Radiation environment laboratory whole-process traceable system and processing method

Abstract

The invention relates to the technical field of process control, in particular to a radiation environment laboratory full-process traceable system and a processing method.According to the radiation environment laboratory full-process traceable system and processing method.By executing timestamp sequence extraction, attribute difference screening and path structure decomposition operation on detector records and shielding cover logs, causal, time sequence and role attachment relation edges between nodes are established; and classifying attribute values to construct a unified mapping set, converting the multi-source heterogeneous data into a flow chart structure with relevance, executing node embedding operation through a graph attention network algorithm, and performing adjacent node feature weighted aggregation and edge weight screening processing based on an embedding result. And calling a Tarjan strong connected component algorithm to analyze a node branch structure, identifying a heterogeneous sub-graph in a flow path by branch index difference, executing priority ranking and task insertion actions in combination with the number of available tokens and lead plate thickness grade parameters, and generating a dynamic scheduling sequence under a multi-task coexistence condition in a control path.

Description

Technical Field

[0001] This invention relates to the field of process control technology, and in particular to a traceable system and processing method for the entire process of a radiation environment laboratory. Background Technology

[0002] The field of process control technology mainly studies multi-step operation processes with clear sequence, state dependencies and parameter adjustment requirements. It uses control algorithms, scheduling logic, state machine modeling, execution feedback mechanisms and other technical means to control the execution sequence, process parameters and node interactions of the process in real time and in a predetermined manner. The goal is to ensure process stability, strong execution consistency, traceability and auditability.

[0003] The purpose of a full-process traceability system for radiation environment laboratories is to solve the problems of opaque processes, unclear responsibilities, missing data, and lack of traceability in traditional radiation experiments. It enables controllable nodes, verifiable data, and traceable responsibilities throughout the entire experimental cycle. It allows for dynamic control over personnel, equipment, experimental tasks, operating procedures, parameter settings, data uploads, and approval status throughout the entire process, thereby achieving the goals of ensuring experimental quality, standardizing processes, and ensuring experimental data compliance.

[0004] Existing process control technologies rely on linear process modeling and static rule-based decision-making to construct state transition paths. Task modeling is mainly achieved through state machines or Boolean conditional transitions, which can only express linear state transitions and lack the ability to model complex structural paths, circular dependencies, or dynamic cross-conditions. This results in the inability to reconstruct process paths in a timely manner in scenarios involving multi-role collaboration or sudden task insertion. Furthermore, the use of fixed priority tables or static resource allocation mechanisms fails to consider the position of tasks in the process structure and the impact of upstream and downstream nodes on the overall execution path. This leads to a mismatch between task scheduling results and process structure, resulting in situations where high-priority nodes are occupied by low-priority resources. There is a lack of explicit dependency pointers between nodes, and paths are only deduced through time sequence, leading to broken behavioral chains or unclear responsibility tracking, making it impossible to trace back. This results in broken chains of processes that are not closed, behaviors are not connected, and data is unusable. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a fully traceable system and processing method for radiation environment laboratories.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a full-process traceability system for radiation environment laboratories includes: Knowledge graph construction module: Based on detector records and shield integrity logs, extract timestamp sequence comparison, screen node attribute differences, decompose event flow paths, construct relation edge weights, classify attribute value mapping, and generate process knowledge graph; Graph embedding module: Based on the process knowledge graph, a graph attention network is used to filter the neighbor weights of detector calibration data, merge and mask the attributes of material detection nodes, reallocate the weights of relation edges, integrate the embedding vector sequence, and generate a node embedding matrix. Scheduling control module: Based on the node embedding matrix, the Tarjan algorithm is used to sort the differences in branch index lengths, verify the number of available tokens, classify lead plate thickness levels, compare priority weights, rearrange the task sequence, insert high-weight tasks first, and generate a scheduling control sequence. The execution history archiving module: Based on the scheduling control sequence, writes a timestamp sequence, records the temperature reading difference of the test chamber, matches the changes in equipment status values, classifies the approval identification number, maps the item index relationship, and obtains the execution history archive; Link tracing module: Based on the execution history file, compare log entry indexes, screen abnormal path nodes, map approval link sets, reconstruct the tracing path order, splice the path chain relationship, and establish a full-process tracing chain.

[0007] As a further aspect of the present invention, the process knowledge graph specifically includes operation event nodes, equipment status nodes, personnel behavior nodes, approval behavior nodes, anomaly record nodes, and time sequence relationships, causal relationships, and role dependency relationships. The node embedding matrix specifically includes a node vector matrix, an adjacency weight matrix, and a structural semantic encoding matrix. The scheduling control sequence includes branch jump instructions, approval sequence instructions, and resource locking instructions. The execution history archive includes a timestamp record set, a temperature reading difference record set, a status change record set, and an approval identifier record set. The full-process traceability chain includes a node sequence chain, a path dependency chain, and an anomaly tracing chain.

[0008] As a further aspect of the present invention, the knowledge graph construction module includes: Time series extraction submodule: Based on detector records and shield integrity logs, it extracts time fields through field parsing, standardizes date formats, sorts timestamps, detects missing intervals, identifies continuous paragraphs, associates node tags, and obtains a fault time series set; Relationship Mapping Submodule: Based on the fault time series set, it splits event nodes by segmented scanning, matches preceding and following nodes, compares attribute values, assigns edge weights, classifies label mappings, reorganizes the structural sequence, and generates a process knowledge graph.

[0009] As a further aspect of the present invention, the map embedding module includes: Neighbor weight filtering submodule: Based on the process knowledge graph, a graph attention network is used to identify data nodes related to detector calibration and extract attribute information, count the number of adjacencies of each node, identify nodes with high connection frequency, filter out low-weight edges, calculate edge weight difference, organize a list of high-weight adjacency edges, and obtain a set of high-weight adjacency structures. Attribute merging submodule: Based on the high-weight adjacency structure set, collect the attribute fields of the shielding material detection node, sort the attribute values ​​of the same type, aggregate the value list, associate the target node identifier, construct the attribute mapping table, and generate attribute fusion feature groups; Vector generation submodule: Based on attribute fusion feature groups, extract node relationship edge weights, sort the node structure sequence, map attribute values ​​to vector dimensions, concatenate feature sequences, correct vector position encoding, and output node embedding matrix.

[0010] As a further aspect of the present invention, the graph attention network first inputs a multi-head attention layer with nodes containing operation events, device status, personnel behavior, approval behavior, and abnormal record nodes, as well as corresponding temporal, causal, and role-dependent edges. Each attention head calculates an attention coefficient based on the features of the current node and the features of neighboring nodes, obtains the edge weight score by scaling the dot product, normalizes the scores of each head, and aggregates the neighbor features by weighting. The aggregated results of each head are then concatenated and a linear transformation is performed to generate the final node embedding. Finally, a list of high-weight neighboring edges is selected based on the edge weight difference value in the embedding matrix, and a set of high-weight neighboring structures is output.

[0011] As a further aspect of the present invention, the scheduling control module includes: Structure reading submodule: Based on the node embedding matrix, the Tarjan algorithm is used to read the node branch index, extract the index length value, calculate the length difference, sort the difference list, filter the significant difference items, summarize the branch node set, and obtain the index sorting result set; Resource Judgment Submodule: Based on the index sorting result set, load token allocation data, calculate the token balance of each node, compare lead plate thickness levels, classify task priority marking, construct resource tag table, and generate resource scheduling tag set; Sequence construction submodule: Based on the resource scheduling tag set, read the task instruction list, adjust the task execution order, insert high-priority task nodes, update the jump instruction sequence, verify the integrity of the closed-loop index, and output the scheduling control sequence.

[0012] As a further aspect of the present invention, the Tarjan algorithm first constructs a directed flowchart based on the node embedding matrix, initializes the index stack and index numbers, and then performs a depth-first search starting from each unvisited node. During the visit, a unique index number and low link value are assigned to each node. When the traversal path backtracks to a certain node and the low link value is equal to the index value, it indicates that a strongly connected component has been found. All nodes in this component are immediately popped from the stack and recorded as a branch structure. The process continues until all nodes have been traversed. Finally, a branch index set composed of multiple strongly connected components is output, the length of each branch index is calculated, significantly different branches are extracted, and a sorted result set is generated.

[0013] As a further aspect of the present invention, the resume archiving module includes: Timestamp writing submodule: Based on the scheduling control sequence, read the sequence time field, parse the timestamp value, sort the time point list, detect missing time periods, fill in the missing node time, save the corrected list, and obtain the timestamp record set; The history mapping submodule loads the test chamber temperature readings based on the timestamp record set, calculates the difference between adjacent readings, matches the equipment status identifier, classifies the approval number, maps the item index relationship, summarizes the node status changes, and obtains the execution history file.

[0014] As a further aspect of the present invention, the link tracing module includes: Index comparison submodule: Based on the execution history file, extract the log entry index, compare index differences, filter abnormal index items, mark the abnormal node list, and obtain the abnormal node set; The path construction submodule loads the approval link mapping table based on the abnormal node set, matches the node approval relationship, sorts the node access order, merges the path chain, splices the node sequence list, and establishes a full-process traceability chain.

[0015] A method for full-process traceability of radiation environment laboratory processing, wherein the method is executed based on the aforementioned full-process traceability system for radiation environment laboratories, and includes the following steps: S1: Extract the timestamps of the detector calibration record and the shielding cover log and compare the differences, screen the attribute field differences and sort them, decompose the event path and count the edge weights and attribute mappings, and generate a knowledge graph framework. S2: Based on the knowledge graph framework, call the graph attention network, calculate the attention weights between nodes, aggregate adjacency features and merge attribute values, concatenate vectors by node index and correct positions to generate a node embedding matrix; S3: Based on the node embedding matrix, the Tarjan strongly connected component algorithm is used to identify the branch structure and calculate the index length. The token balance and lead plate level are compared to generate priority markers. The task sequence is rearranged and high-weight nodes are inserted. The scheduling control sequence is output. S4: Write the timestamp and temperature difference value according to the scheduling control sequence, match the equipment status change and approval number, combine the mapping entry index, and establish an execution history file; S5: Based on the execution history file, compare the abnormal path nodes with the approval link index, reconstruct the path order and splice the chain relationship to establish the traceability link relationship.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by performing timestamp sequence extraction, attribute difference screening, and path structure decomposition operations on detector records and shielding logs, causal, temporal, and role dependency relationships between nodes are established, and attribute values ​​are classified to construct a unified mapping set, transforming multi-source heterogeneous data into a related flowchart structure. Node embedding operations are performed through a graph attention network algorithm, and the structural encoding and explicit weight expression of contextual relationships are achieved through adjacent node feature weighted aggregation and edge weight filtering. In this invention, based on the embedding results, the Tarjan strongly connected component algorithm is called to parse the node branch structure. The heterogeneous subgraphs in the process path are identified by the branch index differences. The priority sorting and task insertion actions are performed in conjunction with the available token quantity and lead plate thickness level parameters to generate a dynamic scheduling sequence under the condition of multiple tasks coexisting in the control path. In this invention, a time series record is established based on the task control path. The equipment status change, approval number and experimental parameter value are integrated to form a multi-factor node entry. The entry order is mapped with the index as a reference. Based on the abnormal node identifier and approval link relationship in the resume file, the behavior chain is spliced ​​to form a traceable structural path. Attached Figure Description

[0017] Figure 1 This is a system flowchart of the present invention; Figure 2 This is a schematic diagram of the method steps of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0020] Example 1 Please see Figure 1 This invention provides a technical solution: a full-process traceability system for radiation environment laboratories, comprising: Knowledge graph construction module: Based on detector records and shield integrity logs, extract timestamp sequence comparison, screen node attribute differences, decompose event flow paths, construct relation edge weights, classify attribute value mapping, and generate process knowledge graph; Graph embedding module: Based on process knowledge graph, it adopts graph attention network to filter the neighbor weights of detector calibration data, merge and mask material detection node attributes, reallocate relation edge weights, integrate embedding vector sequence, and generate node embedding matrix; Scheduling control module: Based on the node embedding matrix, using the Tarjan algorithm, sorting the differences in branch index lengths, verifying the number of available tokens, classifying lead plate thickness levels, comparing priority weights, rearranging the task sequence, inserting high-weight tasks first, and generating the scheduling control sequence; The history archiving module: Based on the scheduling control sequence, writes the timestamp sequence, records the temperature reading difference of the test chamber, matches the changes in equipment status values, classifies the approval identification number, maps the item index relationship, and obtains the execution history archive; Link tracing module: Based on the execution history file, compare log entry index, screen abnormal path nodes, map approval link set, reconstruct the traceability path order, splice the path chain relationship, and establish a full-process traceability chain.

[0021] The process knowledge graph specifically includes operation event nodes, equipment status nodes, personnel behavior nodes, approval behavior nodes, anomaly record nodes, and time sequence relationships, causal relationships, and role dependency relationships. The node embedding matrix specifically includes a node vector matrix, an adjacency weight matrix, and a structural semantic encoding matrix. The scheduling and control sequence includes branch jump instructions, approval sequence instructions, and resource locking instructions. The execution history archive includes a timestamp record set, a temperature reading difference record set, a status change record set, and an approval identifier record set. The full-process traceability chain includes a node sequence chain, a path dependency chain, and an anomaly tracing chain.

[0022] The knowledge graph construction module includes: Time series extraction submodule: Based on detector records and shield integrity logs, it extracts time fields through field parsing, standardizes date formats, sorts timestamps, detects missing intervals, identifies continuous paragraphs, associates node tags, and obtains a fault time series set; The relation mapping submodule: Based on the fault time series set, it splits event nodes by segmented scanning, matches preceding and following nodes, compares attribute values, assigns edge weights, classifies label mappings, reorganizes the structural sequence, and generates a process knowledge graph; Time series extraction submodule: Based on detector records and shield integrity logs, it uses a time field standardization method to perform format recognition operations on the time fields in the records, uniformly converting them into standardized date formats, sorting all timestamp values, performing interval comparison operations based on the difference between adjacent timestamps, determining whether they exceed a set threshold, marking the start and end positions of missing intervals, further identifying continuous time segments, associating each segment with its corresponding node label, generating a fault time series set, and improving the continuity and accuracy of time flow information extraction; The relation mapping submodule, based on the fault time series set, adopts a node relation construction method to split continuous event nodes in time sequence, extracts node information before and after each time period, performs node attribute value comparison operation, assigns initial weight values ​​to relation edges according to the degree of difference, normalizes all weights, generates numerical attributes of directed edges, performs label classification operation on each node pair, merges label sets according to the mapping results and reorganizes structural paths to generate a process knowledge graph, enhancing the clarity of node evolution structure and structural traceability.

[0023] The map embedding module includes: The neighbor weight filtering submodule is based on a process knowledge graph and uses a graph attention network to identify data nodes related to detector calibration and extract attribute information. It counts the number of adjacencies of each node, identifies nodes with high connection frequency, filters out low-weight edges, calculates edge weight difference, organizes a list of high-weight adjacency edges, and obtains a set of high-weight adjacency structures. Attribute merging submodule: Based on the high-weight adjacency structure set, collect the attribute fields of the shielding material detection node, sort the attribute values ​​of the same type, aggregate the value list, associate the target node identifier, construct the attribute mapping table, and generate attribute fusion feature groups; The vector generation submodule extracts node relationship edge weights based on attribute fusion feature groups, sorts the node structure sequence, maps attribute values ​​to vector dimensions, concatenates feature sequences, corrects vector position encoding, and outputs a node embedding matrix. The neighbor weight selection submodule, based on the process knowledge graph, uses a graph attention network algorithm to calculate the attention weights for the connection relationships between each node and its neighboring nodes in the node set. The embedding dimension is set to 64, and the number of attention heads is 4. It iterates through the neighbor set of each node, performs a dot product operation on the node features and the target node features respectively, adds a linear bias term, and normalizes the result to an attention weight value. All edges are sorted by weight, and edges below a set threshold are removed. The number of neighboring edges retained for each node is counted, and the indices of nodes with the highest connection frequency are extracted. The weights of the retained edges are organized, and a weight difference sequence is calculated to generate a high-weight neighbor structure set, enhancing the ability to distinguish structural features. The attribute merging submodule, based on a high-weight adjacency structure set, uses an attribute value aggregation method to extract the material type, detection value, and node label fields from the masked material detection nodes. It groups and merges the detection values ​​in nodes of the same material type, removes duplicate detection values, and sorts them in ascending order of numerical value. It then binds each merging result with the target node index to a label, constructs a mapping table between the node index and the detection value list, and generates attribute fusion feature groups to enhance the consistency of node attribute representation. The vector generation submodule, based on attribute fusion feature groups, uses a vector concatenation method to extract the attribute value vector and structure index vector corresponding to each node. These vectors are arranged in ascending order by node number, and two vectors are concatenated at corresponding positions to form a single feature sequence. The order of the concatenated feature dimensions is adjusted, and the position index dimensions are aligned and corrected to construct a node vector matrix with unified dimensions. The node embedding matrix is ​​then output, enhancing the spatial representation capability of the nodes.

[0024] The graph attention network first inputs nodes containing operation events, device status, personnel behavior, approval behavior, and abnormal record nodes, along with corresponding temporal, causal, and role-dependent edges, into a multi-head attention layer. Each attention head calculates attention coefficients based on the features of the current node and its neighboring nodes. Edge weight scores are obtained by scaling the dot product. Then, the scores of each head are normalized and weighted to aggregate neighbor features. The aggregated results of each head are concatenated and a linear transformation is used to generate the final node embedding. Finally, a list of high-weight neighboring edges is selected based on the edge weight difference in the embedding matrix, and a set of high-weight neighboring structures is output. Graph attention network, according to the formula: ;

[0025] in: This indicates an enhanced attention coefficient. This represents the normalized attention component weights. This indicates the attention temperature parameter. This represents the transpose of the attention parameter vector. Represents the linear mapping matrix of node features. Represents a node eigenvectors, Represents a node eigenvectors, Indicates the weight of edge interaction frequency. Represents a node With nodes Frequency of interaction between them Indicates time consistency weight, This represents the time series consistency coefficient between nodes. Represents spatial proximity weights. Represents the spatial proximity coefficient between nodes; Execution process: First, a multi-node knowledge graph structure based on the laboratory workflow is constructed, and attribute feature vectors of all nodes in the graph are extracted. and And through a linear mapping matrix The results were obtained by conversion respectively. and Then, the splicing operation is performed to form Then, with the transpose of the attention parameter vector Perform inner product calculations, then introduce temperature parameters. The value is scaled and adjusted, then input into LeakyReLU(·) to obtain the normalized attention components, and then the nodes are counted. With nodes Frequency of interactions between them in the flowchart Calculate the time series consistency coefficient between the two during the calibration process. And to evaluate the proximity relationships in the spatial layout of nodes to generate spatial proximity coefficients. Multiply the above four indicators by their respective weighting coefficients. Post-addition generates enhanced attention coefficient Finally, the grid search method is used to find the target value within the specified range. For each weighting coefficient and temperature parameter Combinatorial optimization is performed, and the impact of attention differences on the screening performance of high-weight adjacency structures is evaluated using a validation set. The optimal parameter combination is then determined for the final output of the high-weight adjacency structure set.

[0026] The scheduling and control module includes: Structure reading submodule: Based on the node embedding matrix, using the Tarjan algorithm, it reads the node branch index, extracts the index length value, calculates the length difference, sorts the difference list, filters the items with significant differences, summarizes the branch node set, and obtains the index sorting result set; Resource Judgment Submodule: Based on the index sorting result set, load token allocation data, calculate the token balance of each node, compare lead plate thickness levels, classify task priority marking, construct resource tag table, and generate resource scheduling tag set; The sequence construction submodule reads the task instruction list based on the resource scheduling tag set, adjusts the task execution order, inserts high-priority task nodes, updates the jump instruction sequence, verifies the integrity of the closed-loop index, and outputs the scheduling control sequence. The structure reading submodule, based on the node embedding matrix, uses the Tarjan strongly connected component algorithm to perform a depth-first traversal operation on each node in the graph, recording the access order number and the lowest reachable number of each node. During the backtracking phase, it determines whether the current node is the starting point of a strongly connected subgraph. If so, it forms a branch set by all unpopped nodes in the current stack, traverses all strongly connected subgraphs to calculate the node index length, calculates the numerical difference between any two sets of index lengths, sorts them in descending order by the difference value, selects the items with the largest difference value, extracts the corresponding branch node numbers, and generates an index sorting result set to enhance the salience of the branch evolution path in the process structure. Resource Judgment Submodule: Based on the indexed sorting result set, the analytic hierarchy process is used to construct a judgment matrix by using task urgency, equipment load value, and historical failure rate value as evaluation factors. Factor weight scoring is calculated for each node. The token allocation data is multiplied based on the weight values ​​to determine the order of token reserves for each node. Lead plate thickness level is set as a priority correction reference. Node priority grouping is performed and a classification label table is generated. All node labels are summarized to generate a resource scheduling label set, enhancing the integrity of the association between task sorting criteria and resource marking. The sequence construction submodule, based on the resource scheduling tag set, adopts a preemptive weighted shortest processing time first scheduling algorithm. It traverses the task instruction list, reads the task execution weight and expected processing time data, calculates the sorting weight ratio, and then sorts the tasks in descending order. It checks whether there is a candidate task with a higher weight than the current task in the current scheduling window. If so, it inserts the task and updates subsequent jump instructions. After rearranging the jump paths of all tasks, it compares the start and end point index values ​​and verifies whether there are discontinuous segments, completes the closed-loop path verification, and outputs the scheduling control sequence, which enhances the priority of high-priority task responses and the integrity of scheduling paths in the instruction chain.

[0027] The Tarjan algorithm first constructs a directed flowchart based on the node embedding matrix, initializes the index stack and index numbers, and then performs a depth-first search starting from each unvisited node. During the visit, each node is assigned a unique index number and a low link value. When the traversal path backtracks to a certain node and the low link value is equal to the index value, it indicates that a strongly connected component has been found. All nodes in this component are immediately popped from the stack and recorded as a branch structure. The process continues until all nodes have been traversed. Finally, a branch index set composed of multiple strongly connected components is output. The length of each branch index is calculated, significantly different branches are extracted, and a sorted result set is generated. Tarjan's algorithm, according to the formula: ;

[0028] in: This represents the index difference value between the i-th branch and the j-th branch in the full-process traceability system of the radiation environment laboratory. This represents the weighting coefficient of index length difference in the total difference calculation. This represents the weighting coefficient of key node density differences in the overall difference calculation. This represents the length of the index sequence corresponding to the i-th branch. This represents the length of the index sequence corresponding to the j-th branch. This represents the number of nodes identified as critical nodes in the i-th branch. This represents the number of critical nodes in the j-th branch. This represents the total number of nodes in the i-th branch. This represents the total number of nodes in the j-th branch; Execution process: First, based on the set of strongly connected components identified in the laboratory traceability data graph, the index sequence of each branch is extracted and its length is calculated. At the same time, count the total number of nodes contained in each branch. Then, based on preset judgment criteria such as node degree centrality threshold, the key nodes in each branch are determined and the number of key nodes is counted. Calculate the density value of key nodes in each branch. For any branch pair First, calculate the index length difference. Then calculate the density difference of key nodes. Next, the two types of differences are multiplied by their respective weighting coefficients. and And sum to generate index difference values The weighting coefficients were determined through grid search combined with cross-validation, and the search range was set to... And select the parameter combination with the best clustering effect, and finally all The sorting results generate a list of differences for system source analysis.

[0029] The resume archiving module includes: The timestamp writing submodule reads the sequence time field based on the scheduling control sequence, parses the timestamp value, sorts the time point list, detects missing time periods, fills in the missing node time, saves the corrected list, and obtains the timestamp record set. The history mapping submodule loads test chamber temperature readings based on timestamp record sets, calculates the difference between adjacent readings, matches equipment status identifiers, classifies approval numbers, maps entry index relationships, summarizes node status changes, and obtains execution history files. The timestamp writing submodule, based on the scheduling control sequence, uses a linear interpolation method to read the time field and convert it into a unified format timestamp value. It sorts all time points in ascending order, compares the interval between two adjacent time points, and determines whether the interval exceeds a set threshold of 300 seconds. If it does, it generates a list of time points with equal intervals based on the start and end times, inserts the time points generated in the missing time period into the original sequence, re-sorts the completed time sequence in ascending order and overwrites the original time field, and generates a timestamp record set to enhance the continuity and integrity of the event scheduling execution sequence. The history mapping submodule, based on a timestamp record set, uses a dynamic time warping method to read the test chamber temperature readings corresponding to each time point and calculate the numerical differences between adjacent readings. It loads the equipment status identifier value and approval number information for the corresponding time point, performs multi-field combination matching according to the time point order, associates reading changes, status identifiers, and approval numbers into unified entry content, maps them to the specified index field position, builds the correspondence between the index and the entry, completes node status updates and time path alignment, generates an execution history archive, and enhances the temporal structure consistency of task process traceability information.

[0030] The link tracing module includes: Index comparison submodule: Based on the execution history archive, extract the log entry index, compare index differences, filter abnormal index items, mark the abnormal node list, and obtain the abnormal node set; The path construction submodule loads the approval link mapping table based on the abnormal node set, matches the node approval relationship, sorts the node access order, merges the path chain, splices the node sequence list, and establishes a full-process traceability chain. Index Comparison Submodule: Based on the execution history archive, the Levenshtein distance algorithm is used to perform pairwise log entry index sequence comparison operations. The number of rows and columns of the matrix is ​​initialized to the length of the sequence to be compared plus one. The rows and columns are traversed sequentially. For the same index value, the matrix cell is set to the value of the top left cell. If the index values ​​are different, the minimum value of the left cell and the top cell plus one is calculated and assigned to the current cell. After filling row by row, the matrix is ​​backtracked from the bottom right corner. The aligned and unaligned parts are identified by the backtracking path. Unaligned index items are marked as abnormal and entered into the abnormal node list. The duplicates are removed and sorted into an abnormal node set. The path construction submodule uses the Kahn topology sorting algorithm based on the abnormal node set to perform node in-degree statistics. It adds all abnormal nodes and related nodes in the approval link mapping table as initial nodes to the queue, pops nodes from the queue and appends them to the access sequence, decrements the in-degree value of all target nodes pointing to the node in the mapping table by one, and adds them to the queue if the in-degree is zero. The loop continues until the queue is exhausted. Finally, the access sequence is concatenated into a node sequence list according to the generation order, and the entire process traceability chain is output.

[0031] Please see Figure 2 A traceable treatment method for radiation environment laboratories, based on the aforementioned traceable system for radiation environment laboratories, includes the following steps: S1: Extract the timestamps of the detector calibration record and the shielding cover log and compare the differences, screen the attribute field differences and sort them, decompose the event path and count the edge weights and attribute mappings, and generate a knowledge graph framework. S2: Based on the knowledge graph framework, call the graph attention network to calculate the attention weights between nodes, aggregate the adjacency features and merge the attribute values, concatenate the vectors by node index and correct the position to generate a node embedding matrix; S3: Based on the node embedding matrix, the Tarjan strongly connected component algorithm is used to identify the branch structure and calculate the index length. The token balance and lead plate level are compared to generate priority tags. The task sequence is rearranged and high-weight nodes are inserted. The scheduling control sequence is output. S4: Write timestamps and temperature differences according to the scheduling control sequence, match equipment status changes and approval numbers, combine mapping entry indexes, and establish execution history archives; S5: Based on the execution history file, compare the abnormal path nodes with the approval link index, reconstruct the path sequence and splice the chain relationship to establish the traceability link relationship.

[0032] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A full-process traceability system for radiation environment laboratories, characterized in that, The system includes: Knowledge graph construction module: Based on detector records and shield integrity logs, extract timestamp sequence comparison, screen node attribute differences, decompose event flow paths, construct relation edge weights, classify attribute value mapping, and generate process knowledge graph; Graph embedding module: Based on the process knowledge graph, a graph attention network is used to filter the neighbor weights of detector calibration data, merge and mask the attributes of material detection nodes, reallocate the weights of relation edges, integrate the embedding vector sequence, and generate a node embedding matrix. Scheduling control module: Based on the node embedding matrix, the Tarjan algorithm is used to sort the differences in branch index lengths, verify the number of available tokens, classify lead plate thickness levels, compare priority weights, rearrange the task sequence, insert high-weight tasks first, and generate a scheduling control sequence. The execution history archiving module: Based on the scheduling control sequence, writes a timestamp sequence, records the temperature reading difference of the test chamber, matches the changes in equipment status values, classifies the approval identification number, maps the item index relationship, and obtains the execution history archive; Link tracing module: Based on the execution history file, compare log entry indexes, screen abnormal path nodes, map approval link sets, reconstruct the tracing path order, splice the path chain relationship, and establish a full-process tracing chain.

2. The full-process traceability system for radiation environment laboratories according to claim 1, characterized in that, The process knowledge graph specifically includes operation event nodes, equipment status nodes, personnel behavior nodes, approval behavior nodes, anomaly record nodes, and time sequence relationships, causal relationships, and role dependency relationships. The node embedding matrix specifically includes a node vector matrix, an adjacency weight matrix, and a structural semantic encoding matrix. The scheduling control sequence includes branch jump instructions, approval sequence instructions, and resource locking instructions. The execution history archive includes a timestamp record set, a temperature reading difference record set, a status change record set, and an approval identifier record set. The full-process traceability chain includes a node sequence chain, a path dependency chain, and an anomaly tracing chain.

3. The full-process traceability system for radiation environment laboratories according to claim 1, characterized in that, The knowledge graph construction module includes: Time series extraction submodule: Based on detector records and shield integrity logs, it extracts time fields through field parsing, standardizes date formats, sorts timestamps, detects missing intervals, identifies continuous paragraphs, associates node tags, and obtains a fault time series set; Relationship Mapping Submodule: Based on the fault time series set, it splits event nodes by segmented scanning, matches preceding and following nodes, compares attribute values, assigns edge weights, classifies label mappings, reorganizes the structural sequence, and generates a process knowledge graph.

4. The full-process traceability system for radiation environment laboratories according to claim 1, characterized in that, The map embedding module includes: Neighbor weight filtering submodule: Based on the process knowledge graph, a graph attention network is used to identify data nodes related to detector calibration and extract attribute information, count the number of adjacencies of each node, identify nodes with high connection frequency, filter out low-weight edges, calculate edge weight difference, organize a list of high-weight adjacency edges, and obtain a set of high-weight adjacency structures. Attribute merging submodule: Based on the high-weight adjacency structure set, collect the attribute fields of the shielding material detection node, sort the attribute values ​​of the same type, aggregate the value list, associate the target node identifier, construct the attribute mapping table, and generate attribute fusion feature groups; Vector generation submodule: Based on attribute fusion feature groups, extract node relationship edge weights, sort the node structure sequence, map attribute values ​​to vector dimensions, concatenate feature sequences, correct vector position encoding, and output node embedding matrix.

5. The full-process traceability system for radiation environment laboratories according to claim 4, characterized in that, The graph attention network first inputs nodes containing operation events, device status, personnel behavior, approval behavior, and abnormal record nodes, along with corresponding temporal, causal, and role-dependent edges, into a multi-head attention layer. Each attention head calculates attention coefficients based on the features of the current node and its neighboring nodes, obtains edge weight scores by scaling the dot product, normalizes the scores of each head, and aggregates neighbor features with weights. The aggregated results of each head are then concatenated and linearly transformed to generate the final node embedding. Finally, a list of high-weight neighboring edges is selected based on the edge weight difference in the embedding matrix, and a set of high-weight neighboring structures is output.

6. The full-process traceability system for radiation environment laboratories according to claim 1, characterized in that, The scheduling control module includes: Structure reading submodule: Based on the node embedding matrix, the Tarjan algorithm is used to read the node branch index, extract the index length value, calculate the length difference, sort the difference list, filter the significant difference items, summarize the branch node set, and obtain the index sorting result set; Resource Judgment Submodule: Based on the index sorting result set, load token allocation data, calculate the token balance of each node, compare lead plate thickness levels, classify task priority marking, construct resource tag table, and generate resource scheduling tag set; Sequence construction submodule: Based on the resource scheduling tag set, read the task instruction list, adjust the task execution order, insert high-priority task nodes, update the jump instruction sequence, verify the integrity of the closed-loop index, and output the scheduling control sequence.

7. The full-process traceability system for radiation environment laboratories according to claim 6, characterized in that, The Tarjan algorithm first constructs a directed flowchart based on the node embedding matrix, initializes the index stack and index numbers, and then performs a depth-first search starting from each unvisited node. During the visit, each node is assigned a unique index number and a low-link value. When the traversal path backtracks to a certain node and the low-link value equals the index value, it indicates that a strongly connected component has been found. All nodes in this component are immediately popped from the stack and recorded as a branch structure. This process continues until all nodes have been traversed. Finally, a branch index set composed of multiple strongly connected components is output. The length of each branch index is calculated, significantly different branches are extracted, and a sorted result set is generated.

8. The full-process traceability system for radiation environment laboratories according to claim 1, characterized in that, The resume archiving module includes: Timestamp writing submodule: Based on the scheduling control sequence, read the sequence time field, parse the timestamp value, sort the time point list, detect missing time periods, fill in the missing node time, save the corrected list, and obtain the timestamp record set; The history mapping submodule loads the test chamber temperature readings based on the timestamp record set, calculates the difference between adjacent readings, matches the equipment status identifier, classifies the approval number, maps the item index relationship, summarizes the node status changes, and obtains the execution history file.

9. The full-process traceability system for radiation environment laboratories according to claim 1, characterized in that, The link tracing module includes: Index comparison submodule: Based on the execution history file, extract the log entry index, compare index differences, filter abnormal index items, mark the abnormal node list, and obtain the abnormal node set; The path construction submodule loads the approval link mapping table based on the abnormal node set, matches the node approval relationship, sorts the node access order, merges the path chain, splices the node sequence list, and establishes a full-process traceability chain.

10. A fully traceable treatment method for radiation environment in laboratories, characterized in that, The full-process traceability system for radiation environment laboratories according to any one of claims 1-9 includes the following steps: S1: Extract the timestamps of the detector calibration record and the shielding cover log and compare the differences, screen the attribute field differences and sort them, decompose the event path and count the edge weights and attribute mappings, and generate a knowledge graph framework. S2: Based on the knowledge graph framework, call the graph attention network, calculate the attention weights between nodes, aggregate adjacency features and merge attribute values, concatenate vectors by node index and correct positions to generate a node embedding matrix; S3: Based on the node embedding matrix, the Tarjan strongly connected component algorithm is used to identify the branch structure and calculate the index length. The token balance and lead plate level are compared to generate priority markers. The task sequence is rearranged and high-weight nodes are inserted. The scheduling control sequence is output. S4: Write the timestamp and temperature difference value according to the scheduling control sequence, match the equipment status change and approval number, combine the mapping entry index, and establish an execution history file; S5: Based on the execution history file, compare the abnormal path nodes with the approval link index, reconstruct the path order and splice the chain relationship to establish the traceability link relationship.

Images