Steel structure raw material usage quota control management system

Abstract

The application discloses a control management system based on a steel structure raw material usage quota, which comprises a quota acquisition module, which is used for acquiring a raw material usage quota and a raw material quantitative quota of a current sub-component; a local subgraph representation module, which is used for extracting a local subgraph of the current sub-component in an assembly dependency path graph according to the raw material usage quota and the raw material quantitative quota of the current sub-component, and generating an embedded representation of the local subgraph; and an abnormal usage path output module, which is used for inputting the embedded representation of the local subgraph into an abnormal path recognition model and outputting a corresponding abnormal dependency path of the current sub-component in the assembly dependency path graph. The application realizes the recognition of the aggregation state of the abnormal usage in the assembly path and the analysis of the abnormal trend transmission by constructing the abnormal path recognition model, and enhances the recognition precision and structure positioning capability of the raw material usage abnormality in the graph structure.

Description

Technical Field

[0001] This invention relates to a control and management system, specifically a control and management system based on the usage quota of steel structure raw materials. Background Technology

[0002] In steel structure construction, the amount of raw materials used directly affects structural safety and cost control. Therefore, project management typically involves quantitative control of raw material usage for components, combined with threshold judgments to identify anomalies. Patent publication CN109992800A discloses a management method for a steel structure engineering management system, which solves the problems of untimely and unequal information acquisition and poor communication and coordination among various parties involved in steel structure engineering project management.

[0003] However, existing raw material usage control and management systems generally have the following problems: First, they easily overlook the assembly sequence and structural dependencies between components, making it difficult to determine whether anomalies are structural problems in the assembly chain; second, they cannot identify "tolerable errors" that are within the threshold range but have deviated from the trend, and such errors may become hidden dangers for the accumulation of structural anomalies; and third, it is difficult to identify whether anomalies form a coherent path or cluster area in the structure as a whole, thus limiting the accuracy of anomaly location and intervention. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a control and management system based on the usage quota of steel structure raw materials. It solves the technical problems mentioned in the background by constructing an assembly dependency path diagram and performing anomaly identification modeling on it.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A control and management system based on the usage quota of steel structure raw materials, the management system including:

[0007] The quota acquisition module is used to acquire the raw material usage quota and raw material quantity quota of the current sub-component;

[0008] The local subgraph representation module is used to extract the local subgraph of the current sub-component in the assembly dependency path graph based on the raw material usage amount and raw material quantity amount of the current sub-component, and generate the embedded representation of the local subgraph.

[0009] The abnormal path output module is used to input the embedded representation of the local subgraph into the abnormal path recognition model and output the abnormal dependency path corresponding to the current sub-component in the assembly dependency path graph.

[0010] In some specific embodiments, the modeling steps of the abnormal path identification model include:

[0011] S1. Construct an assembly dependency path diagram for several sub-components in the steel structure;

[0012] The assembly dependency path graph contains several sub-component nodes with assembly dependencies; the assembly dependencies include: path number and node number in each path number.

[0013] S2. Anchor the node to be determined from several sub-component nodes of the assembly dependency path diagram;

[0014] S3. Perform anomaly detection on the raw material usage of the node to be judged;

[0015] S4. If the raw material usage of the node to be judged is normal, then proceed to the next node to be judged according to the assembly dependency path diagram to perform anomaly judgment.

[0016] S5. If the node to be determined is an abnormal usage, then obtain the abnormal dependency path from the assembly dependency path graph and encode the abnormal dependency path into an abnormal path vector.

[0017] S6. Extract the local subgraph of the node corresponding to the abnormal usage and its N preceding adjacent nodes, and extract the embedded representation of the local subgraph.

[0018] S7. The embedding representation of the local subgraph is used as the input of the supervised model, and the abnormal path vector is used as the target variable. After iterative training, an abnormal path recognition model is generated.

[0019] In some specific embodiments, an assembly dependency path diagram of several sub-components in the steel structure is constructed, including:

[0020] S1-1. Obtain the types of components involved in the steel structure of the project;

[0021] S1-2. Based on the types of components involved, the steel structure in the project is broken down into several sub-components;

[0022] S1-3. Assign predefined assembly dependencies among several sub-components;

[0023] S1-4. Define each sub-component as a component node of the graph structure, and define assembly dependencies as directed edges between component nodes;

[0024] S1-5. Based on the directed edges between component nodes, connect any two adjacent component nodes that have assembly dependencies until all assembly dependencies are connected.

[0025] S1-6. Define the overall structure of the component nodes and their directed edges as the assembly dependency path graph of the sub-components.

[0026] In some specific embodiments, the abnormality determination of the raw material usage of the node to be determined includes:

[0027] S3-1. Obtain the raw material quantity and raw material usage quota of the node to be judged.

[0028] S3-2. Calculate the real-time error between the raw material usage quota and the raw material quantity quota of the node to be judged.

[0029] S3-3. Place the real-time quota error into the threshold range and determine whether the real-time quota error is located at the error position of the threshold range.

[0030] S3-4. Determine the amount of raw materials to be used based on the location of the error;

[0031] S3-5. If the error position is within the threshold range, the raw material usage of the node to be judged will be output as the normal usage.

[0032] S3-6. If the error location is outside the threshold range, the raw material usage of the node to be judged will be output as abnormal usage.

[0033] In some specific embodiments, if the node to be determined is an abnormal usage, the abnormal dependency path is obtained from the assembly dependency path graph, including:

[0034] S5-1. Based on the assembly dependency relationship, in the assembly dependency path graph, take the abnormal usage as the center and move forward M adjacent nodes.

[0035] S5-2, Perform anomaly detection on M adjacent nodes;

[0036] S5-3. If the raw material usage of M adjacent nodes is determined to have at least two abnormal usages, then extract the assembly dependency path between the two abnormal usages.

[0037] S5-4. Determine the assembly dependency relationship for the assembly dependency path between two abnormal usages.

[0038] S5-5. If there is an upstream and downstream assembly dependency relationship on the assembly dependency path between two abnormal uses, then export the abnormal uses and the dependency paths with upstream and downstream assembly dependencies as abnormal dependency paths.

[0039] In some specific embodiments, anomaly detection is performed on M subsequent adjacent nodes, including:

[0040] S5-2-1. Obtain the raw material quantity quota and raw material usage quota of M adjacent nodes;

[0041] S5-2-2 Calculate the real-time quota error of M adjacent nodes;

[0042] S5-2-3. Determine the error positions of M real-time quota errors in the next adjacent nodes that are within the threshold range, and obtain the M error positions;

[0043] S5-2-4. Generate the corresponding raw material usage determination for the next adjacent node based on the M error positions.

[0044] In some specific embodiments, the embedding vector of the local subgraph is extracted, including:

[0045] S6-1. Based on the path number and node number of the preceding adjacent node in the local subgraph, determine the assembly dependency feature of the preceding adjacent node in the local subgraph.

[0046] S6-2. Based on the real-time quota error of the preceding adjacent node in the local subgraph, determine the error accumulation characteristics of the preceding adjacent node in the local subgraph.

[0047] S6-3. Using the local subgraph with assembly dependency features and error accumulation features as input to the graph neural network, and after node feature aggregation, the embedding vector of the local subgraph is obtained.

[0048] In some specific embodiments, the step of obtaining the N assembly dependency features at the preceding adjacent nodes includes:

[0049] A6-1. Define each node in front of its adjacent node as an intermediate node for assembly dependency features;

[0050] A6-2. Obtain the path number of the intermediate node;

[0051] A6-3. Based on the path number of the intermediate node, obtain the preceding and following nodes adjacent to the intermediate node;

[0052] A6-4. Obtain the node sequence numbers of the previous and next nodes;

[0053] A6-5 concatenates the path number of the intermediate node, the node number of the previous node, and the node number of the next node to obtain the position number of the preceding adjacent node in the assembly dependency path graph.

[0054] A6-6. Encode the position number into a one-hot vector to generate N assembly dependency features in the preceding adjacent nodes.

[0055] In some specific embodiments, the step of obtaining the N error accumulation features at the previous adjacent nodes includes:

[0056] B6-1. Obtain the real-time rated errors of N preceding adjacent nodes;

[0057] B6-2. Normalize the real-time rated errors of N adjacent nodes to generate N normalized errors.

[0058] B6-3. Encode the N normalized errors into one-hot vectors and generate N error accumulation features in the preceding adjacent nodes.

[0059] In some specific embodiments, an abnormal path identification model is generated after iterative training, including:

[0060] Obtain the embedding representation of the local subgraph in the current batch and input it into the supervised learning model;

[0061] Perform forward propagation on the embedded representation of the local subgraph of the current batch and output its corresponding abnormal path prediction vector;

[0062] Calculate the mean squared error loss between the abnormal path prediction vector and the abnormal path vector;

[0063] Backpropagation is performed based on the mean squared error loss to update the model parameters, and the process continues iterating until the mean squared error loss converges.

[0064] This invention provides a control and management system based on the usage quota of steel structure raw materials, which has the following beneficial effects:

[0065] This invention constructs an abnormal path identification model and uses the assembly dependency structure of local subgraphs as the input representation of the model. Through joint encoding of structural position and error accumulation features, it identifies the potential causes of abnormal nodes in their preceding structures and determines whether there are multiple abnormal nodes in their subsequent adjacent paths that form a directed path with upstream and downstream assembly dependencies. This enables the identification of the aggregation state of abnormal usage in the assembly path and the analysis of abnormal trend transmission, thereby enhancing the identification accuracy and structural positioning capability of abnormal raw material usage in the graph structure. Attached Figure Description

[0066] Figure 1 This is a structural block diagram of the control and management system based on the usage quota of steel structure raw materials of the present invention;

[0067] Figure 2 This is a schematic diagram of the management process of the control and management system based on the usage quota of steel structure raw materials of the present invention;

[0068] Figure 3 This is a schematic diagram of the modeling process of the abnormal path identification model described in this invention;

[0069] Figure 4 This is a schematic diagram of the export process of the abnormal dependency path described in this invention. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] Example 1: Please refer to Figures 1 to 2 This invention provides a control and management system based on the usage quota of steel structure raw materials, the management system comprising:

[0072] The quota acquisition module is used to acquire the raw material usage quota and raw material quantity quota of the current sub-component;

[0073] The local subgraph representation module is used to extract the local subgraph of the current sub-component in the assembly dependency path graph based on the raw material usage amount and raw material quantity amount of the current sub-component, and generate the embedded representation of the local subgraph.

[0074] The abnormal path output module is used to input the embedded representation of the local subgraph into the abnormal path recognition model and output the abnormal dependency path corresponding to the current sub-component in the assembly dependency path graph.

[0075] It should be noted that an assembly dependency path diagram refers to a directed graph structure constructed based on the assembly sequence, connection logic, or mechanical dependencies of components in a steel structure design. Each node in the diagram represents a sub-component, and edges represent the assembly dependencies or construction order between components. This graph is used to express the upstream and downstream relationships of "who depends on whom" in a structure.

[0076] An abnormal dependency path refers to one or more directed paths in an assembly dependency path diagram, consisting of component nodes identified as having abnormal usage. This path is used to reveal whether there is a structural propagation trend of abnormal raw material usage.

[0077] This embodiment extracts a local subgraph embedding representation based on the difference between the raw material usage quota and the raw material quantity quota, and inputs this embedding representation into an abnormal path identification model for inference. Under the structural constraints of the assembly dependency path graph, it can identify abnormal paths in sub-component usage. It effectively integrates the assembly dependencies between sub-components, accurately locating abnormal dependency paths composed of abnormal nodes, thereby improving the ability to perceive the clustering trend of abnormal raw material usage in steel structures.

[0078] Example 2: See Figures 3 to 4 The technical solution of this embodiment 2 differs from that of embodiment 1 in that it discloses the modeling steps of the abnormal path identification model described in embodiment 1. These modeling steps include:

[0079] S1. Construct an assembly dependency path diagram for several sub-components in the steel structure;

[0080] The assembly dependency path graph contains several sub-component nodes with assembly dependencies; the assembly dependencies include: path number and node number in each path number.

[0081] The raw material quota represents the theoretical total amount of materials required for the sub-component under standard assembly conditions, based on the design drawings or budget list; while the raw material usage quota represents the real-time recorded amount of raw materials consumed by the sub-component during actual construction.

[0082] S2. Anchor the node to be determined from several sub-component nodes of the assembly dependency path diagram;

[0083] S3. Perform anomaly detection on the raw material usage of the node to be judged;

[0084] S4. If the raw material usage of the node to be judged is normal, then proceed to the next node to be judged according to the assembly dependency path diagram to perform anomaly judgment.

[0085] S5. If the node to be determined is an abnormal usage, then obtain the abnormal dependency path from the assembly dependency path graph and encode the abnormal dependency path into an abnormal path vector.

[0086] It should be noted that the abnormal path vector is used to characterize the path structure formed by the sub-component nodes identified as having abnormal usage in the assembly dependency path graph. The path is a sequence of nodes with directed dependencies. Its encoding method is as follows: extract the corresponding node number according to the order of appearance of the nodes in the path, and combine it with its position index in the path for position encoding, or perform one-hot encoding, embedding encoding or tensor concatenation operation on the ordered node sequence to form a fixed-dimensional path vector representation.

[0087] S6. Extract the local subgraph of the node corresponding to the abnormal usage and its N preceding adjacent nodes, and extract the embedded representation of the local subgraph.

[0088] S7. The embedding representation of the local subgraph is used as the input of the supervised model, and the abnormal path vector is used as the target variable. After iterative training, an abnormal path recognition model is generated.

[0089] This embodiment achieves supervised modeling of the abnormal path identification model by performing node-by-node anomaly determination of raw material usage at component nodes in the assembly dependency path graph, extracting abnormal dependency paths as target variables, and using the embedded representation of local subgraphs containing pre-dependencies as input.

[0090] Specifically, in this embodiment, step S1 further includes:

[0091] S1-1. Obtain the types of components involved in the steel structure of the project;

[0092] Specifically, component type refers to the different categories of components classified according to the steel structure design drawings, including but not limited to main beams, secondary beams, columns, gusset plates, support rods, connectors, etc. Component type is usually associated with its structural function, stress characteristics and installation location.

[0093] S1-2. Based on the types of components involved, the steel structure in the project is broken down into several sub-components;

[0094] Specifically, "sub-components" refers to the smallest assembly unit after further refining various components according to structural division principles. Each sub-component can be independently represented as a node in the diagram structure, possessing clear assembly start and end relationships and raw material usage information.

[0095] S1-3. Assign predefined assembly dependencies among several sub-components;

[0096] It should be noted that the predefined assembly dependency relationship refers to the pre-setting of the assembly sequence between components based on the assembly order, node connection logic, and construction process rules in the steel structure design drawings, and expressed through path numbers and node numbers in the structural drawings. The path number identifies a complete assembly path, while the node number indicates the assembly sequence position of the sub-component within the corresponding path. Together, they clarify the assembly dependency relationship of the sub-component within the overall assembled structure.

[0097] S1-4. Define each sub-component as a component node of the graph structure, and define assembly dependencies as directed edges between component nodes;

[0098] S1-5. Based on the directed edges between component nodes, connect any two adjacent component nodes that have assembly dependencies until all assembly dependencies are connected.

[0099] S1-6. Define the overall structure of the component nodes and their directed edges as the assembly dependency path graph of the sub-components.

[0100] This embodiment obtains the component type and decomposes it into several sub-components, assigns predefined assembly dependencies among the sub-components, and identifies the dependencies using path numbers and node numbers. Based on this, an assembly dependency path graph is constructed with sub-components as nodes and assembly dependencies as directed edges, thus realizing a structured expression of assembly relationships in steel structures.

[0101] Specifically, in this embodiment, step S3 further includes:

[0102] S3-1. Obtain the raw material quantity and raw material usage quota of the node to be judged.

[0103] S3-2. Calculate the real-time error between the raw material usage quota and the raw material quantity quota of the node to be judged.

[0104] S3-3. Place the real-time quota error into the threshold range and determine whether the real-time quota error is located at the error position of the threshold range.

[0105] S3-4. Determine the amount of raw materials to be used based on the location of the error;

[0106] S3-5. If the error position is within the threshold range, the raw material usage of the node to be judged will be output as the normal usage.

[0107] S3-6. If the error location is outside the threshold range, the raw material usage of the node to be judged will be output as abnormal usage.

[0108] This embodiment obtains the raw material quantity and raw material usage amount of the node to be judged, calculates the real-time quantity error, compares the error with the threshold range, and outputs the normal or abnormal status of raw material usage based on the error position, thereby realizing the judgment of the raw material usage of the node to be judged.

[0109] Specifically, in this embodiment, step S5 further includes:

[0110] S5-1. Based on the assembly dependency relationship, in the assembly dependency path graph, take the abnormal usage as the center and move forward M adjacent nodes.

[0111] S5-2, Perform anomaly detection on M adjacent nodes;

[0112] S5-3. If the raw material usage of M adjacent nodes is determined to have at least two abnormal usages, then extract the assembly dependency path between the two abnormal usages.

[0113] S5-4. Determine the assembly dependency relationship for the assembly dependency path between two abnormal usages.

[0114] S5-5. If there is an upstream and downstream assembly dependency relationship on the assembly dependency path between two abnormal uses, then export the abnormal uses and the dependency paths with upstream and downstream assembly dependencies as abnormal dependency paths.

[0115] This embodiment establishes a directed path filtering mechanism for local assembly areas by centering on abnormal usage nodes in the assembly dependency path graph and combining forward backtracking to the preceding adjacent nodes with backward advancement to the following adjacent nodes. By determining the abnormality of the following adjacent nodes, the assembly dependency path between at least two abnormal nodes is extracted, and it is further determined whether there is a clear upstream and downstream assembly dependency relationship in the path, thereby realizing the identification of structural associations between abnormal usage components.

[0116] Furthermore, step S5-2 further includes:

[0117] S5-2-1. Obtain the raw material quantity quota and raw material usage quota of M adjacent nodes;

[0118] S5-2-2 Calculate the real-time quota error of M adjacent nodes;

[0119] S5-2-3. Determine the error positions of M real-time quota errors in the next adjacent nodes that are within the threshold range, and obtain the M error positions;

[0120] S5-2-4. Generate the corresponding raw material usage determination for the next adjacent node based on the M error positions.

[0121] This embodiment obtains the raw material quantity quota and raw material usage quota of M adjacent nodes, calculates the corresponding real-time quota error, and judges its abnormal state based on the position of the error within the threshold range, thereby realizing the item-by-item usage determination of adjacent nodes on the structural path.

[0122] Specifically, in this embodiment, step S6 further includes:

[0123] S6-1. Based on the path number and node number of the preceding adjacent node in the local subgraph, determine the assembly dependency feature of the preceding adjacent node in the local subgraph.

[0124] S6-2. Based on the real-time quota error of the preceding adjacent node in the local subgraph, determine the error accumulation characteristics of the preceding adjacent node in the local subgraph.

[0125] S6-3. Using the local subgraph with assembly dependency features and error accumulation features as input to the graph neural network, and after node feature aggregation, the embedding vector of the local subgraph is obtained.

[0126] This embodiment extracts the path number and node sequence number of the preceding adjacent node in the local subgraph as assembly dependency features, and combines this with the real-time quota error of the node to form an error accumulation feature. After establishing directed connections between nodes in the graph structure, the local subgraph with the above features is used as input to a graph neural network for node feature aggregation, generating an embedding vector for the local subgraph. This method achieves a fusion representation of structural location information and raw material usage status, enabling the local subgraph to jointly represent assembly dependencies and error distribution.

[0127] Specifically, in this embodiment, the step of obtaining the assembly dependency features includes:

[0128] A6-1. Define each node in front of its adjacent node as an intermediate node for assembly dependency features;

[0129] A6-2. Obtain the path number of the intermediate node;

[0130] A6-3. Based on the path number of the intermediate node, obtain the preceding and following nodes adjacent to the intermediate node;

[0131] A6-4. Obtain the node sequence numbers of the previous and next nodes;

[0132] A6-5 concatenates the path number of the intermediate node, the node number of the previous node, and the node number of the next node to obtain the position number of the preceding adjacent node in the assembly dependency path graph.

[0133] For example, the position number is used to uniquely identify the structural positional relationship of the preceding adjacent node in the assembly dependency path graph, so as to express its upstream and downstream dependency relationship in the assembly path;

[0134] A6-6. Encode the position number into a one-hot vector to generate N assembly dependency features in the preceding adjacent nodes.

[0135] This embodiment uses each preceding adjacent node as an intermediate node, generates a structural position number based on the path number and the node indices of its preceding and following adjacent nodes, and encodes this position number as a one-hot vector to obtain the assembly dependency feature of the preceding adjacent node. This assembly dependency feature represents the structural positional relationship of the node in the assembly dependency path graph, enabling the node input to distinguish and express upstream and downstream dependencies in the assembly path.

[0136] Specifically, in this embodiment, the steps for obtaining the error accumulation feature include:

[0137] B6-1. Obtain the real-time rated errors of N preceding adjacent nodes;

[0138] B6-2. Normalize the real-time rated errors of N adjacent nodes to generate N normalized errors.

[0139] B6-3. Encode the N normalized errors into one-hot vectors and generate N error accumulation features in the preceding adjacent nodes.

[0140] This embodiment acquires the real-time quota errors of N preceding adjacent nodes, normalizes them, and encodes them into one-hot vectors to construct an error accumulation feature representing the intensity of error fluctuations. This feature characterizes the state distinction of each node under raw material usage deviations in a unified numerical expression form, enabling the local subgraph to have the ability to represent a relative tolerance level in the input features.

[0141] In this embodiment, the training steps of the abnormal path identification model include:

[0142] Obtain the embedding representation of the local subgraph in the current batch and input it into the supervised learning model;

[0143] Perform forward propagation on the embedded representation of the local subgraph of the current batch and output its corresponding abnormal path prediction vector;

[0144] Calculate the mean squared error loss between the abnormal path prediction vector and the abnormal path vector;

[0145] Backpropagation is performed based on the mean squared error loss to update the model parameters, and the process continues iterating until the mean squared error loss converges.

[0146] In this embodiment, the abnormal path identification model takes the abnormal usage node as the starting point for identification. Based on its preceding adjacent structure in the assembly dependency path graph, it extracts a local subgraph and generates an embedded representation as the model input. This local subgraph only selects the preceding adjacent nodes of the abnormal node because, in the assembly logic of steel structures, the assembly order of sub-components reflects the directed propagation path of structural dependencies. If the current node has an abnormal raw material usage, its material usage in the preceding assembly nodes may not have triggered the abnormal threshold, but there is already an error accumulation or deviation trend. Although this trend itself is within the "tolerance range," once it enters the subsequent structural assembly chain, its tolerance characteristics may be the preconditions for structural anomalies in subsequent paths. Therefore, by introducing "error accumulation characteristics," the model realizes the modeling of the state of preceding components that have not triggered the threshold but may have abnormal causes, and establishes a semantic space for upstream and downstream abnormal transmission in advance at the input end.

[0147] The model's output design also reflects assembly logic constraints. The output objective is that in the subsequent adjacent structure of the current anomalous node, there exist at least two nodes that have been identified as anomalous usages, and these two anomalous nodes must satisfy a clear upstream and downstream assembly dependency relationship. This means that the model not only focuses on whether anomalous nodes exist, but also on whether these anomalous nodes structurally form a continuous assembly chain. Only when multiple anomalous nodes have assembly path connectivity can they be determined to constitute an anomalous dependency path. This path-level structural identification effectively distinguishes between isolated error triggers and systematic assembly anomaly trends, enhancing the model's ability to determine the structure of anomalous cluster regions.

[0148] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means.

[0149] The computer-readable storage medium can be any available medium that a computer can access, or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives (SSDs).

[0150] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Furthermore, the mutual couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0151] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A control and management system based on the usage quota of steel structure raw materials, characterized in that, The management system includes: The quota acquisition module is used to acquire the raw material usage quota and raw material quantity quota of the current sub-component; The local subgraph representation module is used to extract the local subgraph of the current sub-component in the assembly dependency path graph based on the raw material usage amount and raw material quantity amount of the current sub-component, and generate the embedded representation of the local subgraph. The abnormal path output module is used to input the embedded representation of the local subgraph into the abnormal path recognition model and output the abnormal dependency path corresponding to the current sub-component in the assembly dependency path graph. The modeling steps of the abnormal path identification model include: Construct an assembly dependency path diagram for several sub-components in a steel structure; The assembly dependency path graph contains several sub-component nodes with assembly dependencies; the assembly dependencies include: path number and node number in each path number. Anchor the node to be determined from several sub-component nodes of the assembly dependency path graph; Perform anomaly detection on the raw material usage at the node to be judged; If the raw material usage of the node to be judged is normal, then proceed to the next node to be judged according to the assembly dependency path graph to perform anomaly judgment. If the node to be determined is an abnormal usage, then obtain the abnormal dependency path from the assembly dependency path graph and encode the abnormal dependency path into an abnormal path vector. Extract the local subgraph of the node corresponding to the abnormal usage and its N preceding adjacent nodes, and extract the embedding representation of the local subgraph; The embedding representation of the local subgraph is used as the input to the supervised model, and the abnormal path vector is used as the target variable. After iterative training, an abnormal path recognition model is generated. If the node to be determined is an abnormal usage, then obtain the abnormal dependency path from the assembly dependency path graph, including: Based on the assembly dependency relationship, in the assembly dependency path graph, with the abnormal usage as the center, advance M adjacent nodes to the end. Perform anomaly detection on M adjacent nodes; If the raw material usage of M adjacent nodes is determined to have at least two abnormal usages, then the assembly dependency path between the two abnormal usages is extracted. Determine the assembly dependency relationship for the assembly dependency path between two abnormal usages; If there are upstream and downstream assembly dependencies on the assembly dependency path between two abnormal uses, then the abnormal uses and their dependency paths with upstream and downstream assembly dependencies are exported as abnormal dependency paths. The extraction of the embedding vector of the local subgraph includes: Based on the path number and node number of the preceding adjacent node in the local subgraph, determine the assembly dependency feature of the preceding adjacent node in the local subgraph. Based on the real-time quota error of the preceding adjacent node in the local subgraph, determine the error accumulation characteristics of the preceding adjacent node in the local subgraph. The local subgraphs with assembly dependency features and error accumulation features are used as input to the graph neural network. After node feature aggregation, the embedding vector of the local subgraph is obtained.

2. The control and management system based on the usage quota of steel structure raw materials according to claim 1, characterized in that, Construct an assembly dependency path diagram for several sub-components in the steel structure, including: Obtain the types of steel structure components involved in the project; Based on the types of components involved, the steel structure in the project is broken down into several sub-components; Assign predefined assembly dependencies among several sub-components; Each sub-component is defined as a component node of a graph structure, and assembly dependencies are defined as directed edges between component nodes; Connect any two adjacent component nodes that have assembly dependencies based on the directed edges between component nodes, until all assembly dependencies are connected. The overall structure of a component node and its directed edges is defined as the assembly dependency path graph of the sub-components.

3. The control and management system based on the usage quota of steel structure raw materials according to claim 1, characterized in that, Anomaly detection is performed on the raw material usage at the node to be judged, including: Obtain the raw material quantity and raw material usage quota for the node to be judged. Calculate the real-time error between the raw material usage quota and the raw material quantity quota for the node to be judged; Place the real-time quota error within the threshold range and determine whether the real-time quota error is located within the threshold range error position. The amount of raw materials used is determined based on the location of the error. If the error location is within the threshold range, the raw material usage of the node to be judged will be output as the normal usage. If the error location is outside the threshold range, the raw material usage of the node to be judged will be output as abnormal usage.

4. The control and management system based on the usage quota of steel structure raw materials according to claim 1, characterized in that, Anomaly detection is performed on M adjacent nodes, including: Obtain the raw material quantity quota and raw material usage quota of M adjacent nodes; Calculate the real-time quota error of M adjacent nodes; Determine the error positions of M real-time quota errors in the next adjacent nodes that are within the threshold range, and obtain the M error positions; Based on M error locations, the corresponding raw material usage determination for the next adjacent node is generated.

5. The control and management system based on the usage quota of steel structure raw materials according to claim 1, characterized in that, The steps for obtaining the N assembly dependency features in the preceding adjacent nodes include: Define each intermediate node that is in front of its neighboring node as an assembly dependency feature; Get the path number of the intermediate node; Based on the path number of the intermediate node, obtain the preceding and following nodes adjacent to the intermediate node; Get the node indices of the previous and next nodes; By concatenating the path number of the intermediate node, the node number of the previous node, and the node number of the next node, we can obtain the position number of the preceding adjacent node in the assembly dependency path graph. Encode the position number into a one-hot vector to generate N assembly dependency features in the preceding adjacent nodes.

6. The control and management system based on the usage quota of steel structure raw materials according to claim 1, characterized in that, The steps for obtaining the N error accumulation features at the preceding adjacent nodes include: Obtain the real-time rated errors of N preceding adjacent nodes; Normalize the real-time rated errors of N adjacent nodes to generate N normalized errors; The N normalized errors are encoded into one-hot vectors, generating N cumulative error features in the preceding neighboring nodes.

7. The control and management system based on the usage quota of steel structure raw materials according to claim 1, characterized in that, An abnormal path identification model is generated after iterative training, including: Obtain the embedding representation of the local subgraph in the current batch and input it into the supervised learning model; Perform forward propagation on the embedded representation of the local subgraph of the current batch and output its corresponding abnormal path prediction vector; Calculate the mean squared error loss between the abnormal path prediction vector and the abnormal path vector; Backpropagation is performed based on the mean squared error loss to update the model parameters, and the process continues iterating until the mean squared error loss converges.

Images