Intelligent ship general assembly construction data service calling link tracking method

Abstract

The invention relates to the technical field of ship construction, and discloses an intelligent ship final assembly construction data service calling link tracking method. The method comprises the steps of obtaining a ship general assembly building full life cycle data set, generating a service node topological graph and a load balancing path, performing anomaly detection processing on a task execution result, generating a link tracking feature matrix, further generating a service calling link optimization strategy, and outputting a standardized service calling link. The system comprises a full life cycle data acquisition module, a knowledge graph construction module, a path optimization module, an anomaly detection module, a multi-dimensional tracking module, a link generation module and the like. According to the invention, data can be effectively integrated, a service calling link can be optimized, load balancing and accurate anomaly detection can be realized, the efficiency, quality and stability of shipbuilding can be improved, the construction cost can be reduced, and the requirements of the shipbuilding industry on digital and intelligent management can be met.

Description

Technical Field

[0001] The present invention relates to the technical field of shipbuilding, and particularly to an intelligent method for tracking the data service call link in ship general assembly construction. Background Art

[0002] During the process of ship general assembly construction, the management and tracking of the data service call link face many severe challenges, which seriously restrict the efficiency, quality, and cost control of shipbuilding.

[0003] From the perspective of data management, ship general assembly construction involves numerous devices and complex processes, and the data such as equipment status parameters, service call logs, and task execution results generated are large in scale, complex in structure, and interrelated. Traditional data management methods cannot efficiently integrate and analyze this data, resulting in difficulty in fully exploiting the data value. For example, the data formats and interfaces provided by different equipment suppliers are different, and compatibility problems are likely to occur during data aggregation, making it difficult to ensure the consistency and accuracy of the data, and bringing great difficulties to subsequent analysis and decision-making.

[0004] In terms of the service call link, with the increasing digitization of shipbuilding, the call relationships between various systems and services are becoming more and more complex. Due to the lack of effective link tracking means, it is difficult to accurately locate the fault points and performance bottlenecks in the service call process. When an exception occurs in a certain service, it is impossible to quickly determine whether it is a problem of the service itself or a chain reaction caused by other links in the upstream and downstream service call links. This will not only prolong the fault troubleshooting time but also may cause delays in the entire construction progress and increase unnecessary cost expenditures.

[0005] In terms of load balancing, existing load balancing strategies often cannot be dynamically adjusted according to the real-time service call situation. When some service nodes are overloaded while other nodes are idle, the load cannot be reasonably distributed to the idle nodes in time, resulting in a decline in the overall system performance. Especially during the peak period of shipbuilding, when a large number of tasks are executed concurrently, this load imbalance problem will be more prominent, seriously affecting the construction efficiency.

[0006] In addition, for the anomaly detection of task execution results, traditional methods mainly rely on manual experience or simple threshold judgment, with poor accuracy and timeliness. They cannot comprehensively and deeply analyze complex anomaly situations, are prone to missing potential risk points, and are difficult to detect and solve problems at an early stage, thus having an adverse impact on the quality of shipbuilding.

[0007] With the development of the global shipping industry, higher requirements are put forward for the quality, efficiency, and cost control of shipbuilding. The traditional management method of the data service call link in ship general assembly construction can no longer meet the needs of the industry's development. There is an urgent need for an intelligent method and system to optimize data management, track service call links, achieve efficient load balancing, and accurate anomaly detection, so as to enhance the overall competitiveness of shipbuilding. Summary of the Invention

[0008] The purpose of the present invention is to provide an intelligent method for tracking the data service call link in ship general assembly construction to solve the problems put forward in the above background technology.

[0009] To achieve the above purpose, the present invention provides the following technical solution: An intelligent method for tracking the data service call link in ship general assembly construction, the method includes: Obtain the full life cycle data set of ship general assembly construction, and the data set includes equipment status parameters, service call logs, and task execution results; Based on the equipment status parameters, generate a service node topology graph through a dynamic knowledge graph construction algorithm, and the topology graph includes equipment association relationships and service call path weights; According to the service call logs, generate a load balancing path through a service path optimization algorithm, and eliminate redundant call nodes; Perform anomaly detection processing on the task execution results to generate a service call anomaly marking sequence; Input the service node topology graph, load balancing path, and anomaly marking sequence into a multi-dimensional tracking model to generate a link tracking feature matrix; Based on the link tracking feature matrix, generate a service call link optimization strategy through an adaptive tracking network construction algorithm, and output a standardized service call link; the nodes of the adaptive tracking network represent service call modules, and the edges represent call priorities and stability weights.

[0010] Preferably, the generation of the service node topology graph through the dynamic knowledge graph construction algorithm includes: Perform entity relationship modeling on the equipment status parameters, and extract the collaborative operation characteristics between equipment; Based on the graph embedding algorithm, map the equipment entities into low-dimensional vectors to generate an initial knowledge graph; Use the hierarchical clustering algorithm to group the equipment entities and calculate the connection strength between groups; Encode the low-dimensional vectors, connection strength between groups, and service call path weights into a service node topology graph.

[0011] Preferably, the generation of the load balancing path through the service path optimization algorithm includes: Standardize and clean the service call logs to eliminate invalid call records; Identify high-frequency service nodes based on the dynamic load balancing algorithm, calculate the response time deviation between nodes, and generate a load balancing path according to the response time deviation and the preset stability threshold; Dynamically eliminate redundant call nodes through a distributed hash table and update the load balancing path.

[0012] Preferably, the anomaly detection and processing includes: Segment the task execution results by time series, and extract the task completion indicators for each stage; Use the isolation forest algorithm to detect abnormal segments, generate preliminary anomaly marks, correct the preliminary anomaly marks based on the dynamic threshold adjustment algorithm, and eliminate misdetected marks; Encode the corrected anomaly marks into an anomaly mark sequence in chronological order.

[0013] Preferably, the multi-dimensional tracking model includes a feature alignment module and a matrix generation module, and the feature alignment module includes: Normalize the device association relationships in the service node topology map to obtain a first alignment vector; discretize and map the call priorities in the load balancing path to generate a second alignment vector; perform window sliding statistics on the anomaly mark sequence to extract anomaly frequency features and obtain a third alignment vector; Fuse the first alignment vector, the second alignment vector, and the third alignment vector into a link tracking feature matrix through a tensor decomposition algorithm.

[0014] Preferably, the matrix generation module includes: Reduce the dimension of the link tracking feature matrix to generate a core feature subspace; Extract the topological dependence features between nodes through a graph neural network to generate a local association graph; Perform feature splicing on the core feature subspace and the local association graph to generate a global tracking matrix; Map the global tracking matrix to a standardized service call link through a residual connection layer.

[0015] Preferably, the update method of the dynamic knowledge graph construction algorithm includes: Divide the device state parameters based on a sliding time window and extract the change amount of entity relationships within the window; Update the device association relationships and path weights in the knowledge graph through an incremental learning algorithm, and dynamically adjust the structure of the service node topology map according to the updated knowledge graph.

[0016] Preferably, the parameter adjustment method of the distributed hash table includes: Set the initial number of hash shards according to the number of service call logs; Dynamically adjust the shard distribution through the consistent hashing algorithm to reduce the data migration volume during node expansion; Optimize the hash function parameters based on the shard load monitoring results to improve the efficiency of redundant node removal.

[0017] Preferably, the adaptive tracking network construction algorithm includes: Initialize the node attributes according to the service call module, and generate an edge weight tensor based on the stability weight; Use the link tracking feature matrix as the node input, and the edge weight tensor is composed of the call priority and the stability weight; Iteratively update the path selection strategy of each node through the policy optimization algorithm to optimize the edge weight tensor; Generate the optimal service call link covering all nodes according to the optimized edge weight tensor.

[0018] Preferably, the present invention further includes an intelligent ship general assembly construction data service call link tracking system, and the system includes: Full life cycle data acquisition module: used to acquire the ship general assembly construction full life cycle data set, and the data set includes equipment status parameters, service call logs and task execution results; Knowledge graph construction module: configured to generate a service node topology graph through a dynamic knowledge graph construction algorithm based on the equipment status parameters; Path optimization module: used to generate a load balancing path through a service path optimization algorithm according to the service call logs; Abnormality detection module: perform abnormality detection processing on the task execution results to generate a service call abnormality marking sequence; Multi-dimensional tracking module: input the service node topology graph, the load balancing path and the abnormality marking sequence into a multi-dimensional tracking model to generate a link tracking feature matrix; Link generation module: used to generate a service call link optimization strategy based on the link tracking feature matrix through an adaptive tracking network construction algorithm, and output a standardized service call link.

[0019] Compared with the prior art, the beneficial effects of the present invention are: In terms of data integration and analysis, by obtaining the dataset of the entire life cycle of ship assembly and construction and using the dynamic knowledge graph construction algorithm to generate the topology diagram of service nodes, it is possible to effectively integrate complex equipment status parameters and clearly present the equipment association relationship and service call path weights. This helps engineers comprehensively understand the collaborative relationship between various equipment during the ship construction process and provides strong support for subsequent fault troubleshooting and performance optimization. For example, when a device fails, other related devices can be quickly determined based on the topology diagram, narrowing down the troubleshooting scope and improving the fault resolution efficiency.

[0020] In terms of optimizing the service call link, based on the service call logs, the load balancing path is generated through the service path optimization algorithm, and redundant call nodes are removed, greatly improving the performance and efficiency of the system. The dynamic load balancing algorithm can real-time identify high-frequency service nodes and reasonably allocate loads according to the response time deviation and the preset stability threshold, avoiding the situation where some nodes are overloaded while some nodes are idle. Taking the large amount of data processing tasks during ship construction as an example, the generation of the load balancing path can ensure that each data processing service node can evenly undertake tasks, significantly improving the response speed of the entire system, reducing the task waiting time, and thus accelerating the ship construction progress.

[0021] In terms of anomaly detection and handling, advanced isolation forest algorithm and dynamic threshold adjustment algorithm are used to detect anomalies in the task execution results, generating accurate anomaly marking sequences. This method can timely detect anomalies during task execution and effectively remove false detection marks through dynamic threshold adjustment, improving the accuracy of anomaly detection. In ship construction, timely discovery and handling of anomalies in key tasks can prevent problems from expanding and ensure the quality of ship construction. For example, in the ship structure welding task, if abnormal welding parameters are timely detected through anomaly detection, the welding process can be adjusted in time to avoid welding defects and improve the safety of the ship structure.

[0022] The introduction of the multi-dimensional tracking model integrates the service node topology diagram, the load balancing path, and the anomaly marking sequence to generate a link tracking feature matrix, comprehensively tracking and analyzing the service call link from multiple perspectives. Through the collaborative work of the feature alignment module and the matrix generation module, rich feature information can be extracted, providing more accurate data support for subsequent link optimization.

[0023] Finally, based on the link tracking feature matrix, an adaptive tracking network construction algorithm is used to generate service call link optimization strategies, and standardized service call links are output. These optimization strategies can dynamically adjust service call links according to real-time situations, improving the stability and reliability of the system. During the shipbuilding process, in the face of constantly changing requirements and environments, standardized service call links can ensure smoother collaborative work among all links, reduce the probability of system errors, improve the overall efficiency and quality of shipbuilding, and also help reduce construction costs and enhance the competitiveness of shipbuilding enterprises in the market. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is the working principle diagram of the production scheduling visualization method based on big data according to the present invention; Figure 2 is the working principle diagram of generating a service node topology map by constructing a dynamic knowledge graph; Figure 3 is the flow chart of abnormal detection and processing of task execution results; Figure 4 is the flow chart of the multi-dimensional tracking model matrix generation module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0026] Please refer to Figures 1 - 4 , the present invention provides a technical solution: The present invention relates to an intelligent method for tracking the service call link of ship general assembly construction data, and the specific implementation solution is as follows: During the ship general assembly construction process, through various sensors, monitoring systems, and relevant software records, equipment status parameters, service call logs, and task execution results are collected to form a ship general assembly construction full life cycle data set. For example, equipment status parameters may include real-time data such as the operating temperature, rotation speed, and pressure of the equipment; service call logs record information such as the time, caller, and callee of each service call; task execution results reflect the completion status of each construction task, such as whether it is completed, completion time, and completion quality, etc.

[0027] Based on the obtained equipment status parameters, a service node topology map is generated by using a dynamic knowledge graph construction algorithm. This topology map contains equipment association relationships and service call path weights, which shows the collaborative operation relationships among various equipment during the ship general assembly construction process and the importance degree of service call paths.

[0028] Based on the service call logs, use the service path optimization algorithm to generate the load balancing path. At the same time, identify and eliminate redundant call nodes through the algorithm to improve the efficiency and performance of service calls and reduce the waste of system resources.

[0029] Perform anomaly detection on the task execution results, timely detect anomalies during task execution, and generate a service call anomaly marking sequence from the detection results for subsequent analysis and processing of anomalies.

[0030] Input the service node topology graph, load balancing path, and anomaly marking sequence into the multi-dimensional tracking model. This model processes and fuses these data to generate a link tracking feature matrix, providing data support for subsequent generation of optimization strategies.

[0031] Based on the link tracking feature matrix, use the adaptive tracking network construction algorithm to generate a service call link optimization strategy, and finally output a standardized service call link to optimize the service call link of the ship's total assembly construction data, improving the efficiency and reliability of the entire construction process.

[0032] The technical solution of the present invention will be further described in detail through specific embodiments as follows: Embodiment 1:

[0033] This embodiment focuses on the specific implementation process of generating a service node topology graph through the dynamic knowledge graph construction algorithm. During the ship's total assembly construction process, there are numerous and complex relationships among various types of equipment. Taking a large ship construction project as an example, it involves various types of equipment such as power equipment, welding equipment, and lifting equipment. After obtaining the status parameters of these equipment, entity relationship modeling is first performed. Assuming that equipment A, B, C, etc. are different equipment entities, through the analysis of equipment status parameters, it is found that the temperature change of equipment A during operation will affect the working efficiency of equipment B, indicating that there is a collaborative operation feature between equipment A and equipment B, and this relationship is extracted and recorded.

[0034] Map the equipment entities to low-dimensional vectors based on the graph embedding algorithm. The graph embedding algorithm used here can be the DeepWalk algorithm. Let the equipment entity set be , where e i represents the i-th equipment entity. Through the DeepWalk algorithm, each equipment entity e i is mapped to a low-dimensional vector v i so that the relationship features between equipment entities can be retained in the low-dimensional space, generating an initial knowledge graph.

[0035] Then, use the hierarchical clustering algorithm to group the equipment entities. Let the distance metric function be d(e i, e j ), which represents the device entity e i and e j The distance between. Through the hierarchical clustering algorithm, the device entities are gradually merged into different groups according to the distance metric function . Calculate the connection strength between groups. Assume that the connection strength between group G i and G j is S(G i, G j ), which can be determined by calculating the total number of connections or the sum of connection weights between the device entities in the two groups.

[0036] Finally, encode the low-dimensional vector v i , the inter-group connection strength S(G i, G j ) and the service call path weight w ij (where w ij represents the service call path weight from device to device ) into the service node topology graph. In this way, the association relationship between devices and the weight information of service call paths can be clearly displayed, providing an important basis for subsequent analysis and optimization.

[0037] Example 2:

[0038] In the scenario of ship general assembly construction, the service call log records a large amount of complex data, which is crucial for generating the load balancing path. At the beginning, the service call log needs to be standardized and cleaned. In the actual ship construction process, due to factors such as network fluctuations and system instantaneous failures, many invalid call records will be generated. For example, during a certain period of time, due to a short network interruption, some service call requests failed to be successfully sent to the target server, but left call records in the log, and these records have no corresponding service response data. Such records are invalid records. During cleaning, by checking the response field in the log record, if the response field is empty and the call time is less than the minimum value of the normal response time, it can be determined that the record is invalid and it will be excluded.

[0039] After cleaning, identify high-frequency service nodes based on the dynamic load balancing algorithm. Assume that the service node set is , where n i represents the i-th service node. Through statistical analysis of the cleaned service call log, record the call times C(n i)。Set a high-frequency threshold T, which can be determined according to the data of past construction projects and the actual situation of the current project. For example, in past similar projects, when the number of calls of a certain service node exceeded 100 times per unit time, there would be a performance bottleneck. Considering the scale and complexity of the current project, the high-frequency threshold T is set to 120 times per hour. When the number of calls C(n i of a certain service node n i ) is greater than T, this node is identified as a high-frequency service node.

[0040] Next, calculate the response time deviation between nodes. Let the response times between node n i and n j be R(n i ) and R(n j ) respectively. The response time deviation is calculated by the formula . Here, the response time R(n i ) refers to the time interval from the sending of a service request to the receipt of a service response, which can be accurately obtained from the service call log. Preset a stability threshold S th , for example, according to the system performance requirements and past experience, set S th to 50 milliseconds. When , it means that the load between these two nodes is unbalanced, and the service call path needs to be adjusted to achieve load balancing. For example, if the response time deviation between node n1 and n2 is greater than the stability threshold, some requests originally sent to n1 can be considered to be allocated to n2, or the service resources of n1 can be expanded.

[0041] Finally, dynamically eliminate redundant call nodes through a distributed hash table. A distributed hash table (DHT) is a distributed storage system that stores data on multiple nodes and determines the storage location of data through a hash function. In this embodiment, when the call frequency of a certain call node is extremely low within a period of time and it is found through evaluation that its contribution to the function of the entire service call link is small, this node can be determined as a redundant call node. Utilizing the characteristics of the distributed hash table, according to the unique identifier of the node, calculate the position of the node in the hash table through the hash function h(node_id), and then delete it from the hash table, thereby achieving the dynamic elimination of redundant call nodes and timely updating the load balancing path to ensure that the service call link always maintains an efficient operation state.

[0042] Embodiment 3:

[0043] During the execution of the ship's general assembly construction task, the task execution result is a sequence of data that changes over time. It is of great significance to perform accurate anomaly detection and processing on it. First, segment the task execution result by time series. Assume that the task execution result sequence is , with a fixed time interval Δt as the basis for segmentation. In the ship welding task, with a fixed time interval Δt of 10 minutes, the entire welding task execution process is divided into multiple time periods . Within each time period, the task completion index is extracted.

[0044] After extracting the task completion index, the Isolation Forest algorithm is used to detect abnormal segments. The Isolation Forest algorithm isolates data points by constructing a binary tree. Let the data point set be , where p i represents the task completion index data point for each time period. For each data point p i , calculate its path length l(p i ) in the Isolation Forest. In the Isolation Forest, the shorter the path length l(p i ), the more the data point deviates from the normal data distribution, and the more likely it is to be an abnormal point. Set a path length threshold L th , and this threshold can be obtained through statistical analysis of historical normal task completion data. For example, by constructing an Isolation Forest model for the completion data of a large number of normal welding task time periods, calculating the average and standard deviation of the path lengths of normal data points, and taking the average minus twice the standard deviation as the path length threshold L th . When l(p i ) < L th , mark the time period corresponding to this data point as abnormal and generate a preliminary abnormal mark.

[0045] Since there may be misdetection in the preliminary abnormal mark, it needs to be corrected based on the dynamic threshold adjustment algorithm. Assume the dynamic threshold adjustment factor is α, and this factor can be dynamically adjusted according to the real-time situation of task execution and the change trend of historical data. For example, when it is found that the recent task completion data fluctuates greatly, appropriately increase the value of α to make the threshold more stringent; when the data is relatively stable, decrease the value of α. According to the formula dynamically adjust the path length threshold L th to be , and then re-evaluate the preliminary abnormal mark to eliminate misdetection marks.

[0046] Finally, encode the corrected abnormal marks into an abnormal mark sequence in chronological order. For example, after correction, if abnormalities are detected in time periods and , the abnormal mark sequence can be expressed as , where 1 represents abnormal and 0 represents normal. Such an abnormal mark sequence provides a clear and accurate data basis for in-depth analysis and timely handling of abnormal situations in the ship assembly and construction tasks, and helps to ensure the smooth progress of the entire construction task.

[0047] Example 4:

[0048] This example deeply explores the specific implementation of the feature alignment module and the matrix generation module of the multi-dimensional tracking model. During the process of ship general assembly construction data processing, the equipment association relationship in the service node topology graph is normalized. Let the equipment association relationship matrix be , where a ij represents the association relationship value between equipment i and equipment j (0 ≤ a ij ≤ 1). Through the normalization formula , a ij is normalized to the interval [0, 1] to obtain the first alignment vector.

[0049] The call priorities in the load balancing path are discretely mapped. Assume that the call priority value range is [1, 5], and it is discretized into values between 0 and 1. For example, a priority of 1 is mapped to 0, a priority of 5 is mapped to 1, a priority of 3 is mapped to 0.5, etc., to generate the second alignment vector.

[0050] The window sliding statistics are performed on the abnormal marker sequence to extract the abnormal frequency feature. Let the abnormal marker sequence be , and the window size is w. The number count of abnormal markers with a value of 1 is statistically calculated within each window, and the abnormal frequency , to obtain the third alignment vector.

[0051] The first alignment vector, the second alignment vector, and the third alignment vector are fused into a link tracking feature matrix through the tensor decomposition algorithm. Let the first alignment vector be v1, the second alignment vector be v2, and the third alignment vector be v3. Through the tensor decomposition algorithm , the link tracking feature matrix M is obtained.

[0052] In the matrix generation module, dimensionality reduction processing is performed on the link tracking feature matrix M. Assume that the principal component analysis (PCA) algorithm is used. Let the dimension of the link tracking feature matrix M be d, and it is reduced to d' dimensions through the PCA algorithm to generate the core feature subspace C.

[0053] The topological dependence features between nodes are extracted through a graph neural network to generate a local association graph. Let the input of the graph neural network be the service node topology graph. After the calculation of the graph neural network, the topological dependence features between nodes are obtained, and then the local association graph G is generated.

[0054] The core feature subspace C and the local association graph G are feature-stitched to generate the global tracking matrix T. Through the residual connection layer, the global tracking matrix T is mapped into a standardized service call link to achieve effective tracking and analysis of the service call link.

[0055] Example 5:

[0056] This embodiment details the update method of the dynamic knowledge graph construction algorithm, the parameter adjustment method of the distributed hash table, and the adaptive tracking network construction algorithm. During the overall assembly and construction of a ship, the equipment status parameters change continuously over time. Based on a sliding time window, the equipment status parameters are divided, and the size of the sliding time window is set as ΔT. Within each time window, the change amount of the entity relationship of the equipment status parameters is extracted. For example, at a certain moment, a new device D is installed, and it establishes new collaborative operation relationships with devices A and B, which is the change amount of the entity relationship.

[0057] Update the device association relationships and path weights in the knowledge graph through the incremental learning algorithm. Let the incremental learning algorithm be IL, the knowledge graph be KG, the device association relationship be R, and the path weight be W. According to the change amount of the entity relationship, update the knowledge graph through where ΔR and ΔW represent the change amounts of the device association relationship and the path weight respectively. Dynamically adjust the structure of the service node topology graph according to the updated knowledge graph to ensure that the topology graph can reflect the relationship changes between devices in real time.

[0058] For the parameter adjustment of the distributed hash table, set the initial number of hash shards according to the number of service call logs. Let the number of service call logs be N and the initial number of hash shards be S0. The initial number of hash shards can be determined by the formula (This is just an example formula, and a more appropriate formula can actually be determined based on experience or testing).

[0059] Dynamically adjust the shard distribution through the consistent hashing algorithm to reduce the data migration volume during node expansion. Let the consistent hashing algorithm be CHA and the hash ring be H. When a new node joins or an existing node needs to be expanded, dynamically adjust the hash ring through H = CHA(H, new_node) to achieve a reasonable distribution of shards.

[0060] Optimize the hash function parameters based on the shard load monitoring results to improve the efficiency of redundant node removal. Let the shard load be L i (i represents the shard number), the hash function be h. By monitoring L i , adjust the hash function parameters according to the optimization algorithm to improve the accuracy and efficiency of redundant node removal.

[0061] In the adaptive tracking network construction algorithm, initialize the node attributes according to the service call module and generate an edge weight tensor based on the stability weight. Let the set of service call modules be , each service call module m i corresponds to a node, and the node attribute is attr(m i ). The stability weight is w sij(indicating the stability weight from node i to node j), generate the edge weight tensor T edge .

[0062] Use the link tracking feature matrix as the node input. The edge weight tensor is composed of the call priority and the stability weight. Iteratively update the path selection strategy of each node through the policy optimization algorithm to optimize the edge weight tensor. Let the policy optimization algorithm be POA. After multiple iterations , where input represents input data such as the link tracking feature matrix.

[0063] Generate the optimal service call link covering all nodes according to the optimized edge weight tensor, realize the optimization of the ship assembly construction data service call link, and improve the efficiency and reliability of the entire construction process.

[0064] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0065] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent method for tracking the call link of ship general assembly construction data services, characterized in that, Including: Obtain the dataset of the whole life cycle of ship general assembly construction, where the dataset includes equipment status parameters, service call logs, and task execution results; Based on the equipment status parameters, generate a service node topology graph through a dynamic knowledge graph construction algorithm, where the topology graph contains equipment association relationships and service call path weights; According to the service call logs, generate a load balancing path through a service path optimization algorithm, and eliminate redundant call nodes; Perform anomaly detection processing on the task execution results to generate a service call anomaly marking sequence; Input the service node topology graph, load balancing path, and anomaly marking sequence into a multi-dimensional tracking model to generate a link tracking feature matrix; Based on the link tracking feature matrix, generate a service call link optimization strategy through an adaptive tracking network construction algorithm, and output a standardized service call link; the nodes of the adaptive tracking network represent service call modules, and the edges represent call priorities and stability weights.

2. The intelligent ship general assembly construction data service call link tracking method according to claim 1, wherein, The generation of the service node topology graph through the dynamic knowledge graph construction algorithm includes: Perform entity relationship modeling on the equipment status parameters to extract the collaborative operation characteristics between equipment; Based on the graph embedding algorithm, map equipment entities into low-dimensional vectors to generate an initial knowledge graph; Use the hierarchical clustering algorithm to group equipment entities and calculate the connection strength between groups; Encode the low-dimensional vectors, connection strength between groups, and service call path weights into a service node topology graph.

3. The intelligent ship general assembly construction data service call link tracking method according to claim 1, characterized in that, The generation of the load balancing path through the service path optimization algorithm includes: Perform standardized cleaning on the service call logs to eliminate invalid call records; Based on the dynamic load balancing algorithm, identify high-frequency service nodes, calculate the response time deviation between nodes, and generate a load balancing path according to the response time deviation and a preset stability threshold; Dynamically eliminate redundant call nodes through a distributed hash table and update the load balancing path.

4. The intelligent ship general assembly construction data service call link tracking method according to claim 1, characterized in that The anomaly detection processing includes: Perform time series segmentation on the task execution results and extract the task completion degree indicators for each stage; Use the isolation forest algorithm to detect abnormal segments, generate preliminary anomaly markings, correct the preliminary anomaly markings based on the dynamic threshold adjustment algorithm, and eliminate misdetection markings; Encode the corrected anomaly markings into an anomaly marking sequence in chronological order.

5. The intelligent ship general assembly construction data service call link tracking method according to claim 1, wherein The multi-dimensional tracking model includes a feature alignment module and a matrix generation module. The feature alignment module includes: Normalize the equipment association relationships in the service node topology graph to obtain a first alignment vector; discretely map the call priorities in the load balancing path to generate a second alignment vector; perform window sliding statistics on the anomaly marking sequence to extract anomaly frequency features and obtain a third alignment vector; Fuse the first alignment vector, second alignment vector, and third alignment vector into a link tracking feature matrix through a tensor decomposition algorithm.

6. The intelligent ship general assembly construction data service call link tracking method according to claim 5, wherein, The matrix generation module includes: Perform dimensionality reduction processing on the link tracking feature matrix to generate a core feature subspace; Extract the topological dependence features between nodes through a graph neural network to generate a local association graph; Perform feature splicing on the core feature subspace and the local association graph to generate a global tracking matrix; Map the global tracking matrix to a standardized service call link through a residual connection layer.

7. The intelligent ship general assembly construction data service call link tracking method according to claim 2, wherein The update method of the dynamic knowledge graph construction algorithm includes: Divide the device state parameters based on a sliding time window, and extract the change amount of entity relationships within the window; Update the device association relationships and path weights in the knowledge graph through an incremental learning algorithm, and dynamically adjust the structure of the service node topology graph according to the updated knowledge graph.

8. The intelligent ship general assembly construction data service call link tracking method according to claim 3, wherein, The parameter adjustment method of the distributed hash table includes: Set the initial number of hash shards according to the number of service call logs; Dynamically adjust the shard distribution through a consistent hashing algorithm to reduce the data migration volume during node expansion; Optimize the hash function parameters based on the shard load monitoring results to improve the efficiency of redundant node removal.

9. The intelligent ship general assembly construction data service call link tracking method according to claim 6, wherein The adaptive tracking network construction algorithm includes: Initialize the node attributes according to the service call module, and generate an edge weight tensor based on the stability weights; Use the link tracking feature matrix as the node input, and the edge weight tensor is composed of the call priority and stability weights; Iteratively update the path selection strategies of each node through a policy optimization algorithm to optimize the edge weight tensor; Generate an optimal service call link covering all nodes according to the optimized edge weight tensor.

10. An intelligent data service call link tracking system for ship general assembly construction, characterized in that, Includes: Full-life cycle data acquisition module: used to acquire the full-life cycle data set of ship hull construction, and the data set includes device state parameters, service call logs, and task execution results; Knowledge graph construction module: configured to generate a service node topology graph based on the device state parameters through a dynamic knowledge graph construction algorithm; Path optimization module: used to generate a load-balanced path through a service path optimization algorithm according to the service call logs; Abnormality detection module: perform abnormality detection processing on the task execution results to generate a service call abnormality marking sequence; Multi-dimensional tracking module: input the service node topology graph, load-balanced path, and abnormality marking sequence into a multi-dimensional tracking model to generate a link tracking feature matrix; Link generation module: used to generate a service call link optimization strategy based on the link tracking feature matrix through an adaptive tracking network construction algorithm, and output a standardized service call link.

Images