Production and construction material data sharing method and system based on cloud computing

Abstract

The present invention provides a cloud computing-based production and construction material data sharing method and system. The method generates a multidimensional feature space for production and construction materials, performs cross-level feature association processing on the multidimensional feature space through a cloud computing platform, and generates a material construction sharing association map. A sharing demand prediction model is established based on the material construction sharing association map, and a material data sharing demand sequence for different construction stages is output. A material data sharing task list including priority sorting is generated based on the material data sharing demand sequence. The material data sharing task list is mapped to a preset cloud sharing resource pool, and a material data sharing strategy including resource scheduling parameters is generated. The present invention achieves dynamic matching from demand forecasting to resource scheduling, replacing the traditional static resource allocation method and improving the adaptability of demand and resources during the implementation of material data sharing.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and system for sharing production and construction material data based on cloud computing. Background Art

[0002] With the in-depth application of cloud computing technology in the field of production and construction, material data sharing has received increasing attention. It can realize the interaction and sharing of various material-related data among different participants or systems in the production and construction process. At present, data interaction is achieved through preset sharing rules or matching mechanisms. However, such methods often lack the correlation between the characteristic levels of material data, which easily leads to a low matching degree between sharing requirements and material properties. At the same time, it is difficult to achieve dynamic adaptation of requirements and resources during the sharing implementation process, which affects the efficiency and accuracy of material data sharing. Based on this, how to optimize the material data sharing process to improve the sharing quality has become an urgent problem to be solved in the current production and construction field. Summary of the Invention

[0003] In view of this, the present invention provides a method for sharing production and construction material data based on cloud computing.

[0004] The technical solution of the embodiment of the present invention is achieved as follows:

[0005] On the one hand, an embodiment of the present invention provides a cloud computing-based production and construction material data sharing method, including: generating a multi-dimensional feature space for production and construction materials, the multi-dimensional feature space including a material basic feature layer, a construction-related feature layer, and a sharing demand feature layer; performing cross-level feature association processing on the multi-dimensional feature space through a cloud computing platform to generate a material construction sharing association map; establishing a sharing demand prediction model based on the material construction sharing association map to output a material data sharing demand sequence for different construction stages; generating a material data sharing task list including priority sorting according to the material data sharing demand sequence; mapping the material data sharing task list to a preset cloud sharing resource pool to generate a material data sharing strategy including resource scheduling parameters.

[0006] On the other hand, the present invention provides a computer system including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor implements the steps in the above method when executing the program.

[0007] The cloud computing-based production and construction material data sharing method provided by the present invention generates a multi-dimensional feature space including a material basic feature layer, a construction-related feature layer and a sharing demand feature layer, hierarchically integrates the features of the three dimensions of material attributes, construction scene adaptability and data sharing requirements that were originally independent, and forms a feature system with inherent logical associations, thereby avoiding the limitation of the binary separation of attributes and requirements in traditional material data processing and improving the integrity of the associations between the feature dimensions of material data; through the cloud computing platform, cross-level feature association processing is performed on the multi-dimensional feature space to generate a material construction sharing association map, and the implicit associations between features are converted into explicit structured maps, replacing the traditional Relying on manually defined rules or simple supply and demand matching methods, the structured expression ability of material feature associations is improved; a dynamic shared demand prediction model is established based on the material construction shared association map, and the structured association relationship between features in the map is integrated into the prediction process, breaking through the limitation of traditional prediction models that only rely on single-dimensional data, and improving the accuracy of material data sharing demand prediction at different construction stages; a task list with priority sorting is generated according to the shared demand sequence and mapped to the cloud shared resource pool to generate an implementation strategy, realizing dynamic matching from demand prediction to resource scheduling, replacing the traditional static resource allocation method, and improving the adaptability of demand and resources during the implementation of material data sharing. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A schematic diagram of the implementation flow of a cloud computing-based production and construction material data sharing method provided in an embodiment of the present invention.

[0009] Figure 2 A schematic diagram of a hardware entity of a computer system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0010] An embodiment of the present invention provides a method for sharing production and construction material data based on cloud computing, which can be executed by a processor of a computer system, wherein the computer system can be a cloud server. Figure 1 A schematic diagram of the implementation flow of a cloud computing-based production and construction material data sharing method provided in an embodiment of the present invention is shown in FIG. Figure 1 As shown, the method includes the following steps:

[0011] Step S100: Generate a multi-dimensional feature space for production and construction materials, where the multi-dimensional feature space includes a material basic feature layer, a construction-related feature layer, and a shared demand feature layer.

[0012] The multidimensional feature space for production and construction materials is a comprehensive feature system that describes data on production and construction materials from multiple perspectives. The material base feature layer focuses on the inherent properties of the material itself. These properties are the material's essential attributes and do not change with external factors such as the construction scenario. The construction-related feature layer emphasizes the relationship between materials and construction scenarios, reflecting the material's adaptability and performance under different construction environments and process requirements. The shared demand feature layer predicts the demand for material data at different construction nodes based on historical data and current construction plans.

[0013] As an implementation manner, step S100 can be specifically implemented as the following steps S110 to S150:

[0014] Step S110: Collecting full material attribute data, which includes material characteristic data, production process data, and quality inspection data.

[0015] Material property data is a collection of data that comprehensively describes various types of production and construction materials. Material property data refers to data reflecting the physical, chemical, and mechanical properties of the material itself. These properties determine the material's basic performance and application. Production process data records information such as the process methods, process parameters, and process control indicators used in the material production process, reflecting the stability of the material production process and production quality. Quality inspection data is data obtained through material quality testing and evaluation, including quality characteristic parameters and quality defect records, and is an important basis for determining whether the material meets quality standards. Material property data can be measured and analyzed using specialized testing equipment. For example, a densitometer can be used to measure material density, a chemical analyzer can be used to analyze the material's chemical composition, and mechanical properties testing equipment can be used to test mechanical properties such as strength and toughness. Production process data can be obtained from the control system of the production equipment, including process parameter sequences and process control indicators, or from production records. Quality inspection data can be obtained from test reports and records provided by the quality inspection department.

[0016] Step S120: performing structured feature extraction on the material characteristic data to generate a material basic feature layer including a physical characteristic vector, a chemical composition vector, and a mechanical property vector.

[0017] Material property data exists in a raw, unstructured form and requires structured feature extraction to convert it into a feature vector with a clear mathematical representation. The physical property vector is a quantitative representation of the material's physical properties. It is obtained by processing the physical description information in the material property data and includes information on the material's morphological characteristics, density characteristics, and thermal properties. The chemical composition vector reflects the material's chemical composition and is obtained by extracting and analyzing elemental features from the component description information in the material property data. The mechanical property vector is a quantitative representation of the material's mechanical properties and is obtained by extracting and processing performance indicators from the mechanical test data in the material property data.

[0018] As an implementation manner, step S120 can be specifically implemented as the following steps S121 to S124:

[0019] Step S121: Perform parameter conversion on the physical description information in the material property data, convert the text description into quantified physical property parameters, which include morphological characteristic parameters, density characteristic parameters and thermal characteristic parameters; standardize the physical characteristic parameters to eliminate dimensional differences and generate physical property vectors.

[0020] The physical descriptions in material property data, for example, exist in natural language. Parametric transformation involves converting these textual descriptions into specific numerical parameters for subsequent calculations and analysis. Morphological parameters describe the material's geometry and dimensions, such as length, diameter, and thickness; density parameters reflect the relationship between a material's mass and volume; and thermal parameters include thermal conductivity and specific heat capacity.

[0021] Normalization eliminates dimensional differences between physical property parameters, making them numerically comparable. For example, normalization involves subtracting the mean from each parameter and dividing it by its standard deviation. This results in all parameters having a mean of 0 and a standard deviation of 1, placing them on the same scale.

[0022] Step S122: Extract elemental features from the component description information in the material characteristic data, identify the composition ratio of the main components and trace components, and generate chemical composition parameters; compare the chemical composition parameters with the standard component database, calculate the component similarity score, and generate a chemical composition vector based on the composition ratio.

[0023] The composition description information in the material property data lists the various chemical elements contained in the material and their approximate content ranges. Element feature extraction involves accurately identifying the main and trace components from this description information and determining their composition ratios. Chemical composition parameters are a specific quantitative representation of the material's chemical composition, including the names of the various elements and their corresponding composition ratios. The standard composition database is a pre-established database that contains chemical composition information for various standard materials. By comparing the chemical composition parameters with the standard composition database, a similarity score between the material's composition and the standard composition can be calculated. The higher the similarity score, the closer the material's chemical composition is to the standard composition. Combining the composition ratio and similarity score can generate a chemical composition vector that more comprehensively reflects the material's chemical composition characteristics.

[0024] Step S123: extracting performance indicators from the mechanical test data in the material characteristic data, including strength characteristic parameters, elastic characteristic parameters, and toughness characteristic parameters; normalizing the strength characteristic parameters, elastic characteristic parameters, and toughness characteristic parameters to generate a mechanical performance vector.

[0025] The mechanical test data in material property data is the raw data obtained by mechanical property testing of the material, from which representative performance indicators need to be extracted. Strength characteristic parameters reflect the material's ability to resist damage, such as the material's tensile strength and compressive strength. Elastic characteristic parameters describe the material's deformation characteristics within the elastic range, such as the elastic modulus. Toughness characteristic parameters reflect the material's ability to absorb energy and undergo plastic deformation before fracture, such as impact toughness. Normalization is the process of unifying different performance indicators into the same numerical range, such as the [0,1] interval. This eliminates dimensional differences and numerical size differences between different performance indicators, making them equally important in subsequent calculations and analyses.

[0026] Step S124: The physical property vector, chemical composition vector and mechanical property vector are spliced ​​in a preset feature dimension order to generate a material basic feature layer.

[0027] The preset feature dimension order is a predefined rule that dictates the order and arrangement of the physical property vectors, chemical composition vectors, and mechanical property vectors when they are stitched together. By stitching the three vectors together in this order, they are integrated into a unified feature representation, the material base feature layer.

[0028] Step S130: Perform construction scene association processing on the production process data and quality inspection data, and generate a construction association feature layer including a process adaptability vector, a construction environment sensitivity vector, and a quality fluctuation tolerance vector in combination with the construction process standard library.

[0029] Construction scenario association processing is to associate and analyze production process data and quality inspection data with specific construction scenarios to determine the performance and performance of materials in different construction scenarios. The construction process standard library is a database that contains standard parameters and requirements for various construction processes, which provides a reference for evaluating the compatibility of materials and construction processes. The process adaptability vector reflects the degree of matching between the material's production process and different construction processes, and is obtained by comparing and analyzing the production process data and the construction process standard library. The construction environment sensitivity vector reflects the sensitivity of the material's quality characteristics to different construction environment factors, and is obtained by performing sensitivity analysis on the quality inspection data and the construction environment factor database. The quality fluctuation tolerance vector determines the tolerance range for material quality fluctuations in different construction stages based on quality inspection data and construction quality requirements.

[0030] As an implementation manner, step S130 can be specifically implemented as the following steps S131 to S136:

[0031] Step S131: extracting process parameter sequences and process control indicators from the production process data. The process parameter sequences include temperature parameters, pressure parameters, and time parameters during the production process. The process control indicators include process stability indicators and process consistency indicators.

[0032] Production process data contains a wealth of information, from which key process parameter sequences and process control indicators must be extracted. The process parameter sequence records important parameters in the production process. Temperature parameters reflect the temperature conditions during production and have a significant impact on material performance and quality; pressure parameters reflect the pressure environment during production; and time parameters record the duration of each step in the production process. Process control indicators are used to assess the stability and consistency of the production process. Process stability indicators reflect the fluctuations in process parameters during production, while process consistency indicators reflect the degree of process consistency across batches of materials.

[0033] Step S132: Compare the process parameter sequence with the standard process parameter range in the construction process standard library, and calculate the process parameter matching degree.

[0034] The construction process standard library specifies the standard process parameter ranges required for various construction processes. By comparing the extracted process parameter sequence with the standard process parameter ranges, we can determine whether each process parameter is within the standard range and calculate the process parameter matching degree. The process parameter matching degree is a comprehensive indicator that reflects the overall degree of compatibility between the material production process and the construction process standard.

[0035] Step S133: Generate a process suitability vector based on the process control index and the matching degree of the process parameters. The process suitability vector includes the suitability scores of different construction processes.

[0036] Process control indicators reflect the stability and consistency of the production process, while process parameter matching reflects the degree of alignment between the production process and construction process standards. Taking these two factors into account, a process suitability vector is generated, which contains the suitability scores for the material and different construction processes. Different construction processes may have different requirements for the material's production process, so the suitability score for each material and construction process needs to be calculated separately.

[0037] Step S134: extract the quality characteristic parameters and quality defect records in the quality inspection data, combine them with the construction environment factor database, calculate the sensitivity coefficient of the quality characteristic parameters to each environmental factor through the sensitivity analysis algorithm, and generate the construction environment sensitivity vector.

[0038] Quality inspection data includes various material quality characteristic parameters and quality defect records, reflecting the material's quality status. The construction environmental factor database contains information on various construction environmental factors, such as temperature, humidity, and vibration. A sensitivity analysis algorithm is used to analyze the sensitivity of one variable to changes in another. By correlating quality characteristic parameters with the construction environmental factor database, the sensitivity analysis algorithm can be used to calculate the sensitivity coefficients of the quality characteristic parameters to various environmental factors.

[0039] The construction environment sensitivity vector is a quantitative representation of the sensitivity of material quality characteristics to different environmental factors, including temperature sensitivity parameters, humidity sensitivity parameters, and vibration sensitivity parameters.

[0040] Step S135: Based on the quality defect records and construction quality requirement standards, determine the tolerance range for material quality fluctuations in different construction stages and generate a quality fluctuation tolerance vector.

[0041] The quality defect record details various quality issues encountered during the material production and testing process. The construction quality standards specify the material quality requirements for different construction stages. Based on the quality defect record and construction quality standards, analysis and evaluation can determine the permissible fluctuation range for material quality at different construction stages. This fluctuation range reflects the requirements for material quality stability during the construction process, and different construction stages may have different tolerance levels for quality fluctuations. The quality fluctuation tolerance vector quantifies the quality fluctuation tolerance ranges for different construction stages, including the corresponding tolerance values ​​for each construction stage.

[0042] For example, in a construction project, during the foundation construction phase, as the foundation primarily bears stress, a relatively high tolerance is imposed on concrete strength fluctuations. Suppose, based on quality defect records and construction quality standards, the tolerance for concrete strength fluctuations during the foundation construction phase is determined to be ±10% of the design strength. During the main structure construction phase, however, the strength requirements are more stringent, and the tolerance range may be ±5% of the design strength. Other concrete quality characteristics, such as slump, also have different tolerance ranges at different construction stages. Suppose the tolerance range for slump fluctuations is ±20mm during the foundation construction phase and ±10mm during the main structure construction phase. By organizing and quantifying these tolerance ranges for different quality characteristics at different construction stages, a quality fluctuation tolerance vector can be generated, such as [10% strength tolerance during foundation construction, 20mm slump tolerance during foundation construction, 5% strength tolerance during main structure construction, 10mm slump tolerance during main structure construction].

[0043] Step S136: Perform feature fusion processing on the process adaptability vector, the construction environment sensitivity vector, and the quality fluctuation tolerance vector to generate a construction-related feature layer.

[0044] Feature fusion integrates and optimizes different types of feature vectors to generate a more comprehensive and representative feature layer. The process suitability vector reflects the compatibility of the material production process with different construction techniques. The construction environment sensitivity vector reflects the sensitivity of material quality characteristics to construction environment factors. The quality fluctuation tolerance vector specifies the tolerance range for material quality fluctuations during different construction phases. Feature fusion combines the information contained in these three vectors, eliminating redundant information and highlighting key features, thereby generating a construction-related feature layer that more comprehensively and accurately reflects the relationship between materials and construction scenarios.

[0045] Feature fusion can be performed using a variety of methods, such as weighted summation and principal component analysis. Taking the weighted summation method as an example, different weights are first assigned to the process adaptability vector, the construction environment sensitivity vector, and the quality fluctuation tolerance vector. These weights are determined based on the importance of each vector in reflecting the relationship between materials and construction scenarios.

[0046] Step S140: Based on the historical shared record data and the current construction plan data, the material data demand intensity of different construction nodes is calculated through the demand forecasting algorithm, and a shared demand feature layer including the demand urgency vector, the data completeness demand vector and the sharing frequency vector is generated.

[0047] Historical sharing records document the sharing of material data during past construction processes, including information such as the time of sharing, the type of data shared, and the parties involved. Current construction plan data clearly defines the various construction nodes, construction schedule, and the material and data requirements for each node in the current project. The demand forecasting algorithm, based on data analysis and model building, predicts the intensity of material data demand at different future construction nodes based on historical data and current conditions.

[0048] The demand urgency vector reflects the degree of urgency of the need for material data at different construction nodes. For example, at certain critical construction nodes, immediate access to relevant material data may be necessary to ensure smooth construction progress, and the demand urgency is relatively high. At non-critical construction nodes, the need for material data may be relatively less urgent. The data completeness requirement vector reflects the requirements for material data integrity at different construction nodes. Some construction nodes may require comprehensive and detailed material data, while others may only require partial key data. The sharing frequency vector indicates how frequently material data is shared at different construction nodes. At nodes with faster construction progress or frequent data updates, the sharing frequency may be higher.

[0049] Step S150: performing spatial mapping processing on the material basic feature layer, the construction-related feature layer, and the shared demand feature layer according to the feature hierarchical relationship to generate a multi-dimensional feature space with hierarchical association weights.

[0050] The material basic feature layer, construction-related feature layer, and shared demand feature layer describe the data of production and construction materials from different perspectives, and there is a certain hierarchical correlation between them. The material basic feature layer is the most basic level, describing the inherent characteristics of the material itself; the construction-related feature layer is based on the material basic feature layer and reflects the relationship between the material and the construction scene; the shared demand feature layer further considers the need for sharing material data during the construction process. Spatial mapping processing integrates and maps these three feature layers according to the hierarchical correlation between them, converting them from different feature spaces into a unified multi-dimensional feature space. In this process, it is necessary to assign corresponding weights to the correlations between different levels. These hierarchical correlation weights reflect the importance and degree of correlation between the features at different levels.

[0051] Step S200: Perform cross-level feature association processing on the multi-dimensional feature space through the cloud computing platform to generate a material construction sharing association map.

[0052] The multi-dimensional feature space includes a material basic feature layer, a construction-related feature layer, and a shared demand feature layer. There are potential correlations between these layers, but these correlations need to be mined and clarified through processing. Cross-level feature association processing is to find the connections and interactions between features at different levels, and integrate the discrete features in the multi-dimensional feature space into an organic whole. The material construction shared association map is a graphical representation that shows the correlation between features at different levels of production and construction materials in the form of a map, including correlation paths and correlation strengths. By generating a material construction shared association map, one can intuitively understand the sharing of material data during the construction process and the mutual influence between different features, providing clearer guidance for the sharing and management of material data.

[0053] As an implementation manner, step S200 can be specifically implemented as the following steps S210 to S270:

[0054] Step S210: inputting the material basic feature layer, the construction related feature layer and the shared demand feature layer in the multi-dimensional feature space into the feature association component of the cloud computing platform.

[0055] The feature association component of the cloud computing platform is a module specifically designed to handle feature association tasks, with the capabilities of receiving, processing, and analyzing data. Inputting the material-based feature layer, construction-related feature layer, and shared-requirement feature layer in the multi-dimensional feature space into the feature association component allows the component to obtain feature information at these different levels for subsequent cross-level feature association processing. During the input process, the accuracy and completeness of the data must be ensured, and the data format must be converted and processed according to the requirements of the feature association component. For example, the feature vectors of the material-based feature layer, construction-related feature layer, and shared-requirement feature layer are input into the feature association component in a specific format (such as a matrix) so that the component can recognize and process this data.

[0056] Step S220: In the feature association component, a graph neural network algorithm is used to perform node initialization processing on features at each level, and each feature dimension is converted into a graph node.

[0057] The graph neural network algorithm is a neural network algorithm specifically designed for processing graph-structured data. It can automatically learn the feature representations of nodes and edges in a graph, as well as the relationships between them. In the feature association component, the graph neural network algorithm is used to initialize nodes at each level of the feature space. This abstracts each feature dimension in the multidimensional feature space into a node in the graph. Each graph node represents a specific feature. For example, a dimension (such as density) of the physical property vector in the material-based feature layer can be a graph node, and an element (such as the suitability score for a specific construction process) of the process suitability vector in the construction-related feature layer can also be a graph node. Through node initialization, these feature dimensions are converted into graph nodes. During initialization, each graph node is assigned an initial feature vector. These feature vectors can be set based on the values ​​of the original feature dimensions. For example, the density value in the physical property vector can be used as an element of the initial feature vector for the corresponding graph node. Nodes can also be assigned different initial weights based on their importance or other prior knowledge.

[0058] Step S230: Calculate the association strength between nodes at different levels, where the association strength is calculated based on the mutual information value and conditional probability between features.

[0059] The strength of the association between nodes at different levels reflects their closeness and the degree of mutual influence. Mutual information is a metric used to measure the degree of information sharing between two random variables and can reflect the dependency between two features. Conditional probability represents the probability that one feature will take on a specific value given that another feature takes on a specific value.

[0060] By combining mutual information and conditional probability to calculate association strength, we can more comprehensively and accurately assess the associations between nodes at different levels. Calculating association strength is a complex data analysis and calculation process that requires processing and statistical analysis of large amounts of data.

[0061] For example, when calculating the strength of the association between a physical property node in the material basic feature layer and a process suitability node in the construction-related feature layer, the feature data corresponding to the two nodes is first extracted, and then the mutual information value and conditional probability between them are calculated. Assume that the physical property node represents the strength of the steel material, and the process suitability node represents the suitability score of the steel material and the welding process. By analyzing historical data, the mutual information value between the steel material strength and the welding process suitability score is calculated, as well as the conditional probability of the welding process suitability score under different steel strength values. Finally, the mutual information value and conditional probability are weighted and fused to obtain the strength of the association between the two nodes.

[0062] As an implementation manner, step S230 may be specifically implemented as the following steps S231 to S237:

[0063] Step S231: extract any feature node in the material basic feature layer as a source node, and any feature node in the construction associated feature layer as a target node.

[0064] When calculating the correlation strength between nodes at different levels, it is necessary to clearly identify the node pairs involved in the calculation. Any feature node in the material base feature layer is used as the source node, and any feature node in the construction-related feature layer is used as the target node. This selection method comprehensively considers the correlation between features at different levels. Because the material base feature layer describes the inherent properties of the material, while the construction-related feature layer reflects the relationship between the material and the construction scenario, calculating the correlation strength between the nodes can reveal the impact of material properties on the construction process and the requirements of the construction process on material selection.

[0065] Step S232: Calculate the mutual information value between the source node and the target node. The mutual information value is used to measure the degree of information sharing between two feature nodes.

[0066] Mutual information (MI) is a key concept in information theory that measures the degree of dependency and information sharing between two random variables. For two features represented by a source node and a target node, a larger MI indicates a closer relationship between them, meaning that changes in one feature can provide more information about the other.

[0067] Mutual information calculations are typically based on probability distributions. First, the joint probability distribution of the features represented by the source and target nodes and their respective marginal probability distributions must be obtained. Then, the joint and marginal probability distributions are substituted into the mutual information formula for calculation.

[0068] For example, for the two nodes in the bridge construction project mentioned above—the elastic modulus of the steel and the compatibility score between the steel and the bridge welding process—we collected a large amount of historical data and counted the frequencies of occurrence of different elastic modulus and fitness scores to estimate their joint probability distribution and marginal probability distribution. Suppose, through statistical analysis, we determine that the probability of an elastic modulus value of E1 is P(E1), the probability of a fitness score value of S1 is P(S1), and the joint probability of both an elastic modulus of E1 and a fitness score of S1 is P(E1, S1). The mutual information between these two nodes is calculated using the mutual information formula I(E, S) = ∑∑P(E, S)log(P(E, S) / (P(E)P(S))) (where E represents the elastic modulus and S represents the fitness score).

[0069] Step S233: Generate a joint probability distribution model for the source and target nodes based on the historical data, and calculate the conditional probability of the target node taking a specific value given the source node taking a specific value. The joint probability distribution model describes the probability distribution of the two features represented by the source and target nodes taking the same value. Through analysis and statistics of historical data, a joint probability distribution model for the source and target nodes can be established.

[0070] Conditional probability is the probability that a target node will take a specific value given that a source node takes a specific value. Conditional probability can be calculated based on a joint probability distribution model and marginal probability distribution. According to the definition of conditional probability, the conditional probability that target node Y will take a specific value y given that source node X takes a specific value x is P(Y=y|X=x) = P(X=x, Y=y) / P(X=x). For example, taking the two nodes in a bridge construction project—the elastic modulus of steel and the suitability score for the steel and bridge welding process—a joint probability distribution model is established based on historical data. Assuming the probability P(E1) that the elastic modulus is E1 and the joint probability P(E1,S1) that the elastic modulus is E1 and the suitability score is S1, then the conditional probability that the suitability score is S1, given that the elastic modulus is E1, is P(S1|E1) = P(E1,S1) / P(E1).

[0071] Step S234: Perform weighted fusion processing on the mutual information value and the conditional probability to generate a preliminary correlation strength value.

[0072] The mutual information value and conditional probability reflect the association relationship between the source node and the target node from different perspectives. By weighted fusion processing, we can comprehensively consider these two factors and obtain a more comprehensive and accurate representation of the association strength.

[0073] The weighted fusion method assigns different weights to the mutual information value and conditional probability, then multiplies them by the corresponding weights and adds them together. The weights can be adjusted based on actual conditions and experience. For example, if the mutual information value is considered more important in measuring associations, a higher weight can be assigned to it.

[0074] Step S235: normalize the preliminary association strength value and map it into a preset association strength range.

[0075] The range of preliminary association strength values ​​may be relatively broad. To facilitate subsequent processing and comparison, they need to be normalized. Normalization converts preliminary association strength values ​​to a preset range, such as the interval [0, 1]. Linear normalization can be used for normalization. Normalization provides a uniform scale and range for association strength values ​​between different nodes, making comparison and analysis easier. Furthermore, normalized association strength values ​​better meet the requirements for subsequent feature association network construction and community discovery.

[0076] Step S236: traverse all feature nodes and calculate the correlation strength between all cross-level feature nodes.

[0077] In order to fully understand the association relationship between nodes at different levels in the multi-dimensional feature space, it is necessary to traverse all feature nodes and calculate the association strength between each pair of cross-level feature nodes.

[0078] The traversal process is a process of processing each node one by one, starting from the first node of the material basic feature layer, calculating the association strength with each node of the construction association feature layer and the shared demand feature layer in turn, and then processing the next node of the material basic feature layer until all nodes are traversed.

[0079] During the calculation process, according to the method described above, the source nodes and target nodes are first extracted, the mutual information value and conditional probability between them are calculated, weighted fusion processing is performed, and then normalization processing is performed to obtain the correlation strength between each pair of cross-level feature nodes.

[0080] Step S237: All calculated association strengths are stored in an association strength matrix, where rows of the association strength matrix represent source nodes, columns represent target nodes, and matrix elements represent corresponding association strength values.

[0081] The association strength matrix is ​​a structured data format used to store and represent the association strengths between nodes at different levels. By storing all calculated association strengths in the association strength matrix, these association strengths can be easily managed, analyzed, and used. The rows of the matrix correspond to source nodes, and the columns correspond to target nodes. Each element in the matrix represents the association strength between the source and target nodes. This storage method makes the representation of association strengths more intuitive and clear, facilitating the subsequent construction and analysis of feature association networks.

[0082] Step S240: Generate cross-level feature association edges based on the association strength to generate a preliminary feature association network.

[0083] A feature-association edge connects nodes at different levels, indicating an association between them. The strength of the association determines the weight of this edge. Based on the previously calculated association strength matrix, when the association strength exceeds a preset threshold, a feature-association edge is generated between the corresponding source and target nodes, and the association strength serves as the edge weight.

[0084] The preliminary feature association network, consisting of graph nodes and feature association edges, intuitively displays the associations between features at different levels. By generating this preliminary feature association network, discrete nodes in the multidimensional feature space and their associations can be graphically presented, providing a foundation for subsequent community discovery and association graph generation.

[0085] For example, for the aforementioned association strength matrix, if the preset threshold is 0.4, then feature association edges will be generated between nodes corresponding to elements in the association strength matrix greater than 0.4. In this example, feature association edges will be generated between A1 and B1, A2 and B1, A2 and B2, and A3 and B1, with edge weights of 0.7, 0.5, 0.6, and 0.8, respectively. This forms a preliminary feature association network, in which nodes A1, A2, and A3 are connected to nodes B1 and B2 via feature association edges. The thickness or color of the edges can be visualized based on the strength of the association.

[0086] Step S250: performing community discovery processing on the preliminary feature association network to identify feature node clusters that meet the association requirements.

[0087] Community discovery is a method used to analyze network structure. Its goal is to divide a preliminary feature-correlation network into distinct communities, where nodes within each community have high correlations, while nodes between different communities have relatively low correlations. Community discovery can identify clusters of characteristic nodes that meet the required correlation requirements. Nodes within these clusters share similarities or correlations in function or properties, forming relatively independent subnetworks.

[0088] As an implementation manner, step S250 may be specifically implemented as the following steps S251-S256:

[0089] Step S251: performing threshold filtering on the association strength in the preliminary feature association network, retaining feature association edges whose association strength values ​​exceed a preset association threshold, and generating a simplified feature association network.

[0090] The initial feature association network may contain some feature association edges with weak association strengths, which have little impact on the overall structure of the network and the associations between nodes. By threshold filtering the association strengths, setting a preset association threshold and retaining only feature association edges with association strengths exceeding the threshold, we can remove unimportant edges, reduce network complexity, and generate a more concise and effective streamlined feature association network.

[0091] Step S252: A community discovery algorithm based on modularity optimization is used to divide the simplified feature association network into communities. The community discovery algorithm takes maximizing network modularity as the optimization goal and achieves the optimal division of the network community structure by iteratively adjusting the communities to which nodes belong.

[0092] A commonly used community partitioning algorithm is the modularity-based community discovery algorithm. Modularity is a metric used to measure the quality of network community partitioning. It represents the difference between the connection strength between nodes within a community and the expected connection strength under random connections. The goal of a community discovery algorithm is to maximize the modularity of the network by iteratively adjusting the communities to which nodes belong, thereby achieving optimal partitioning of the network's community structure.

[0093] During the iteration process, the algorithm attempts to move nodes from one community to another and calculates the change in network modularity after each move. If the modularity increases after the move, the move is retained; otherwise, the node is restored to its original community. This continues until modularity stops increasing or the preset number of iterations is reached.

[0094] For example, in a simplified feature association network, each node initially belongs to a separate community. The algorithm begins iterating, attempting to move nodes between different communities. Suppose that moving node A from community C1 to community C2 increases the network modularity from 0.3 to 0.4. This move is retained. Iterations continue, constantly adjusting the node's community affiliation, ultimately yielding a community partition with the highest modularity.

[0095] Step S253: Calculate the internal association density and external association sparsity of each initial community. The internal association density is the average value of the association strength between nodes in the community, and the external association sparsity is the average value of the association strength between nodes in the community and nodes outside the community.

[0096] Internal connection density reflects the closeness between nodes within a community. It is calculated by summing the connection strengths between all nodes within the community and dividing it by the number of node pairs within the community. External connection sparsity indicates the degree of connection between nodes within the community and nodes outside the community. It is calculated by summing the connection strengths between nodes within the community and nodes outside the community and dividing it by the number of node pairs within the community and nodes outside the community.

[0097] Calculating internal connection density and external connection sparsity helps further evaluate the structure and stability of the community. A good community should have high internal connection density and low external connection sparsity, that is, the nodes within the community are closely connected, but have relatively few connections with nodes outside the community.

[0098] Step S254: setting a community merging threshold and a community splitting threshold according to the internal association density and the external association sparsity, splitting the initial communities whose internal association density is lower than the community splitting threshold, and merging adjacent initial communities whose external association sparsity is lower than the community merging threshold.

[0099] The community merging threshold and community splitting threshold are set based on internal connection density and external connection sparsity to determine whether communities need to be merged or split. If the internal connection density of an initial community is lower than the community splitting threshold, it indicates that the connections between nodes in the community are not close enough. It may be necessary to split the community into multiple smaller communities to improve the connection within the community. If the external connection sparsity of adjacent initial communities is lower than the community merging threshold, it indicates that the connection between the two communities is relatively close. Merging them into a larger community can better reflect the connection between the nodes.

[0100] Step S255: extracting core nodes from the processed community structure. Core nodes are a preset number of nodes with the highest total association strength within the community. The subnetwork containing the core nodes and the nodes directly connected to them is determined as a characteristic node cluster that meets the association requirement.

[0101] Core nodes hold a crucial position in a community, possessing a high sum of connection strengths with other nodes within the community. They play a crucial role in the community's structure and function. By extracting the core nodes from the processed community structure, a preset number of nodes with the highest sum of connection strengths within the community are selected as core nodes. This preset number can be adjusted based on actual circumstances and needs. For example, the top three nodes with the highest sum of connection strengths can be selected as core nodes.

[0102] The sub-network containing the core node and the nodes directly connected to it is determined as a characteristic node cluster with the correlation requirement. The nodes in these clusters have a high correlation and together constitute a relatively independent and closely related sub-network.

[0103] Step S256: Calculate the cluster association strength of each feature node cluster. The cluster association strength is the ratio of the sum of the association strengths of all feature association edges in the cluster to the number of nodes in the cluster. Retain feature node clusters whose cluster association strength exceeds a preset cluster threshold.

[0104] Cluster correlation strength is a metric that measures the closeness of connections within a feature node cluster. It is calculated by summing the correlation strengths of all feature-related edges within a cluster and dividing it by the number of nodes in the cluster. By calculating the cluster correlation strength for each feature node cluster, we can assess the quality and effectiveness of each cluster.

[0105] The preset cluster threshold is a standard for screening feature node clusters. Only feature node clusters whose cluster correlation strength exceeds the threshold will be retained, while clusters below the threshold may be considered to have insufficient internal correlation and do not meet the correlation requirements and will be removed.

[0106] Step S260: assigning an association weight to each feature node cluster, where the association weight is determined based on the sum of the association strengths of the nodes in the cluster.

[0107] The association weight reflects the importance and influence of each feature node cluster within the entire feature association graph. Since the sum of the association strengths of nodes within a cluster reflects the closeness and interaction strength between nodes within the cluster, it is reasonable to determine the association weight based on this. A greater sum of association strengths indicates a closer connection between nodes within the cluster, potentially greater impact on the overall system, and accordingly, a higher association weight is assigned. Conversely, a cluster with a smaller sum of association strengths is assigned a lower association weight.

[0108] Specifically, we first calculate the sum of the association strengths between all nodes within each feature node cluster. For example, if a feature node cluster contains five nodes, the association strength between node A and node B is 0.6, the association strength between node A and node C is 0.7, and so on. The sum of the association strengths between all nodes is then added to obtain the total association strength for the cluster. Then, based on the relative size of the sum of the association strengths across all feature node clusters, we assign an association weight to each cluster. We can use a normalization method to divide the sum of the association strengths of each cluster by the sum of the association strengths of all clusters to obtain the association weight for each cluster.

[0109] Step S270: Integrate and process the feature node clusters, feature association edges, and association weights to generate a material construction sharing association graph including inter-level association paths and association strengths.

[0110] The material construction shared association map is a comprehensive graphical representation that integrates the feature node clusters, feature association edges, and association weights assigned to each cluster obtained in the previous processing, and comprehensively displays the association relationship between the features of production and construction materials at different levels, including the association path and association strength between levels.

[0111] During the integration process, the position and role of each feature node cluster in the graph must be clearly defined, and they must be displayed as essential components of the graph. Feature association edges connect different feature node clusters, reflecting the relationships between them. Edge thickness and color can be used to visualize the strength of the association, providing an intuitive understanding of the closeness of the connections between clusters. Association weights assign a measure of importance to each feature node cluster, and are reflected in the graph through node size, color depth, and other methods.

[0112] For example, in a data sharing scenario for construction materials, the generated material construction sharing association graph may contain multiple feature node clusters, such as clusters representing material physical properties and clusters representing construction process adaptability. These clusters are interconnected by feature association edges, and the thickness of the edges reflects the strength of the association between them. Each cluster is presented in the graph as a node of different sizes based on its association weight. The association path between the layers clearly shows the association relationship from the material basic feature layer to the construction association feature layer and then to the shared demand feature layer, allowing users to understand at a glance the flow and association of production and construction material data between different layers.

[0113] Step S300: Establish a sharing demand prediction model based on the material construction sharing association map, and output the material data sharing demand sequence of different construction stages.

[0114] The material construction sharing association map contains rich association information about the characteristics of different levels of production and construction materials. Establishing a sharing demand prediction model based on this information can make full use of the inherent correlation between material data and more accurately predict the sharing demand for material data in different construction stages.

[0115] The goal of the sharing demand prediction model is to predict the characteristics and intensity of material data sharing required at different construction stages based on the material construction sharing association map and related construction schedule data. The output material data sharing demand sequence lists the material data sharing demand at each stage in the order of construction stages, including information such as the urgency of the demand, data completeness requirements, and the frequency of sharing.

[0116] As an implementation manner, step S300 may be specifically implemented as the following steps S310-S370:

[0117] Step S310: Input the material construction sharing association graph into the graph embedding model, and convert the nodes and associated edges in the graph into target dimension vector representation.

[0118] The graph embedding model is a model for processing graph-structured data. Its purpose is to convert complex graph structures, such as the material construction shared association graph, into a vector representation that computers can process. By mapping the nodes and associated edges in the graph into a vector space of the target dimension, the graph's topology and the associations between nodes can be encoded into vector form, facilitating subsequent machine learning and data analysis.

[0119] Graph embedding models are typically built using deep learning algorithms, such as graph neural networks (GNNs). During this process, the model learns the characteristic representations of nodes and edges in the graph, placing nodes with similar structures and relationships closer together in the vector space. The dimensionality of the target dimension vector can be adjusted based on specific needs and model design. Generally speaking, higher dimensionality allows for richer information to be represented, but also increases computational complexity.

[0120] Step S320: extract key construction nodes and time node information from the construction schedule data and generate a construction phase division result.

[0121] Construction schedule data details the timing and task allocation for the entire construction project. Key construction nodes are specific construction tasks or events that have a significant impact on construction progress and quality, while time nodes clearly define the start and end times of these key construction nodes. Extracting key construction node and time node information requires careful analysis and organization of the construction schedule data. Relevant information can be extracted from construction schedule documents or databases using techniques such as text parsing and data mining. For example, in the construction schedule of a bridge construction project, key construction nodes may include foundation pouring, pier construction, and beam erection, each with a corresponding start and end time.

[0122] Based on the extracted key construction nodes and time nodes, the entire construction process can be divided into different construction phases. The principles for this division can be determined based on factors such as the nature of the construction task and the construction sequence. For example, the foundation construction phase can be defined as the period from the start of foundation pouring to its completion, while the main structure construction phase can be defined as the period from the start of pier construction to the completion of beam erection. The generated construction phase division results clearly define the start and end times of each construction phase and the key construction nodes involved, providing a time and task framework for subsequent shared demand forecasting.

[0123] Step S330: Based on the target dimension vector representation and the construction stage division results, the input features of the shared demand prediction model are generated; the shared demand prediction model includes a time series prediction module and an association reasoning module. The time series prediction module is used to capture the time evolution law of shared demand, and the association reasoning module is used to capture the association relationship between material construction sharing.

[0124] The target dimension vector representation is obtained by transforming the material construction shared association graph using a graph embedding model. It contains information about the different characteristics of the material data. The construction phase division results provide a time and task framework for the construction process. Combining these two pieces of information can generate input features for the shared demand prediction model.

[0125] Input features can include node vectors, associated edge vectors, and time information corresponding to different construction phases. For example, for each construction phase, the target dimension vectors corresponding to the key construction nodes involved in that phase can be combined, and the time coding information of that phase can be added to form a comprehensive input vector.

[0126] The time series prediction module and the associative reasoning module of the sharing demand forecasting model have different functions. The time series prediction module, typically based on a recurrent neural network (RNN) or its variants, such as a long short-term memory network (LSTM), analyzes the temporal evolution of sharing demand. It learns past sharing demand patterns and uses these patterns to predict future basic sharing demand for different construction phases. The associative reasoning module, based on techniques such as the graph attention mechanism, analyzes the relationships between material and construction sharing. It considers the strength and paths of connections between different nodes and generates associative sharing demand that takes these relationships into account.

[0127] Step S340: Input the input features into the time series prediction module, learn the historical sharing demand data through the long short-term memory network, and predict the basic sharing demand in different construction stages during the target time period.

[0128] Long Short-Term Memory (LSTM) networks are a special type of recurrent neural network that effectively handles long-term dependencies in sequential data. In shared demand forecasting, historical shared demand data is a time series that includes information about material data sharing during different construction phases, such as frequency, data volume, and demand urgency.

[0129] After input features are fed into the time series prediction module, the LSTM learns from historical sharing demand data. Through its internal memory cells and gating mechanism, it retains past information and predicts future sharing demand based on current input and past memory. Specifically, the LSTM processes the input at each time step, updates its internal memory state, and outputs a predicted value. This process iterates across the entire historical sharing demand data sequence, continuously adjusting the model parameters to minimize the error between the predicted and actual values.

[0130] Step S350: Input the input features into the association reasoning module, reason on the material construction sharing association graph based on the graph attention mechanism, and generate association sharing requirements that take into account the association relationship.

[0131] The graph attention mechanism is designed for processing graph-structured data. It automatically assigns attention weights based on the strength of associations between nodes, effectively capturing the relationships between shared materials and construction projects. In the association inference module, the input features include a vector representation of the shared material and construction association graph, as well as information about the construction phase.

[0132] As an implementation manner, step S350 may be specifically implemented as the following steps S351-S358:

[0133] Step S351: Input the shared material construction association graph and the target dimension vector representation into the association reasoning module's graph attention layer, initializing the attention weight parameters for each graph node. The graph attention layer is a core component of the association reasoning module, responsible for processing the nodes and relationships in the shared material construction association graph. After inputting the shared material construction association graph and the target dimension vector representation converted from the graph embedding model into the graph attention layer, it is necessary to initialize the attention weight parameters for each graph node.

[0134] Attention weight parameters are used to measure the importance of each node in the associative reasoning process. These parameters can be initialized using random initialization methods, such as using a Gaussian distribution to randomly generate a set of initial values. These initial values ​​are continuously adjusted during subsequent training, allowing the model to automatically assign appropriate attention weights based on the relationships between nodes. For example, for a material construction sharing association graph containing multiple nodes, each node is assigned an initial attention weight vector. The dimension of the vector can be set according to the model design, such as 64 dimensions. These initial weight vectors play a critical role in the calculation of the graph attention layer, influencing the subsequent aggregation and reasoning of node features.

[0135] Step S352: For each graph node, collect the feature vectors and corresponding association strength values ​​of its adjacent nodes to generate an adjacent feature set. In the graph attention mechanism, the feature update of each graph node depends not only on its own features but also on the influence of its adjacent nodes. Therefore, it is necessary to collect the feature vectors and corresponding association strength values ​​of its adjacent nodes for each graph node.

[0136] Adjacent nodes are nodes directly connected to the current node via associated edges. For each graph node, we traverse all of its adjacent nodes, obtain their target dimension vector representations, and record the association strength value between each adjacent node. Combining the feature vectors and association strength values ​​of these adjacent nodes generates the node's adjacent feature set.

[0137] Step S353: The adjacent node feature vectors in the adjacent feature set are weighted using an attention calculation function. The attention calculation function comprehensively considers node feature similarity, association strength, and node degree centrality. The attention calculation function assigns an attention weight to each adjacent node based on the adjacent node's feature vector, association strength, and node degree centrality.

[0138] Node feature similarity reflects the degree of similarity between the features of adjacent nodes and the current node. It can be measured by methods such as calculating the cosine similarity between the target dimension vectors of two nodes. The association strength value indicates the closeness of the association between the current node and the adjacent node, which directly affects the importance of the adjacent node in feature aggregation. The node degree centrality parameter indicates the importance of the node in the entire graph. The higher the degree centrality of a node, the greater its influence on other nodes.

[0139] Step S354: Perform weighted summation on the adjacent feature set based on the assigned attention weight to generate an aggregated feature vector for the graph node.

[0140] After assigning an attention weight to each adjacent node in the adjacent feature set, the feature vectors of these adjacent nodes need to be weighted and summed to generate the aggregate feature vector of the current graph node. The weighted summation process is to multiply the feature vector of each adjacent node by its corresponding attention weight and then add all the weighted feature vectors together.

[0141] Step S355: Concatenate the aggregated feature vector with the target dimension vector representation of the graph node itself, and generate an updated node feature representation through nonlinear transformation.

[0142] After obtaining the aggregated feature vector of the graph node, it needs to be concatenated with the target dimension vector representation of the node itself to fully integrate the node's own features and the information of adjacent nodes.

[0143] Concatenation involves sequentially concatenating the aggregated feature vector and the node's own target dimension vector to form a longer vector. For example, if the dimension of the aggregated feature vector is 128 and the node's own target dimension vector is also 128, the concatenated vector will have a dimension of 256. The concatenated vector requires further processing through nonlinear transformations to introduce nonlinear capabilities into the model and enhance its expressiveness. Nonlinear transformations can use common activation functions, such as the ReLU (Rectified Linear Unit) function. The ReLU function is defined as f(x) = max(0, x) and can set negative values ​​in the concatenated vector to 0 while retaining positive values, thereby introducing nonlinear characteristics.

[0144] Step S356: Repeat the adjacent feature collection, attention weight allocation, feature aggregation and node feature update steps until the preset number of iterations is reached or the node feature representation converges.

[0145] To fully learn the relationships within the shared material construction association graph, multiple iterations are required to perform the steps of neighbor feature collection, attention weight assignment, feature aggregation, and node feature update. In each iteration, each graph node updates its feature representation based on the information of its neighboring nodes. As the number of iterations increases, the node feature representation gradually converges, meaning it no longer changes significantly. The preset number of iterations is a pre-set upper limit; if the node feature representation converges before reaching this number of iterations, the iteration can be stopped early.

[0146] Convergence can be determined by calculating the difference between node feature representations between two consecutive iterations. For example, the sum of the Euclidean distances of all node feature representations can be calculated. If this sum is less than a preset threshold, the node feature representation is considered to have converged.

[0147] For example, assuming the preset number of iterations is 10, if the sum of the Euclidean distances of all node feature representations is less than the preset threshold at the eighth iteration, iteration can be stopped, and the node feature representation obtained at this point is the final result. Through multiple iterations, the model can more deeply learn the associations in the graph, so that the node feature representation can more accurately reflect its role and relationship in the entire graph.

[0148] Step S357: extract the feature representation of the shared demand feature layer node after iterative processing, input it into the fully connected neural network for demand intensity prediction processing, and generate preliminary associated shared demand.

[0149] After multiple iterations, the feature representations of the nodes in the shared demand feature layer have fully integrated the associated information in the graph. The feature representations of these nodes are extracted and input into a fully connected neural network for demand intensity prediction.

[0150] Each neuron in a fully connected neural network layer is connected to all neurons in the next layer. In demand intensity prediction, a fully connected neural network can learn the mapping relationship between node feature representation and demand intensity.

[0151] First, the feature representations of the nodes in the shared demand feature layer are passed as input to the input layer of a fully connected neural network. The number of neurons in the input layer matches the dimensionality of the node feature representation. Then, after processing through several hidden layers, the neurons in each hidden layer perform linear transformations and nonlinear activations on the input to extract higher-level features. Finally, the neurons in the output layer output predicted values ​​for demand intensity. For example, assuming the feature representation dimension of the nodes in the shared demand feature layer is 256, the input layer of the fully connected neural network has 256 neurons. After two hidden layers, each with 128 neurons, the output layer has 1 neuron, which outputs the predicted value for demand intensity. By training on a large amount of historical data, the fully connected neural network can learn the complex relationship between node feature representations and demand intensity, thereby generating preliminary correlated shared demands.

[0152] Step S358: Fuse the preliminary association sharing requirements with the hierarchical association path information in the material construction sharing association map to generate association sharing requirements that take into account the conductive relationship between features.

[0153] Initially associated shared demands are derived by predicting the feature representations of nodes in the shared demand feature layer using a fully connected neural network. However, this may not fully account for the inter-level transmission relationships in the material construction shared association graph. Therefore, it is necessary to fuse the initial associated shared demands with the hierarchical association path information.

[0154] Hierarchical association path information records the association paths and association strengths between features at different levels, reflecting the conductive relationships between features. For example, in a material-construction shared association graph, certain features in the material base feature layer may affect the shared requirement feature layer through the construction-related feature layer. By fusing preliminary shared requirements with hierarchical association path information, the final shared requirements can more accurately reflect the mutual influence between features.

[0155] Fusion processing can be performed using a weighted summation approach. Different weights are assigned to the preliminary association sharing requirements and the hierarchical association path information, and then they are summed according to the weights. The weight distribution can be adjusted based on experimentation or experience to ensure that the fused association sharing requirements better reflect the actual situation.

[0156] Step S360: The basic sharing requirements and the associated sharing requirements are integrated to generate a material data sharing requirement sequence for different construction stages.

[0157] Basic sharing requirements are derived by learning from historical sharing requirements data through the time series prediction module, reflecting the evolution of sharing requirements over time. Associated sharing requirements are derived by the associated reasoning module by considering the relationships within the material and construction sharing association graph, reflecting the mutual influence of features. By integrating these two requirements, we can comprehensively consider time factors and associations to generate a more accurate sequence of material data sharing requirements for different construction stages.

[0158] As an implementation manner, step S360 may be specifically implemented as the following steps S361-S366:

[0159] Step S361: extracting the demand intensity parameter and demand time distribution parameter from the basic shared demand, and extracting the associated impact factor and associated demand weight from the associated shared demand. The basic shared demand includes the demand intensity parameter and demand time distribution parameter.

[0160] The demand intensity parameter indicates the intensity of the demand for material data at different construction stages. For example, the demand intensity may be higher at certain key construction nodes. The demand time distribution parameter describes the temporal distribution of demand, such as whether the demand is concentrated in a certain time period or dispersed across multiple time periods.

[0161] An associated shared requirement consists of an associated impact factor and an associated demand weight. The associated impact factor reflects the degree to which the association between different features affects the requirement. For example, the association between a material's physical properties and its construction process may affect the demand for that material's data. The associated demand weight quantifies the impact of the association and is used to determine the importance of the associated shared requirement during the fusion process.

[0162] These parameters can be extracted by parsing the vector representations of the basic shared demands and the associated shared demands. For example, suppose the basic shared demands are a multidimensional vector, with some dimensions corresponding to demand intensity parameters and some to demand temporal distribution parameters. The associated shared demands are also a multidimensional vector, with some dimensions corresponding to associated impact factors and some to associated demand weights. By indexing and extracting specific dimensions from these vectors, the corresponding parameters can be obtained.

[0163] Step S362: Correct the demand intensity parameter based on the associated impact factor to generate a corrected demand intensity parameter.

[0164] The correlation impact factor reflects the impact of the correlation between different characteristics on demand intensity. By applying the correlation impact factor to the demand intensity parameter, the demand intensity can be corrected to make it more accurately reflect the actual demand situation.

[0165] The correction process can be performed using multiplication or addition. For example, if the correlation impact factor is greater than 1, it means that the correlation relationship will increase the demand intensity; if the correlation impact factor is less than 1, it means that the correlation relationship will weaken the demand intensity.

[0166] Step S363: adjusting the demand time distribution parameter according to the associated demand weight to generate an adjusted demand time distribution parameter.

[0167] The weight of an associated demand reflects the importance of the associated shared demand within the overall demand. It can be used to adjust the demand time distribution parameters to make the demand time distribution more consistent with the actual relationship. Adjusting the demand time distribution parameters can redistribute the time distribution of demands based on the associated demand weights. For example, if the associated demand weight is large, it indicates that the relationship has a significant impact on the timing of the demand. Therefore, it may be necessary to adjust the demand time distribution to concentrate it more closely in time periods with strong relationships.

[0168] Assume that the demand time distribution parameters are represented as a time series, with each time point corresponding to a demand value. Based on the weights of the associated demands, this time series can be weighted to adjust the demand value at each time point. This adjustment of the demand time distribution parameters better reflects the impact of the associated relationships on demand time.

[0169] Step S364: Integrate the corrected demand intensity parameter and the adjusted demand time distribution parameter to generate a preliminary shared demand sequence.

[0170] After obtaining the corrected demand intensity parameters and the adjusted demand time distribution parameters, they need to be integrated to generate a preliminary shared demand sequence.

[0171] Integration processing can combine the modified demand intensity parameters and the adjusted demand time distribution parameters according to certain rules. For example, the demand intensity parameters can be used as the demand intensity value at each time point, and matched with the time information in the adjusted demand time distribution parameters to form a sequence that contains both demand intensity and time information.

[0172] Step S365: Perform time axis alignment processing on the preliminary shared demand sequence to make the demand data of different construction stages consistent in the time dimension.

[0173] The initial shared demand sequence may be inconsistent in the time dimension due to different calculation methods for basic shared demand and associated shared demand. The purpose of timeline alignment is to unify these demand data onto a common timeline, making them comparable in time dimension.

[0174] Timeline alignment can be achieved through methods such as interpolation and sampling. For example, if the time intervals between demand data points in some construction phases of the preliminary shared demand sequence are large, while the time intervals between other phases are small, new demand data points can be inserted into the areas with large time intervals through interpolation to make the time intervals of the entire sequence uniform.

[0175] Step S366: Smoothing the aligned shared demand sequence using a sliding average algorithm, dividing the smoothed shared demand sequence into subsequences corresponding to different construction stages according to the construction stage division results, and generating the final material data shared demand sequence.

[0176] The sliding average algorithm is a method of smoothing data. It reduces data fluctuations by calculating the average value within a certain window size in the data sequence, making the sequence smoother.

[0177] For the aligned shared demand sequence, select an appropriate window size, such as three time points. Slide this window across the sequence, calculate the average demand data within each window, and use this average as the new demand value at the window's center time point. As the window slides, calculate the average value for each window in turn to obtain a smoothed shared demand sequence.

[0178] Based on the construction phase division results, the smoothed shared demand sequence is divided into subsequences corresponding to different construction phases. The construction phase division results clearly define the start and end times of each construction phase. By intercepting the smoothed shared demand sequence based on this time information, the subsequence corresponding to each construction phase can be obtained.

[0179] Step S370: Smoothing the material data sharing demand sequence to obtain the final sharing demand prediction result.

[0180] Although the shared demand sequence has already been smoothed in step S366, it is necessary to smooth the material data shared demand sequence again to further reduce noise and fluctuations in the data and improve the stability and reliability of the forecast results. This can be done using a sliding average algorithm similar to that used in step S366, or other smoothing methods such as exponential smoothing. Exponential smoothing is a weighted average method that assigns a higher weight to recent data and a lower weight to long-term data, allowing it to more promptly reflect data trends.

[0181] Step S400: Generate a material data sharing task list including priority sorting according to the material data sharing requirement sequence.

[0182] The material data sharing requirements sequence details the material data sharing needs at different construction stages. Based on this information, a prioritized material data sharing task list can be generated. Prioritization helps rationally arrange material data sharing tasks, ensuring that material data sharing at critical construction nodes and urgent needs is prioritized within limited resources.

[0183] As an implementation manner, step S400 may be specifically implemented as the following steps S410-S460:

[0184] Step S410: parsing the material data sharing requirement sequence, extracting the requirement characteristics of each sharing requirement, wherein the requirement characteristics include the urgency of the requirement, the importance of the data, and the sharing scope.

[0185] The material data sharing requirement sequence is a sequence of multiple sharing requirements, each corresponding to a construction phase and material data type. Parsing this sequence requires analyzing each element and extracting the requirement characteristics related to the sharing requirement.

[0186] The urgency of the demand reflects the time urgency of the sharing demand. For example, when certain key construction nodes are approaching, the demand for relevant material data may be very urgent. The importance of data indicates the importance of the material data to the construction process. Some key material performance data may be more important than ordinary material specification data. The sharing scope specifies the objects and scope of the material data that need to be shared.

[0187] Parsing a material data sharing requirement sequence can be done by analyzing and extracting the data structure within the sequence. For example, if the sequence is stored in a list format, with each element being an object containing multiple attributes, then information such as the urgency of the requirement, the importance of the data, and the scope of sharing can be obtained by accessing the object's attributes.

[0188] Step S420: Generate priority evaluation indicators based on demand urgency, data importance and sharing scope, and determine the weight coefficient of each evaluation indicator through hierarchical analysis.

[0189] To prioritize each sharing requirement, a priority evaluation metric needs to be generated. This metric should take into account factors such as the urgency of the requirement, the importance of the data, and the scope of sharing. This can be generated by building a mathematical model.

[0190] Hierarchical analysis is a method for determining weight coefficients. It compares and judges the relative importance of different factors, constructs a judgment matrix, and then solves for the matrix's eigenvectors to obtain the weight coefficients. For example, through expert evaluation or historical data statistics, demand urgency, data importance, and sharing scope are compared pairwise to determine their relative importance. Assuming that demand urgency is slightly more important than data importance, and data importance is more important than sharing scope, a judgment matrix is ​​constructed according to the rules of hierarchical analysis. The eigenvector corresponding to the largest eigenvalue of the matrix is ​​solved. The resulting vector elements, after normalization, are the weight coefficients for each evaluation indicator.

[0191] Step S430: Calculate the comprehensive priority score of each shared requirement based on the weight coefficient and the requirement characteristics.

[0192] After obtaining the weight coefficients of each evaluation indicator and the demand characteristics of each shared demand, the comprehensive priority score of each shared demand can be calculated according to the priority evaluation indicator formula generated previously.

[0193] Step S440: Sorting the shared requirements to generate a priority sorting result, and determining the planned execution time of each shared requirement based on the priority sorting result and the construction schedule data.

[0194] Shared demands are sorted according to the comprehensive priority score of each shared demand. Common sorting algorithms such as bubble sort and quick sort can be used for sorting, and the shared demands are arranged in descending order according to the comprehensive priority score to generate a priority sorting result. When determining the planned execution time of each shared demand, it is necessary to combine the construction schedule data. The construction schedule data clarifies the time arrangement and key nodes of each construction task. According to the priority sorting results, shared demands with higher priorities are arranged to be executed at the appropriate time. For example, for the shared demand with the highest priority, if the corresponding construction task is about to start, then the shared demand can be arranged to be completed before the construction task starts; for shared demands with lower priority, they can be arranged according to resource conditions and idle time of the construction progress.

[0195] Step S450: Integrate the sharing requirements, comprehensive priority scores, and planned execution time to generate a priority-sorted material data sharing task list.

[0196] The shared requirements, overall priority scores, and planned execution times obtained from the previous processing are integrated to form a complete material data sharing task list. This integration process can organize and store this information in a specific format. For example, a table or database can be used to record the detailed information of each shared requirement, including the required material data type, urgency, data importance, sharing scope, overall priority score, and planned execution time.

[0197] Step S460: Conflict detection is performed on the material data sharing task list to identify shared tasks with resource conflicts or time conflicts, and adjustments and optimizations are performed.

[0198] Shared tasks in the material data sharing task list may conflict due to limited resources or inappropriate time scheduling. Resource conflicts occur when multiple shared tasks require the same resources, such as computing resources, storage resources, or network bandwidth, resulting in insufficient resources. Time conflicts occur when the scheduled execution times of multiple shared tasks overlap, preventing them from being executed simultaneously.

[0199] Conflict detection can be achieved by analyzing and comparing the resource requirements and planned execution times in the material data sharing task list. For example, the resource requirements of each shared task are checked, and the total demand for various resources at the same time is calculated. If the total demand exceeds the upper limit of available resources, a resource conflict exists. The planned execution times are also checked, and if the time ranges of multiple shared tasks overlap, a time conflict exists.

[0200] Once conflicting shared tasks are identified, adjustments and optimizations are necessary. Resource conflicts can be resolved by adjusting the execution order of shared tasks, allocating more resources, or optimizing resource utilization. Time conflicts can be resolved by rescheduling shared tasks to avoid overlap.

[0201] Step S500: Map the material data sharing task list to a preset cloud shared resource pool, and generate a material data sharing strategy including resource scheduling parameters.

[0202] The preset cloud shared resource pool is a cloud computing platform that provides computing resources, storage resources, network bandwidth and other resources. The purpose of mapping the material data sharing task list to the cloud shared resource pool is to allocate appropriate resources to each shared task to ensure that the task can be executed efficiently.

[0203] As an implementation manner, step S500 may be specifically implemented as the following steps S510-S560:

[0204] Step S510: parsing the material data sharing task list, extracting resource requirement parameters of each sharing task, where the resource requirement parameters include computing resource requirements, storage resource requirements, and network bandwidth requirements.

[0205] The material data sharing task list records the relevant information of each sharing task in detail, which needs to be parsed to extract the resource requirement parameters of each sharing task.

[0206] Computing resource requirements refer to the CPU computing power required during the execution of shared tasks. For example, some complex data analysis tasks may require higher CPU performance. Storage resource requirements refer to the storage space required by shared tasks, which is used to store material data and intermediate calculation results. Network bandwidth requirements refer to the network bandwidth required by shared tasks during data transmission. For example, a large number of data sharing tasks may require higher network bandwidth.

[0207] Parsing the material data sharing task list can be done by analyzing and extracting the data structure in the list. For example, if the list is stored in a table format, the computing resource requirements, storage resource requirements, and network bandwidth requirements of each sharing task can be obtained by accessing the corresponding columns of the table.

[0208] Step S520: Obtain resource status information of a preset cloud shared resource pool, where the resource status information includes available computing resources, available storage resources, and available network bandwidth.

[0209] The preset cloud shared resource pool is a dynamic resource environment, and its resource status will change with time and task execution. Therefore, it is necessary to obtain the resource status information of the cloud shared resource pool in real time in order to allocate appropriate resources to shared tasks.

[0210] Resource status information can be obtained through the cloud shared resource pool's management interface or monitoring system. The management interface provides information about the current resource usage of the cloud shared resource pool, including the number of CPU cores, available storage capacity, and available network bandwidth. The monitoring system monitors resource usage in real time and feeds this information back to the resource scheduling module.

[0211] Step S530: matching the resource requirement parameters with the resource status information to determine a candidate resource combination that meets the resource requirement of each shared task.

[0212] The resource requirement parameters of each shared task are matched against the resource status information of the cloud shared resource pool to identify all possible resource combinations that can meet the shared task's resource requirements. This matching process is achieved by comparing the resource requirement parameters and resource status information. This matching process is repeated for each shared task to identify all candidate resource combinations that meet its resource requirements. Multiple candidate resource combinations may exist, and further selection is required based on factors such as resource utilization and shared task priority.

[0213] Step S540: Based on resource utilization and shared task priority, the candidate resource combinations are optimized and selected through a resource scheduling algorithm to determine the optimal resource allocation strategy.

[0214] After obtaining the candidate resource combinations for each shared task, these combinations need to be optimized and selected based on resource utilization and shared task priority to determine the optimal resource allocation strategy.

[0215] Resource utilization refers to the degree to which resources in a cloud's shared resource pool are effectively utilized. Improving resource utilization can reduce resource costs and improve system efficiency. The priority of shared tasks reflects the importance and urgency of each shared task. Shared tasks with higher priorities should be allocated resources first. Resource scheduling algorithms can employ a variety of methods, such as greedy algorithms and genetic algorithms. Greedy algorithms prioritize the optimal choice at each step. Based on resource utilization and shared task priorities, they select the optimal resource combination for each shared task, maximizing overall resource utilization while prioritizing shared tasks.

[0216] Step S550: Generate resource scheduling parameters including resource allocation ratio, resource allocation time and resource release time according to the optimal resource allocation strategy. After determining the optimal resource allocation strategy, it is necessary to generate detailed resource scheduling parameters according to the strategy.

[0217] Resource scheduling parameters include information such as resource allocation ratio, resource allocation time, and resource release time. These parameters specify when and in what ratio each shared task acquires and releases resources from the cloud shared resource pool. The resource allocation ratio indicates the relative amount of various resources allocated to each shared task. For example, for a shared task, the proportion of computing resources allocated to it relative to the available computing resources in the cloud shared resource pool, or the proportion of storage resources allocated to it relative to the available storage resources, etc. The resource allocation time is the time when a shared task begins using resources, and the resource release time is the time when resources are released after the shared task completes.

[0218] Based on the resource allocation plan in the optimal resource allocation strategy, the resource allocation ratio of each shared task can be calculated. Combined with the planned execution time of the shared task, the resource allocation time and resource release time are determined.

[0219] Step S560: Associating the resource scheduling parameters with the material data sharing task list to generate a material data sharing strategy.

[0220] The resource scheduling parameters generated previously are associated with the material data sharing task list to form a complete material data sharing strategy. The association process can add information such as the resource allocation ratio, resource allocation time, and resource release time in the resource scheduling parameters to the material data sharing task list, so that each shared task has clear resource allocation and time arrangements. For example, in the material data sharing task list, columns such as "resource allocation ratio", "resource allocation time", and "resource release time" are added to each shared task, and the corresponding resource scheduling parameters are filled in these columns. In this way, the generated material data sharing strategy records in detail the resource requirements, priority ranking, planned execution time, and specific information on resource allocation and release of each shared task, providing comprehensive and detailed guidance for the sharing of production and construction material data, ensuring that material data can be efficiently shared at the right time and with the right resources, and promoting the smooth progress of construction projects.

[0221] It is understandable that the various algorithms involved in the above-mentioned introductions of the embodiments of the present invention can be learned from the relevant content in the prior art. In order to save space, they will not be expanded too much in the embodiments of the present invention. In addition, when implementing the scheme of the present invention, those skilled in the art can supplement the details according to the common knowledge in the field. For example, according to the common knowledge in the field, normalization can be used to eliminate dimensional conflicts before feature fusion, interpolation can be used to eliminate dimensional differences, and thresholds can be reasonably set based on historical data, experience or business scenario requirements. The model can be trained based on a general model training method, and the number of layers in the model structure can be set based on actual needs, the activation function can be selected, etc. The present invention will no longer provide redundant introductions to the overly detailed implementation process.

[0222] Figure 2 A hardware entity diagram of a computer system provided by an embodiment of the present invention is as follows Figure 2 As shown, the hardware entity of the computer system 1000 includes: a processor 1001 and a memory 1002, wherein the memory 1002 stores a computer program that can be run on the processor 1001, and when the processor 1001 executes the program, the steps in the method of any of the above embodiments are implemented.

[0223] The memory 1002 stores computer programs that can be run on the processor. The memory 1002 is configured to store instructions and applications executable by the processor 1001. It can also cache data to be processed or processed by the processor 1001 and various modules in the computer system 1000 (for example, image data, audio data, voice communication data, and video communication data). This can be achieved through flash memory (FLASH) or random access memory (RAM).

[0224] When the processor 1001 executes the program, the steps of any of the above-mentioned cloud computing-based production and construction material data sharing methods are implemented. The processor 1001 generally controls the overall operation of the computer system 1000.

[0225] The above description is only an embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for sharing production and construction material data based on cloud computing, characterized in that: include: Generate a multi-dimensional feature space for production and construction materials, wherein the multi-dimensional feature space includes a material basic feature layer, a construction-related feature layer, and a shared demand feature layer; Perform cross-level feature correlation processing on the multi-dimensional feature space through a cloud computing platform to generate a material construction shared correlation map; Establishing a sharing demand prediction model based on the material construction sharing association map to output the material data sharing demand sequence at different construction stages; generating a material data sharing task list including priority sorting according to the material data sharing requirement sequence; Mapping the material data sharing task list to a preset cloud shared resource pool to generate a material data sharing strategy including resource scheduling parameters; The step of generating a multi-dimensional feature space for production and construction materials includes: Collecting full material attribute data, including material property data, production process data, and quality inspection data; Performing structured feature extraction on the material property data to generate a material basic feature layer including a physical property vector, a chemical composition vector, and a mechanical property vector; Performing construction scenario association processing on the production process data and the quality inspection data, and combining them with a construction process standard library to generate a construction-related feature layer including a process adaptability vector, a construction environment sensitivity vector, and a quality fluctuation tolerance vector; Based on historical shared record data and current construction plan data, the demand intensity of material data at different construction nodes is calculated through a demand forecasting algorithm, generating a shared demand feature layer that includes a demand urgency vector, a data completeness demand vector, and a sharing frequency vector. The material basic feature layer, the construction-related feature layer, and the shared demand feature layer are spatially mapped according to a feature hierarchical relationship to generate a multi-dimensional feature space with hierarchical association weights.

2. The method according to claim 1, characterized in that The structured feature extraction of the material characteristic data to generate a material basic feature layer including a physical characteristic vector, a chemical composition vector, and a mechanical property vector includes: Performing parameter conversion on the physical description information in the material property data, converting the text description into quantified physical property parameters, wherein the physical property parameters include morphological characteristic parameters, density characteristic parameters, and thermal characteristic parameters; and normalizing the physical property parameters to eliminate dimensional differences and generate physical property vectors. Extracting elemental features from the component description information in the material characteristic data, identifying the composition ratios of the main components and trace components, and generating chemical composition parameters; comparing the chemical composition parameters with a standard component database, calculating a component similarity score, and generating a chemical composition vector based on the composition ratios; Extracting performance indicators from the mechanical test data in the material characteristic data, including strength characteristic parameters, elastic characteristic parameters, and toughness characteristic parameters; normalizing the strength characteristic parameters, the elastic characteristic parameters, and the toughness characteristic parameters to generate a mechanical property vector; splicing the physical property vector, the chemical composition vector, and the mechanical property vector in a preset characteristic dimension order to generate a material basic characteristic layer; The production process data and the quality inspection data are subjected to construction scenario association processing, and combined with a construction process standard library to generate a construction association feature layer including a process adaptability vector, a construction environment sensitivity vector, and a quality fluctuation tolerance vector, including: Extracting a process parameter sequence and a process control index from the production process data, wherein the process parameter sequence includes a temperature parameter, a pressure parameter, and a time parameter in the production process, and the process control index includes a process stability index and a process consistency index; Comparing the process parameter sequence with the standard process parameter range in the construction process standard library to calculate the process parameter matching degree; generating a process suitability vector according to the process control index and the matching degree of the process parameters, wherein the process suitability vector includes suitability scores of different construction processes; Extracting quality characteristic parameters and quality defect records from the quality inspection data, combining them with the construction environment factor database, and calculating the sensitivity coefficients of the quality characteristic parameters to various environmental factors through a sensitivity analysis algorithm to generate a construction environment sensitivity vector; Based on the quality defect records and construction quality requirement standards, determining the tolerance range for material quality fluctuations in different construction stages and generating a quality fluctuation tolerance vector; The process adaptability vector, the construction environment sensitivity vector, and the quality fluctuation tolerance vector are subjected to feature fusion processing to generate a construction-related feature layer.

3. The method according to claim 1, characterized in that The cross-level feature association processing of the multi-dimensional feature space is performed on the cloud computing platform to generate a material construction shared association map, including: Inputting the material basic feature layer, the construction related feature layer and the shared demand feature layer in the multi-dimensional feature space into the feature association component of the cloud computing platform; In the feature association component, a graph neural network algorithm is used to initialize nodes of features at each level, converting each feature dimension into a graph node; Calculate the association strength between nodes at different levels, where the association strength is calculated based on the mutual information value and conditional probability between features; generating cross-level feature association edges according to the association strengths to generate a preliminary feature association network; Performing community discovery processing on the preliminary feature association network to identify feature node clusters that meet the association requirements; Assigning an association weight to each feature node cluster, wherein the association weight is determined based on the sum of the association strengths of the nodes in the cluster; The characteristic node clusters, characteristic association edges and association weights are integrated and processed to generate a material construction shared association map including inter-level association paths and association strengths.

4. The method according to claim 3, characterized in that The calculation of the association strength between nodes at different levels includes: Extract any feature node in the material basic feature layer as the source node, and any feature node in the construction-related feature layer as the target node; Calculate the mutual information value between the source node and the target node, where the mutual information value is used to measure the degree of information sharing between the two feature nodes; Generate a joint probability distribution model of the source node and the target node based on historical data, and calculate the conditional probability of the target node taking a specific value under the condition that the source node takes a specific value; Performing weighted fusion processing on the mutual information value and the conditional probability to generate a preliminary association strength value; Normalizing the preliminary association strength value and mapping it to a preset association strength range; Traverse all feature nodes and calculate the correlation strength between all cross-level feature nodes; All calculated association strengths are stored in an association strength matrix, where rows of the association strength matrix represent source nodes, columns represent target nodes, and matrix elements represent corresponding association strength values.

5. The method according to claim 1, wherein The method of establishing a sharing demand prediction model based on the material construction sharing association map and outputting a material data sharing demand sequence for different construction stages includes: Inputting the material construction shared association graph into a graph embedding model, and converting the nodes and associated edges in the graph into target dimension vector representations; Extract key construction nodes and time node information from construction schedule data to generate construction phase division results; Based on the target dimension vector representation and the construction stage division result, the input features of the shared demand prediction model are generated; the shared demand prediction model includes a time series prediction module and an association reasoning module, the time series prediction module is used to capture the time evolution law of shared demand, and the association reasoning module is used to capture the association relationship between material construction sharing; Inputting the input features into the time series prediction module, learning the historical sharing demand data through the long short-term memory network, and predicting the infrastructure sharing demand at different construction stages in the target time period; Inputting the input features into the association reasoning module, reasoning on the material construction sharing association graph based on the graph attention mechanism, and generating association sharing requirements considering the association relationship; The basic sharing requirements and the associated sharing requirements are integrated to generate a material data sharing requirement sequence for different construction stages; The material data sharing demand sequence is smoothed to obtain a final sharing demand prediction result.

6. The method according to claim 5, characterized in that The basic sharing requirements and the associated sharing requirements are integrated to generate a material data sharing requirement sequence for different construction stages, including: Extracting the demand intensity parameter and the demand time distribution parameter from the basic shared demand, and extracting the associated impact factor and the associated demand weight from the associated shared demand; Correcting the demand intensity parameter based on the associated influencing factor to generate a corrected demand intensity parameter; Adjusting the demand time distribution parameter according to the associated demand weight to generate an adjusted demand time distribution parameter; Integrating the corrected demand intensity parameter and the adjusted demand time distribution parameter to generate a preliminary shared demand sequence; Performing time axis alignment processing on the preliminary shared demand sequence so that the demand data of different construction stages are consistent in the time dimension; The aligned shared demand sequence is smoothed using the sliding average algorithm. According to the construction stage division results, the smoothed shared demand sequence is divided into subsequences corresponding to different construction stages to generate the final material data shared demand sequence.

7. The method according to claim 1, characterized in that Generating a priority-ordered material data sharing task list according to the material data sharing requirement sequence includes: Parsing the material data sharing requirement sequence and extracting the requirement characteristics of each sharing requirement, wherein the requirement characteristics include the urgency of the requirement, the importance of the data, and the sharing scope; Generate priority evaluation indicators based on the urgency of the demand, the importance of the data and the sharing scope, and determine the weight coefficient of each evaluation indicator through hierarchical analysis; Calculating a comprehensive priority score for each shared requirement based on the weight coefficient and the requirement characteristics; Sorting the shared requirements to generate a priority ranking result, and determining the planned execution time of each shared requirement based on the priority ranking result and in combination with the construction schedule data; Integrate the sharing requirements, the comprehensive priority scores, and the planned execution time to generate a priority-sorted material data sharing task list; Conflict detection is performed on the material data sharing task list to identify shared tasks with resource conflicts or time conflicts, and adjustments and optimizations are performed.

8. The method according to claim 1, characterized in that Mapping the material data sharing task list to a preset cloud shared resource pool to generate a material data sharing strategy including resource scheduling parameters includes: Parsing the material data sharing task list to extract resource requirement parameters of each sharing task, wherein the resource requirement parameters include computing resource requirements, storage resource requirements, and network bandwidth requirements; Obtaining resource status information of a preset cloud shared resource pool, wherein the resource status information includes available computing resources, available storage resources, and available network bandwidth; Matching the resource requirement parameters with the resource status information to determine a candidate resource combination that meets the resource requirements of each shared task; Based on resource utilization and shared task priority, the candidate resource combinations are optimized and selected through a resource scheduling algorithm to determine the optimal resource allocation strategy; Generate resource scheduling parameters including resource allocation ratio, resource allocation time and resource release time according to the optimal resource allocation strategy; The resource scheduling parameters are associated with the material data sharing task list to generate a material data sharing strategy.

9. A computer system comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, wherein: When the processor executes the program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images