts comparing the importance of each parameter pairwise to construct a judgment matrix A; then, the largest eigenvalue of matrix A is calculated. And the corresponding feature vector v; finally, the weight coefficients are obtained by normalizing the feature vector v. Simultaneously, a consistency test is performed, requiring a consistency ratio (CR) < 0.1, where CR = CI / RI. RI is the random consistency index; due to the various system parameters Risk transformation function with different physical dimensions Converting it to a dimensionless normalized risk value, defined as: ,in, The minimum safe value for the parameter. For parameter safety threshold, For the critical threshold parameter. All The values ​​of are all dimensionless values ​​in the range of [0,1], representing the risk level from the safe state (0) to the critical state (1), ensuring that the dimensions of the parameters of different physical quantities are consistent when they are added together.

[0024] Furthermore, parameter safety threshold and parameter critical threshold The method for determining it is: for each system parameter Based on historical operational data and relevant national safety standards (such as GB / T 1254-2017 "Regulations for the Safe Transportation of Radioactive Materials"), Set it to the upper limit of the normal operating range of this parameter. This parameter is set as the critical value at which system instability begins to occur. Specifically, for pressure parameters, To design 80% of the work pressure, The design working pressure is 95%; for temperature parameters, 85% of the maximum permissible operating temperature It is 97% of the maximum permissible operating temperature; for radiation level parameters... 70% of the annual dose limit It is 90% of the annual dose limit; The system critical state evaluation model is trained using a supervised learning method, and the loss function consists of two parts: mean squared error loss and safety constraint loss. ; Differences between mean squared error loss calculation-based predicted risk indicators and labeled risk indicators: ; in, The number of training samples. For the first The true risk index of each sample This is a risk indicator predicted by the model.

[0025] Safety constraint losses are used to ensure the conservatism of risk assessments and avoid underestimating risks. ; in, The loss function, defined as the tolerance threshold, penalizes underestimation of risk, ensuring the safety and reliability of the evaluation results. Model parameters are updated via backpropagation and the Adam optimizer, with the learning rate dynamically adjusted using a cosine annealing strategy.

[0026] Furthermore, the allowable error threshold The value is set to 0.05, meaning that the model-predicted risk index is allowed to be up to 0.05 (5%) higher than the actual risk index, but underestimation of risk is not allowed to exceed this threshold. This value was determined through Monte Carlo simulation and expert evaluation to ensure a safety margin while avoiding reduced operational efficiency due to excessive conservatism. It should be noted that the calculation of the above-mentioned system risk state quantification indicators can also be based on the structural characteristics and physical properties of the radioactive tank, and can be represented using state space: ; in, for The system state vector at time t. To control the input vector, Here is the state transition matrix. For the control matrix, This represents system noise. The quantification index of system risk state can be obtained through the state vector. The distance metric from the preset security domain is determined.

[0027] Step 300: Analyze operator behavior data based on the error pattern library to identify potential operational error types; In step 300, the operator behavior data obtained in step 100 is matched and analyzed with a pre-established error pattern library to identify potential operational error types. The error pattern library contains typical operational error patterns that may occur during the dismantling of radioactive tanks. Each error pattern includes attributes such as error description, triggering conditions, probability of occurrence, and severity.

[0028] The error pattern matching process employs a semantic distance-based pattern recognition algorithm to calculate the similarity between the current operation and error patterns in the database. : The aforementioned semantic distance-based pattern recognition algorithm converts operational behavior data and error pattern data into feature vectors, and performs matching by calculating the distance in the semantic space. Specifically, the algorithm's input consists of the operational behavior feature vector and the feature vectors of each error pattern in the error pattern library, and the output is a similarity score and the possible matched error patterns. The semantic distance-based pattern recognition algorithm employs a weighted similarity calculation method, giving higher weight to key features to enhance matching accuracy.

[0029] ; in, As a behavioral feature dimension, For the current operation 3D eigenvalues For the first error mode 3D eigenvalues For feature weights, This is the similarity calculation function.

[0030] Furthermore, feature weights The method for determining this is based on information entropy theory, and is determined according to the information gain of each feature: ,in Features Information gain, calculated using the following formula: , Entropy for error types, Known features Conditional entropy of error types under given conditions. For key features, such as operation sequence and safety interlock status, the weight range is [0.15, 0.25]; for secondary features, such as operation speed and operation force, the weight range is [0.05, 0.15]. To ensure comparability between different types of features, the operation behavior data needs to be preprocessed as follows before similarity calculation: Numerical behavioral characteristics (such as operation duration, operation speed, etc.) are Z-score standardized: ; in, The mean of this feature. Standard deviation; Categorical features (such as operation type, operation order, etc.) are converted into vector representations through one-hot encoding; Sequence-type features (such as operational trajectories) are used to calculate the distance between sequences using the Dynamic Time Warping (DTW) algorithm, and then converted into similarity values ​​using a Gaussian kernel function. ; in, Represents an exponential function. Furthermore, the standard deviation parameter in the Gaussian kernel function An adaptive method is used to determine the distribution characteristics of sequence distances in the training dataset: , where the set This represents the sequence distance value of the i-th training sample. The number of training samples. This method, which represents the median operation, allows the similarity function to be better adapted to different feature scales. when Exceeding the preset threshold If this occurs, it is considered a potential error. Preset threshold. The threshold was set to 0.75, which was determined through ROC curve analysis. At this threshold, the sensitivity of false alarm detection was 0.92, the specificity was 0.88, and the F1 score was 0.90, which can control the false alarm rate within an acceptable range while ensuring the detection rate. In this embodiment of the application, to improve the real-time performance of operational error identification, step 300 may further include: Step 310: Process the operational behavior monitoring data using a multi-sensor data fusion algorithm to identify abnormal operational behaviors in real time; In step 310, multiple sensors (including vision sensors, motion capture sensors, force feedback sensors, etc.) are deployed to collect real-time behavioral data of operators. A data fusion algorithm integrates this multi-source heterogeneous data to form a unified representation of operational behavior, which is then compared with standard operating procedures to identify abnormal operational behaviors in real time. The data fusion employs a hierarchical adaptive fusion algorithm, including data-level fusion, feature-level fusion, and decision-level fusion. The aforementioned hierarchical adaptive fusion algorithm employs time synchronization and data standardization methods at the data layer to transform heterogeneous data from different sensors into a unified time reference and numerical range. Feature layer fusion uses a deep autoencoder to extract effective features from the standardized data, generating a data representation in a unified feature space. Decision layer fusion combines the decision results from each sensor through an adaptive weighting mechanism to generate the final judgment.

[0031] The algorithm takes heterogeneous data streams from multiple sensors as input and outputs unified operational behavior representations and abnormal operation detection results. It employs an adaptive learning approach, dynamically adjusting the fusion weights based on the data quality from different sensors to improve the freshness and reliability of the data fusion.

[0032] ; in, For the fusion result, , , These are the fusion results of the data layer, feature layer, and decision layer, respectively. , , These are the weighting coefficients for each layer. It is a non-linear mapping function.

[0033] Furthermore, the weight coefficients of each layer , , The determination adopts an adaptive weight allocation strategy, which dynamically adjusts the quality indicators based on the data from each sensor: ,in The quality metric for the i-th layer fusion is calculated by comprehensively considering signal-to-noise ratio, data integrity, and fusion accuracy. The initial weights are set to... , , As the system accumulates information, the weight coefficients are continuously and adaptively updated; this nonlinear mapping function Implemented using a multilayer perceptron architecture, defined as: ; in, , This is the weight matrix. , Here, is the bias vector, and ReLU is the modified linear unit activation function. The nonlinear mapping function of the multilayer perceptron structure can capture the nonlinear relationship between the fusion results of different levels, improving the expressive power of multi-sensor data fusion.

[0034] Step 400: Analyze the interaction between system risk status and operational error type using the system-human factor risk coupling model, and generate a risk coupling index; In step 400, the system risk state quantification index obtained in step 200 and the potential operational error types identified in step 300 are input into the system-human risk coupling model to analyze the interaction between system state and human operation, and generate a risk coupling index. The system-human risk coupling model is constructed based on an improved Bayesian network to characterize the impact of system state on operational difficulty and the impact of operational errors on system state.

[0035] The improved Bayesian network consists of a set of system state nodes. Operation behavior node set and coupling relationship edge set Composition, its probabilistic graphical model is represented as ,in A set of nodes. Each system state node. This represents a system parameter, and each operation behavior node... This represents a type of operation behavior. Directed edges between nodes. This indicates a conditional dependency.

[0036] The joint probability distribution of the system-human-factor risk coupling model is as follows: ; in, Represents a node The set of parent nodes, It is a conditional probability distribution.

[0037] The system-human-cause risk coupling model is trained using the maximum likelihood estimation method, with the loss function being the negative log-likelihood. ; in, For the training dataset, For data samples The node values ​​are determined. The optimization uses gradient descent to update the model parameters.

[0038] Risk Coupling Index The calculation formula is: ; in, For time parameters, for System risk indicators at any given time for Both are dimensionless normalized values, representing human-caused risk indicators at any given time, and their values ​​range from [0,1]. for The system-human coupling coefficient at any given time. The calculation considers the influence of system state on operational difficulty. Factors influencing the system state and operational errors : ; Furthermore, the temporal evolution characteristics of the risk coupling index are manifested in the following aspects: System risk indicators Timing calculation: ,in This represents the base of the natural exponential function. The system risk attenuation coefficient (with a value of 0.05) is used. The time step is 0.5 hours. This represents the system risk value calculated based on real-time data at the current moment. Human-made risk indicators Considering operator fatigue: ,in Based on the risk value of the basic person, This is the start time of the operator's current shift. The standard shift duration is 8 hours. The fatigue coefficient (valued at 0.3) is used in this formula to show that as the continuous working time of operators increases, the risk of human error gradually rises. System-human coupling coefficient This reflects the dynamic interaction between system state and human operation, calculated using the sliding time window method: ,in The width of the time window (with a value of 6). The sampling time interval (value is 10 minutes). for The instantaneous coupling coefficient at time t; in, , , Let be the weighting coefficient, satisfying and Impact Factor and The risk coupling index is obtained through standardization of system state parameters and operational error characteristics. These are dimensionless values ​​within the range [0,1], ensuring dimensional consistency during weighted combination. It is also a dimensionless normalized value, with a range of [0,1], representing the degree of coupling risk from low risk (0) to high risk (1).

[0039] Furthermore, weighting coefficients , and Based on historical accident data analysis and expert experience, the initial value was set to... , , This indicates that the impact of operational errors on the system state is slightly greater than the impact of the system state on the operational difficulty; the interaction effect between the two has a relatively small but not negligible weight. (Influence Factor) and The specific calculation method is as follows: ,in For system parameters The function affecting the difficulty of operation; ,in Number of operation error types Operational error Influence functions on system parameters. These influence functions are constructed based on piecewise linear mappings, converting inputs of different dimensions into dimensionless outputs within the interval [0,1]. In this embodiment of the application, in order to more accurately predict the impact of operational errors on the system state, step 400 may further include: Step 410: Utilize the error consequence propagation simulator to process operational error data and system parameters, and predict system state changes caused by erroneous operations; In step 410, an error consequence propagation simulator is used to simulate and analyze the impact path and extent of potential operational errors identified in step 300 on various system parameters, and to predict the trajectory of system state changes. The error consequence propagation simulator consists of a dynamic influence graph network, including a state representation module, a disturbance propagation module, and a state prediction module.

[0040] The state characterization module encodes system parameters into state vectors: ; in, for The system state vector at time t. For encoding functions, This represents the implicit state representation. The encoding function of the state representation module is implemented using a bidirectional long short-term memory (BiLSTM) network, defined as follows: ; in, This is the parameter set of the BiLSTM network. This encoding function can extract the temporal features and key attributes of the system state and generate a compact state representation.

[0041] Furthermore, the specific structural configuration of the BiLSTM network is as follows: the input dimension equals the number of system state parameters (typically 20-30 dimensions), the hidden layer dimension is 128, the number of layers is 2, and the forward and backward LSTM outputs are merged through concatenation. The network also employs a residual connection structure to directly connect the input to the output to alleviate the gradient vanishing problem; batch normalization is used to accelerate network convergence and improve stability; and temporal attention is used to highlight features at key time points, with the attention weights calculated using the following formula: ,in , and For learnable parameters, This indicates transpose. This structural design can effectively capture the long-term dependencies and short-term fluctuations of the system state; The disturbance propagation module calculates the propagation effect of disturbances introduced by operational errors in the system: ; in, For operation error vectors, For the propagation function, Let be the perturbation effect vector. The propagation function is implemented using a gated graph convolutional network (GGCN) and is defined as follows: ; in, This represents the concatenation of the state representation and the error vector. The adjacency matrix represents the system parameter dependencies. As parameters of a graph convolutional network, the propagation function of a gated graph convolutional network (GGCN) can simulate the propagation effect of disturbances in the system parameter network.

[0042] Furthermore, the specific implementation details of the Gated Graph Convolutional Network (GGCN) are as follows: The network contains 3 layers of graph convolutions, with feature dimensions of 64, 32, and 64 for each layer; the mathematical form of each graph convolution is... ,in For degree matrix, It is an adjacency matrix. For the first Layer node characteristics, The weight matrix is ​​used; the gating mechanism introduces an update gate. and reset door To achieve this, the calculation formula is: , The output is ,in Candidate activation values, This represents element-wise multiplication. The gating mechanism of Gated Graph Convolutional Networks (GGCNs) enables the network to adaptively control the flow of information and selectively update node states, thereby more accurately simulating the complex interactions between system parameters and the effects of disturbance propagation. The state prediction module predicts the future system state based on the current state and the disturbance effect: ; in, To control the input vector, For the prediction function, The predicted future state. The prediction function is implemented using a multi-layer feedforward neural network and is defined as: ; in, This represents the concatenation of state representation, disturbance effects, and control inputs. , This is the weight matrix. , As the bias vector, the multilayer feedforward neural network prediction function can comprehensively consider the current state, disturbance effects, and control input to accurately predict the future state of the system.

[0043] The error consequence propagation simulator is trained using a supervised learning method, with the loss function being the mean squared error between the predicted and actual states. ; in, This represents the number of training samples. Model parameters are updated via backpropagation and the Adam optimizer.

[0044] The aforementioned simulation of error consequence propagation is based on a system dynamics model and employs state transition equations: ; in, for The system state vector at time t. For normal control input, Disturbances introduced by operational errors These are the system dynamic equations. By solving these equations through numerical integration, the evolution of the system state after an operational error can be predicted.

[0045] Furthermore, the definition and range of values ​​for the time parameter in the state transition equation are as follows: Time variable Physical meaning: Represents the cumulative time from the start of demolition work, in hours, with a range of values. ,in The estimated total duration of the entire demolition operation (typically hundreds to thousands of hours). Time step Settings: The basic time step is set to 10 minutes (i.e., (Hours) is used for evolution prediction of general states; for the critical state region, an adaptive time step is used. ,in The sensitivity coefficient (with a value of 10) ensures that a smaller time step is used to improve prediction accuracy when the state changes rapidly; Integral Time Window : Indicates the time range for state prediction; the prediction window length is typically set to 0. The timeframe can be dynamically adjusted based on the specific task being predicted. System dynamics equations Discretization implementation: Numerical integration is performed using the fourth-order Runge-Kutta method (RK4), i.e.: ; ; ; ; ; The truncation error of this integration method is It can provide high-precision state prediction results; Step 500: Based on the risk coupling index and system state prediction results, generate risk prevention and control intervention strategies; In step 500, based on the risk coupling index calculated in step 400 and the system state prediction results, targeted risk prevention and intervention strategies are generated. These strategies include two aspects: system parameter adjustment suggestions and operational behavior guidance suggestions.

[0046] The pre-control intervention strategy is generated using a weighted multi-objective adaptive optimization algorithm, with the objective function being: The aforementioned weighted multi-objective adaptive optimization algorithm solves the problem by transforming the two objectives of risk coupling index and intervention cost into a weighted single-objective problem. The input to the weighted multi-objective adaptive optimization algorithm is the current system state, the risk coupling index, and a set of possible intervention operations; the output is the optimal parallel vector of intervention strategies.

[0047] The algorithm employs an adaptive weight adjustment strategy, dynamically adjusting the weight ratios of the risk index and cost in the objective function based on the current risk level. When the risk level reaches a severe level, the weight of the risk index is increased, while the impact of cost factors is reduced; when the risk is within an acceptable range, minimizing intervention costs is considered. The algorithm uses an improved particle swarm optimization method to solve the problem, rapidly converging to the optimal solution that satisfies the constraints through a combination of global search and local refinement.

[0048] ; in, As the intervention action vector, The time parameter indicates that at Intervention actions implemented at all times In order to be in Intervention actions should be taken at all times. The subsequent risk coupling index, The cost function for the intervention action (converted to a dimensionless value through normalization). and These are the weighting coefficients.

[0049] Furthermore, optimizing the time dimension of the objective function is reflected in the following aspects: Dynamic risk coupling index: Over time Dynamic changes reflect the time evolution characteristics of the system state, and the calculation formula is: ,in express The change in risk at any given moment; Time-series differentiated costs: intervention costs ,in This represents the base of the natural exponential function. The total duration of the demolition task. and The weights are 0.5 and 0.2 respectively. This function makes the weights decrease as the task nears its end. near Interventions implemented at lower costs encourage early preventative measures. Time lag in intervention effects: The effects of intervention actions have a time lag, as shown by the effect function. Modeling, meaning in Intervention actions implemented at all times exist The effect of time, including time delay Time unit; Furthermore, weighting coefficients and Based on the current risk coupling index Horizontal adaptive adjustment: when (In the high-risk zone) , Prioritize risk reduction; when (In the medium-risk range) , Balancing risks and costs; when (In the low-risk range) , Risks and costs are considered equally. Intervention cost function. The weighted sum, including economic cost, time cost, and operational complexity, is converted into a dimensionless value in the [0,1] interval using Min-Max normalization. The constraints of the optimization problem include: ; Furthermore, the completeness of the constraints in the optimization problem is reflected in the following five aspects: Risk and security constraints: Requires the entire prediction time domain Within this range, the risk coupling index does not exceed the safety threshold. ,in To predict the length of the time domain (usually set to 24 hours); Action space constraints: The range of values ​​for each component in the intervention action vector was defined; Motion smoothness constraints: To ensure that intervention actions do not change drastically between adjacent time points, This is set to the maximum permissible range of change (usually 0.2), which helps to avoid system oscillations and operational chaos. Resource budget constraints: The total resource consumption of the intervention was limited, among which For the first The unit resource consumption coefficient of each intervention action This is the upper limit of the total resource budget; Time constraints: The time limit for intervention actions was restricted. The range, of which and These are the minimum and maximum allowed execution times, respectively. in, This represents an acceptable threshold for the risk coupling index. and These are the lower and upper bound vectors for the intervention action, respectively.

[0050] Furthermore, the acceptable threshold for the risk coupling index. Based on the safety level settings for the dismantling of radioactive storage tanks, for dismantling work at the standard safety level... For high-security-level demolition work, For demolition work at an extremely high safety level, Upper and lower bound vectors of intervention actions and Based on the system design parameters and operational safety specifications, the adjustment range of system parameters is limited to 70% to 90% of the design safety boundary; the scope of operational behavior guidance is determined based on standard operating procedures and the qualifications and capabilities of operators. Furthermore, intervention action vectors It is A dimensional vector containing two parts: a subvector for adjusting system parameters. and operational behavior guidance subvector ,in This represents the total dimension of the intervention action vector. Each dimension of the system parameter adjustment sub-vector corresponds to an adjustable system parameter (such as pressure, temperature, ventilation rate, etc.), typically a normalized value within the range of [0, 1]. Each dimension of the operational behavior guidance sub-vector corresponds to the strength or priority of an operational guidance strategy, also represented by a normalized value within the range of [0, 1]. The timestamp attribute of the intervention action vector. The execution time of the action is defined, and its value range is: ,in For the current time, This is the predicted timeframe (typically the next 8 hours). The vector also includes the execution duration. This indicates the effective duration of the intervention, with a value range of [value missing]. Hour; The abstract intervention strategy output by the optimization algorithm needs to be decoded and transformed into specific, executable operation instructions. The decoding process includes: System parameter adjustment decoding: The optimized parameter adjustment vector is mapped to specific system control commands, such as pressure regulation values, temperature control setpoints, and radiation protection measures, in the following form: ; in, Indicates the system parameters Adjust to target value Specific control commands; Operational behavior guidance decoding: The operational behavior optimization vector is converted into a standard operating procedure (SOP) and warning prompts, in the following form: ; in, This is a description of the operation steps. Warning messages are issued for specific risk points. The decoded intervention strategy is presented to operators through a human-machine interface system to guide the safe execution of subsequent demolition operations.

[0051] In this embodiment of the application, in order to improve the foresight of the risk prevention and control strategy, step 500 may further include: Step 510: Analyze historical data and current status using the risk coupling evolution prediction algorithm to identify potential risk amplification paths; In step 510, based on risk event cases in historical data and the current system-human coupling state, a risk coupling evolution prediction algorithm is applied to identify evolutionary paths that may lead to risk amplification. The risk coupling evolution prediction algorithm uses a spatiotemporal graph neural network model to construct a risk state transition network.

[0052] The spatiotemporal graph neural network model consists of a temporal embedding layer, a spatial graph convolutional layer, and an evolution prediction layer.

[0053] Furthermore, the specific parameter configuration of the spatiotemporal graph neural network model is as follows: The network consists of three main components, and the parameter configuration of each component is optimized according to the characteristics of the input data and the model requirements. For the input data, the historical window size... The system is set to 10 time steps, analyzing the risk status within the most recent 10 time units; the batch size is 32; training uses the Adam optimizer with an initial learning rate of 0.0005, employing a learning rate scheduling strategy that decays to 80% of the original rate every 20 epochs; the dropout rate is 0.15 to prevent overfitting; the model uses an early stopping strategy, stopping training if performance does not improve for 5 consecutive epochs on the validation set; to address the imbalanced sample problem, oversampling is used for high-risk samples to increase their proportion in the training data, making the ratio of high, medium, and low-risk samples approximately 1:2:2. The temporal embedding layer encodes historical risk state sequences into temporal features: ; in, for The risk status at all times, For the size of the history window, For time coding functions, This is the temporal feature vector. The temporal encoding function is implemented using a temporal attention mechanism and is defined as follows: ; in, For extracting time features, Attention weights are calculated as follows: ; in, Represents an exponential function. in, Here is the attention parameter matrix. As an attention vector, the temporal attention mechanism's temporal encoding function can adaptively extract temporal features based on the importance of the state.

[0054] Furthermore, the specific parameters of the temporal embedding layer are set as follows: the input state dimension is... Time feature extraction matrix The dimension is This involves mapping the input state to a 64-dimensional latent space; attention parameter matrix The dimension is Attention vector The dimension is This layer also employs positional encoding to enhance the model's ability to perceive temporal order. Positional encoding is implemented using a sine-cosine function. , ,in For location index, For dimensional indexing, For model dimensions; Spatial graph convolutional layers capture the spatial dependencies between system parameters and operational behavior: ; in, For system parameter set, A collection of events performed by human operators. It is a spatial convolution function. This represents the spatiotemporal feature vector. The spatial convolution function is implemented using a heterogeneous graph convolutional network (HGCN), defined as: ; in, The constructed heterogeneous graph consists of nodes including system parameter nodes and operation event nodes, with edges representing the interaction relationships between them. and These are the feature matrices for system parameter nodes and operation event nodes, respectively. As parameters for graph convolutional networks, the spatial convolution function of Heterogeneous Graph Convolutional Networks (HGCN) can effectively capture the complex interaction relationships between different types of nodes.

[0055] Furthermore, the specific implementation method of Heterogeneous Graph Convolutional Network (HGCN) is as follows: First, the heterogeneous graph contains two types of nodes (system parameter nodes and operation event nodes) and three types of edges (parameter-parameter edges, event-event edges, and parameter-event edges); different transformation matrices are defined for each type of edge to form message passing of specific relationships; the message passing formula is, where Indicates the relation type, It is a relationship-specific transformation matrix with dimension . (Input feature dimensions) Output feature dimension Then, for each node, aggregate messages from different relations: where Indicates the relationship Connect to node The neighbor set; finally, node feature updates employ a gating mechanism: The GRU is a gated recurrent unit used to adaptively fuse current features and aggregated neighbor messages; the HGCN network is stacked with 3 layers, and batch normalization and residual connections are added after each layer to improve training stability and alleviate the gradient vanishing problem. The evolutionary prediction layer generates the risk state transition probabilities for the next time step: ; in, The prediction function is defined as follows: The evolutionary prediction layer prediction function is implemented using a variational autoencoder (VAE) structure. ; in, As latent variables, generated by the encoder: ,in and According to the encoder and The calculation yielded: ; Variational autoencoder (VAE) prediction functions can model the uncertainty of risk state transitions and generate probabilistic prediction results.

[0056] The model is trained using the conditional likelihood maximization method, with the loss function being the negative log-likelihood. ; in, The sequence length is given. Optimization uses the RMSprop algorithm, and gradient clipping is employed to prevent gradient explosion. After model training, the risk evolution path is predicted using Monte Carlo simulation. ; in, for The risk status at all times, For system parameter set, A collection of events performed by human operators. Let be the state transition function. The state transition function for risk evolution path prediction, based on a trained spatiotemporal graphical neural network model, is defined as: ; in, To increase the number of integrated models, For the first The weight coefficients of each model, For the first The spatiotemporal feature vectors of each model are used. The state transition function for risk evolution path prediction is improved in terms of robustness and accuracy through model ensemble. Monte Carlo simulation is used to predict possible future risk evolution paths, calculate the probability of each path, and identify high-risk paths.

[0057] In this embodiment, the aforementioned Monte Carlo simulation method employs importance sampling to improve simulation efficiency. The input to the importance sampling Monte Carlo simulation algorithm is the conditional probability distribution generated by the evolutionary prediction model. and current risk status The output is a series of possible risk evolution paths and their probability of occurrence.

[0058] The algorithm first bases itself on the current state. Multiple state transition samples are generated, and then transition samples are recursively generated from each newly generated state to form a state transition tree. For each path, its cumulative probability and risk index are calculated. To improve computational efficiency, the algorithm employs an importance sampling technique, prioritizing the exploration of paths with higher risk indices, and normalizes the path probabilities to ensure the accuracy of the calculation results.

[0059] In a nuclear power plant decommissioning project, a waste liquid storage tank needed to be dismantled. The tank was 6 meters in diameter, 8 meters high, and had a volume of approximately 200 cubic meters. It contained waste liquid containing radioactive materials such as Cs-137 and Sr-90. The dismantling work consisted of three main stages: waste liquid extraction, equipment dismantling, and tank cutting. The operating environment was complex, and there was a significant interaction between system status and human operational errors.

[0060] Step 100 Implementation: Acquisition of multi-condition operation data; During the waste liquid extraction phase, system parameters and operator behavior data are acquired through a deployed sensor network. After data preprocessing, a basic dataset is formed, as shown in Table 1: Table 1. Real-time monitoring data of system parameters; ; Operator behavior data is shown in Table 1A: Table 1A: Operator Behavior Data; ; Step 200: System critical state assessment; Multi-condition operating data are input into the system critical state assessment model to calculate the risk state quantification index at each time point, as shown in Table 2: Table 2. Quantitative results of system risk status; ; Step 300: Operational error identification; Based on the analysis of operator behavior data using an error pattern library, potential operational error types are identified, as shown in Table 3: Table 3. Results of Operational Error Identification; ; Step 400 Implementation: Risk Coupling Analysis; The interaction between system risk status and operational errors was analyzed using a system-human factor risk coupling model, as shown in Table 4: Table 4. Calculation results of the risk coupling index; ; Step 410 Implementation: Simulation of Error Consequence Propagation; Using an error consequence propagation simulator, the system state change trajectory is predicted for potential operational errors identified in step 300, as shown in Table 4A: Table 4A: Predicted results of the impact of operational errors on system state; ; Step 500 Implementation: Generation of risk prevention and control intervention strategies; Based on the risk coupling index and system state prediction results, an intervention strategy is generated using a weighted multi-objective adaptive optimization algorithm, as shown in Table 5: Table 5 Risk prevention and control intervention strategies; .

Claims

1. A method for optimizing the dismantling of large radioactive tanks based on digital twins, characterized in that, Includes the following steps: Acquire multi-condition operating data and operator behavior data of large radioactive tanks to generate the basic dataset for digital twin models; The system critical state assessment model is used to analyze multi-condition operating data and generate quantitative indicators of system risk status. Based on error pattern analysis, analyze operator behavior data to identify potential operational error types; The interaction between system risk status and operational error type is analyzed using a system-human factor risk coupling model to generate a risk coupling index. Based on the risk coupling index and system state prediction results, risk prevention and control intervention strategies are generated. Among them, the system risk status quantification index is calculated by accumulating the weighted risk values ​​of each system parameter, and the risk value of each parameter is converted into a normalized risk value through a risk transformation function.

2. The method according to claim 1, characterized in that, In the data acquisition step, the acquired data is preprocessed, including outlier detection and handling, data normalization, operation behavior data encoding, and time series alignment.

3. The method according to claim 1, characterized in that, The system critical state assessment model is an adaptive multi-level parameter analysis model, which consists of a feature extraction layer, a parameter correlation layer, and a risk quantification layer. The risk quantification layer quantifies the impact of each parameter on system stability into risk indicators, and obtains the system risk state quantification index by weighting the calculation results of the risk transformation function of each parameter.

4. The method according to claim 1, characterized in that, The analysis of operator behavior data based on an error pattern library employs a semantic distance-based pattern recognition algorithm. By calculating the similarity between the current operation and the error patterns in the library, a potential error is identified when the similarity exceeds a preset threshold.

5. The method according to claim 1, characterized in that, The system-human risk coupling model is constructed based on an improved Bayesian network. The risk coupling index is calculated by considering the product of three factors: system risk index, human risk index, and system-human coupling coefficient.

6. The method according to claim 1, characterized in that, The method also includes: using a multi-sensor data fusion algorithm to process operational behavior monitoring data and identify abnormal operational behaviors in real time; the data fusion algorithm adopts a hierarchical adaptive fusion strategy, including data layer fusion, feature layer fusion and decision layer fusion.

7. The method according to claim 1, characterized in that, The method also includes: using an error consequence propagation simulator to process operational error data and system parameters, and predicting system state changes caused by erroneous operations; the error consequence propagation simulator consists of a state characterization module, a disturbance propagation module, and a state prediction module.

8. The method according to claim 1, characterized in that, The method also includes: using a risk coupling evolution prediction algorithm to analyze historical data and current state to identify potential risk amplification paths; the risk coupling evolution prediction algorithm uses a spatiotemporal graph neural network model and uses Monte Carlo simulation to predict possible future risk evolution paths.

9. The method according to claim 1, characterized in that, Risk prevention and control intervention strategies include system parameter adjustment suggestions and operational behavior guidance suggestions, which are generated by a weighted multi-objective adaptive optimization algorithm. The optimization objective function takes into account the weighted sum of risk coupling index and intervention cost.

10. A system for executing the digital twin-based optimization method for the dismantling of large radioactive tanks as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire multi-condition operating data and operator behavior data of the radioactive tank; The system risk assessment module is used to generate quantitative indicators of system risk status. Operation error identification module, used to identify potential operation error types; The risk coupling analysis module is used to generate risk coupling indices; The risk prevention and control strategy generation module is used to generate risk prevention and control intervention strategies.

Images